Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Drawings
The drawings are objected to as failing to comply with 37 CFR 1.84(p)(5) because they do not include the following reference sign(s) mentioned in the description: 300 (¶[025], "method 300"); and 601 (¶[031], "synthetic data 601").
The drawings are objected to as failing to comply with 37 CFR 1.84(p)(5) because they include the following reference character(s) not mentioned in the description: 404 (Fig. 4); 410 (Fig. 4); and 411 (Fig. 4).
Corrected drawing sheets in compliance with 37 CFR 1.121(d), or amendment to the specification to add the reference character(s) in the description in compliance with 37 CFR 1.121(b) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either "Replacement Sheet" or "New Sheet" pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
s
Specification
The disclosure is objected to because of the following informalities: (1) ¶[004] recites "bottle neck", which should be --bottleneck--; (2) ¶[006] recites "based on a bin edges", which should be --based on bin edges-- (or --based on a bin edge--); (3) ¶[007] recites "data rage", which should be --data range--; (4) ¶[009] lacks a terminating period after "synthetic data generation"; (5) ¶[018] recites "received though an input/output interface" (should be --through--), "The processed data is may be stored" (should be --The processed data may be stored--); (6) ¶[019] refers to element 103 as "synthetic data generator 103" while ¶[030] refers to the same element as "data generator 103"; consistent terminology should be used; (7) ¶[020] recites "configured provide training results", which should be --configured to provide training results--; (8) ¶[024] recites "in a window of a same size or adjusted based on the processing threshold second memory portion 200b", which is grammatically unclear (e.g., --in a window of the same size, or a size adjusted based on the processing threshold, of second memory portion 200b--); (9) ¶[025] describes U.S. Patent Application Ser. No. 16/151,385 as "recently allowed"; the status of the incorporated application should be updated (e.g., by patent number); (10) ¶[028] states that, at 403, the amount of stored data is monitored to determine whether it exceeds the processing threshold and that, if the threshold is met, the stored data is processed at 405, whereas Fig. 4 depicts monitoring at 403 and the threshold determination at 404, and the "Yes" branch 410 and "No" branch 411 of determination 409; the description should be conformed to the drawings; and (11) the thresholds are labeled inconsistently: ¶[005] and ¶[021] recite a "first threshold" as the "maximum available threshold" and ¶[022] recites a "second threshold" as the "processing threshold", yet ¶[028] and Fig. 4 (step 404) compare the stored data to the "processing threshold"; and ¶[031] compares the number of samples at 605 to an "accuracy threshold" while ¶[032] states that the same step 605 uses a "generator threshold".
Appropriate correction is required.
Claim Objections
Claims 21, 30, and 37 are objected to because of the following informalities: (1) claims 21, 30, and 37 each recite "determining a size of stored streaming data has reached a first threshold", which should read --determining that a size of stored streaming data has reached a first threshold--; (2) claim 37 recites "creating a plurality of bins, having respective data ranges", in which the comma after "bins" is unnecessary (as demonstrated in claims 21 and 30). Appropriate correction is required.
Claim 32 is objected to under 37 CFR 1.75(c) as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends. Claim 30, from which claim 32 depends, already recites that the processing is "iteratively performed based on a window size specified by a processing threshold"; claim 32 recites that "the processing is based on a window size specified by a processing threshold" and thus adds no further limitation. Applicant may cancel the claim, amend the claim to place the claim in proper dependent form, rewrite the claim in independent form, or present a sufficient showing that the dependent claim complies with the statutory requirements.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 29 and 39 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim 29 (depending from claim 28) and claim 39 (depending from claim 38) each recite "wherein the edge widths are configured to minimize error associated with approximating an original distribution." This limitation defines the claimed bin edge widths solely by a desired result, namely minimizing the error of approximating an original distribution. It does not recite any particular value, rule, structure, or algorithm by which the edge widths are configured to achieve that result. Under the broadest reasonable interpretation, the limitation encompasses any technique for selecting or adjusting histogram bin edge widths so as to minimize any measure of error between the binned representation and an original distribution of the stored streaming data, for any type of streaming data (e.g., numeric, image, video, or audio data; see claims 24-25).
The only disclosure addressing this limitation is a single sentence in the specification at ¶[033]: "Edge widths may define data ranges for each of the bins and may be changed to minimize the overall error of approximating the original distribution." The same paragraph adds that "[e]ach bin may have different edge widths" and that "[e]ach bin edge width may be modified independently from total variation of the data model using similar techniques e.g. manually or automatically, dynamically or at the specified points." These statements merely restate the claimed result. The specification does not identify what error is measured or how the "original distribution" is represented or estimated from the samples accumulated in the processing window. Nor does it identify any cost function to be minimized, how the edge widths are computed or iteratively adjusted to minimize that error, or at what point in the disclosed processing (Fig. 5, steps 502-506) and populating (Fig. 6, steps 602-608) operations any such minimization occurs. Fig. 5 and its description at ¶[029] disclose only that "a plurality of bins are created, with each bin having a data range assigned to it" (step 504); they are silent as to how the bin ranges or edge widths are selected, let alone selected so as to minimize approximation error. The example in ¶[033] of generating "more data around the peaks of the histogram instead of edges" concerns where synthetic data is generated, not how the edge widths are configured.
"[O]riginal claims may lack written description when the claims define the invention in functional language specifying a desired result but the specification does not sufficiently describe how the function is performed or the result is achieved. For software, this can occur when the algorithm or steps/procedure for performing the computer function are not explained at all or are not explained in sufficient detail (simply restating the function recited in the claim is not necessarily sufficient)." MPEP § 2161.01(I). Further, "[i]t is not enough that one skilled in the art could write a program to achieve the claimed function because the specification must explain how the inventor intends to achieve the claimed function to satisfy the written description requirement." Id. (citing Vasudevan Software, Inc. v. MicroStrategy, Inc., 782 F.3d 671, 681-683 (Fed. Cir. 2015)). See also Ariad Pharm., Inc. v. Eli Lilly & Co., 598 F.3d 1336, 1351 (Fed. Cir. 2010) (en banc); MPEP § 2163.
Here, the specification states the desired outcome of configuring the edge widths but does not describe any algorithm, steps, or procedure by which the inventor configured the edge widths to minimize error associated with approximating an original distribution. Accordingly, the specification fails to reasonably convey to one of ordinary skill in the art that the inventor or a joint inventor, at the time the application was filed, had possession of the full scope of the subject matter of claims 29 and 39.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 22, 26, 27, 29, 31, 32, 35, 36, and 39 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 22 and 31 recite "analyzing the plurality of bins" (line 1 of each claim). There is insufficient antecedent basis for this limitation in the claim. Parent claims 21 and 30 recite no analyzing operation, and it is unclear what operation of the parent claim is being further limited. For purposes of examination, "analyzing the plurality of bins" is interpreted under BRI as any examination of the created bins that yields a minimum and maximum value for each bin, performed as part of the processing of claim 21".
Claim 32 recites "the processing is based on a window size specified by a processing threshold" (line 1). Claim 30, from which claim 32 depends, already recites a window size specified by a processing threshold. It is unclear whether claim 32 introduces a second window size and processing threshold or refers to the same ones, so the scope of claim 32 cannot be determined. For purposes of examination, claim 32 is interpreted under BRI as referring to the same window size and processing threshold already recited in claim 30, so that claim 32 adds no further limitation.
Claims 26 and 35 recite "respective iterations" (line 2). Claim 27 inherits this language from claim 26. The parent claims recite no iterative generating of synthetic data, so the "iterations" lack antecedent basis. In claim 35, it is unclear whether the iterations are those of the iteratively performed processing of claim 30 or separate iterations of the generating step. For purposes of examination, "respective iterations" is interpreted under BRI as repeated passes of the synthetic-data generating step, in each of which the number of data points generated is specified by the generator threshold.
Claims 27 and 36 recite "pre-set or set based on at least one of determined minimum and maximum values of the assigned samples or a number of the assigned samples". No determining of minimum and maximum values is recited in the claims or in any parent claim. The "at least one of A and B or C" construction is ambiguous as to whether the minimum value and maximum value may be used separately or only together (See MPEP § 2173.05(h)). For purposes of examination, the limitation is interpreted under BRI to cover a generator threshold that is preset, or set based on any one or more of: the minimum value of the assigned samples, the maximum value of the assigned samples, or the number of the assigned samples.
Claim 36 recites "the generator threshold" (line 1). There is insufficient antecedent basis for this limitation in the claim. A generator threshold is introduced only in claim 35, from which claim 36 does not depend. For purposes of examination, "the generator threshold" is interpreted under BRI as the threshold, per specification ¶[032], that "specifies how many data points are generated at each iteration" by the synthetic data generator, i.e., the same threshold as in claim 35.
Claims 29 and 39 recite "configured to minimize error associated with approximating an original distribution" (line 2). The claims do not identify the "original distribution," state how error is measured, or state whether "minimize" requires a global minimum or only a reduction. The specification (¶[033]) provides no objective measure of the error, so the metes and bounds cannot be ascertained (See MPEP § 2173.05(b)). For purposes of examination, the limitation is interpreted under BRI as edge widths that are capable of being adjusted so as to reduce any measure of difference between the binned (histogram) representation and the distribution of the original streaming data.
Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 21-40 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims follows the 2019 Revised Patent Subject Matter Eligibility Guidance (“2019 PEG”) as incorporated into MPEP 2106, including MPEP 2106.04(d)(1) and 2106.05(a) as revised in light of Ex Parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025) (Appeals Review Panel Decision).
CLAIM 21
Step 1: Claim 21 recites “A system for synthetic data generation, comprising: at least one processor; and at least one non-transitory memory storing instructions,” which falls within the statutory category of a machine. See MPEP 2106.03. Accordingly, claim 21 satisfies Step 1 of the eligibility analysis.
Step 2A, Prong 1: Claim 21 is directed to an abstract idea. Specifically, claim 21 recites mathematical concepts and mental processes. The claim recites: “determining a size of stored streaming data has reached a first threshold”; “in response to the size determination, processing the stored streaming data, the processing comprising: creating a plurality of bins having respective data ranges; and assigning samples from the stored streaming data to the plurality of bins”; “populating the bins with synthetic data, the populating comprising: . . . assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins”; and “creating a processed dataset based on the populated bins.” Creating a plurality of bins having respective data ranges and assigning the samples to the bins sets forth, in words, the construction of a frequency distribution (i.e., a histogram): a span of numerical values is partitioned into intervals, and each sample is allocated to the interval whose range contains the sample’s value. Assigning the synthetic data points to the bins “based on values of the synthetic data points and data ranges of the bins” likewise recites the mathematical relationship between a value and the lower and upper bounds of an interval. These limitations recite mathematical relationships and calculations, i.e., organizing information and manipulating information through mathematical correlations. Creating a processed dataset based on the populated bins does not itself set forth a further calculation; it compiles the information organized by those correlations into a new form and, as explained below, is also an evaluation that a person can practically perform with pen and paper. See MPEP 2106.04(a)(2)(I); Digitech Image Techs., LLC v. Elecs. for Imaging, Inc., 758 F.3d 1344, 1350-51 (Fed. Cir. 2014) (“Without additional limitations, a process that employs mathematical algorithms to manipulate existing information to generate additional information is not patent eligible.”); SAP America, Inc. v. InvestPic, LLC, 898 F.3d 1161, 1163-65 (Fed. Cir. 2018) (statistical analysis generating a distribution). The specification confirms the mathematical character of the claimed bins, explaining that they correspond to the bins of a histogram whose edge widths may be changed “to minimize the overall error of approximating the original distribution” (¶[033]) and that the processed data is used for calculations requiring “accurate statistical characteristics such as distributions (pdf, cdf), percentiles, etc.” (¶[034]).
The same limitations also recite mental processes. Under the broadest reasonable interpretation, the claim does not require any particular volume or rate of data, and the stored streaming data encompasses a stream of single numerical values such as salary, age, or tax figures (¶[027]). At the level of generality claimed, a person can practically (i) count the values that have been collected and judge whether the count has reached a chosen threshold (an observation and evaluation; see ¶[029], determining “a total number of samples”), (ii) in response, write down a set of value ranges (e.g., salary ranges such as $0-50,000 and $50,001-100,000) and tally each collected value into the range in which it falls, (iii) tally each additional value into the range in which it falls by comparing the value to the range boundaries, and (iv) write out the resulting set of tallied values as a new dataset. These are observations, evaluations, and judgments that can practically be performed in the human mind or with pen and paper. See MPEP 2106.04(a)(2)(III); Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350, 1353-55 (Fed. Cir. 2016); CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1372-73 (Fed. Cir. 2011). That the claim performs these steps using a processor does not preclude the limitations from reciting a mental process. MPEP 2106.04(a)(2)(III)(C).
Considered together (MPEP 2106.04, subsection II(B)), these limitations recite the abstract idea of collecting data until a threshold amount is reached, organizing the collected values into value-range bins (a histogram), supplementing the bins with additional data values assigned according to their values, and compiling a processed dataset from the supplemented bins.
Step 2A, Prong 2: The judicial exception is not integrated into a practical application. The additional elements, considered individually and as an ordered combination, do not integrate the judicial exception into a practical application. The additional elements are: “at least one processor”; “at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations”; the recitation that the data operated on is “stored streaming data”; and “generating, by a synthetic data generator, a plurality of synthetic data points”.
Consistent with Ex Parte Desjardins and MPEP 2106.04(d)(1) and 2106.05(a), the specification has been evaluated to determine whether it describes an improvement in the functioning of a computer or in another technology or technical field, and the claim has been evaluated to determine whether it reflects any such improvement. The specification describes three kinds of benefits. First, it describes increasing “the accuracy of the statistical approximations” and of the data model by supplementing the collected data with synthetic data (¶¶[026], [030], [033]), and producing synthetic data having “statistical characteristics similar to those of the original data” (¶[003]). These are improvements in the accuracy of the mathematical analysis itself, i.e., improvements in the abstract idea, and “an improvement in the abstract idea itself . . . is not an improvement in technology.” MPEP 2106.05(a), subsection II; see SAP America, 898 F.3d at 1163 (an advance that “lies entirely in the realm of abstract ideas” is ineligible). Moreover, even if the statistical fidelity of the synthetic data were regarded as a technological benefit, claim 21 does not include the steps that the specification credits with providing it, such as generating synthetic values based on the minimum and maximum values determined for the bins, updating the per-bin sample counts, and pausing the generator when an accuracy threshold is met (¶¶[030]-[031]), or concentrating generated data around the peaks of the histogram (¶[033]); claim 21 requires only that whatever data points are generated be assigned to bins according to their values. See MPEP 2106.05(a). Second, it describes avoiding the regulatory burden of training artificial intelligence systems on sensitive real-world data (¶¶[002]-[003]), which is a legal and business benefit rather than a technological one. Third, it describes a memory-handling scheme for processing a continuous data stream in real time with limited memory (¶[004]), in which data accumulates in a first memory portion (“window”) 200a sized by the processing threshold, is processed while new data accumulates in a second memory portion 200b, is stored in long-term memory or sent to another system, and the first memory portion is then purged for reuse (¶¶[017], [021]-[024], [034]-[035]; Fig. 2), which the specification states “allows for efficient memory management and real time processing” (¶[024]).
Even assuming that one of ordinary skill in the art would recognize this memory-handling scheme as an improvement in computer functionality, claim 21 does not reflect it. Claim 21 does not recite allocating memory portions or windows, accumulating newly received data in a second memory portion while the stored data is processed, storing the processed dataset in a second storage device, or purging or releasing the memory occupied by the processed data (compare ¶[035]). Claim 21 merely conditions the performance of the abstract analysis on the amount of stored data reaching a threshold, and nothing in the claim requires that any memory be freed or that the stream be processed in real time. Because the claim does not include the components or steps of the invention that provide the described improvement, it does not reflect an improvement in technology. See MPEP 2106.05(a); Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1316 (Fed. Cir. 2016) (claims lacked limitations addressing the asserted improvements). Nor does the claim reflect an improvement to the synthetic data generator or to any machine-learning model. Specification at ¶[001] states that the disclosed embodiments “also concern improvements to generative adversarial network models and adversarially learned inference models,” but the present disclosure provides no technical explanation of any such improvement: it describes the generator only generically (¶¶[019]-[020], [036]) and states that the generation process is “described more completely in” a separate application incorporated by reference (¶[025]). A bare assertion of an improvement, without the detail necessary for the improvement to be apparent to one of ordinary skill in the art, does not establish that the claimed invention improves technology. MPEP 2106.04(d)(1); MPEP 2106.05(a). In any event, even if the incorporated application were found to describe an improvement in how a generative model operates, claim 21 recites the generator solely in terms of its result and includes none of the components or steps of any such improvement, so the claim does not reflect it. Compare Ex Parte Desjardins (claims reflected a disclosed improvement in how the machine learning model itself operates).
Turning to the individual additional elements, the “at least one processor” and the “at least one non-transitory memory storing instructions” executed by the processor are recited at a high level of generality and are described in the specification as generic computing components (e.g., a CPU or GPU, and RAM, a hard disk drive, or flash memory, ¶[016]). They are used merely as tools to perform the abstract idea, which amounts to mere instructions to apply the exception using a generic computer. MPEP 2106.05(f); Alice Corp. v. CLS Bank Int’l, 573 U.S. 208, 223-24 (2014). A generic computer is not a particular machine (MPEP 2106.05(b)), and manipulating data values into a processed dataset is not a transformation of a particular article to a different state or thing (MPEP 2106.05(c), citing CyberSource, 654 F.3d at 1372 n.2).
The recitation that the analyzed data is “stored streaming data” specifies the source and type of the data that is analyzed and requires only that the data has been stored (i.e., buffered) before it is analyzed. Storing data in preparation for analysis is mere data gathering, which is insignificant extra-solution activity (MPEP 2106.05(g)), and “merely selecting information, by content or source, for collection, analysis, and display does nothing significant to differentiate a process from ordinary mental processes.” Electric Power Group, 830 F.3d at 1355. Limiting the analysis to data received as a stream merely links the abstract idea to a particular technological environment (MPEP 2106.05(h)), and buffering content is a generic function performed by computer hardware. See MPEP 2106.05(a), subsection II (discussing Affinity Labs of Tex., LLC v. DirecTV, LLC, 838 F.3d 1253, 1264 (Fed. Cir. 2016)). Even if the size determination (“determining a size of stored streaming data has reached a first threshold”) were treated as an additional element rather than as part of the abstract idea, it would not integrate the exception: monitoring how much data has been buffered and initiating processing when a threshold is reached is recited at a high level of generality, without any particular technique for allocating, monitoring, or releasing memory, and amounts to mere instructions to apply the exception using a generic computer function (MPEP 2106.05(f)), i.e., the selection of a generic hardware function from among well-known functions (MPEP 2106.05(a), subsection II (discussing Affinity Labs, 838 F.3d at 1264)).
The limitation “generating, by a synthetic data generator, a plurality of synthetic data points” is recited at a high level of generality, solely in terms of its result. The claim specifies no model, algorithm, training, or constraint by which the synthetic data points are generated; it recites only the idea of obtaining additional data values, rather than a particular way of generating them, and invokes the generator merely as a tool to supply the values that are then binned by the abstract assignment step. See MPEP 2106.05(f)(1)-(2). Whether the term “synthetic data generator” is given its plain meaning (a component that generates data) or is construed under 35 U.S.C. 112(f) to cover the corresponding structure described in the specification (computing devices executing a generative model, e.g., a recurrent neural network or a generative adversarial network, ¶¶[019]-[020], [036]) and equivalents thereof, the generator is used only for its ordinary function of generating data. Claims that do no more than apply generic machine learning to a new data environment, without any improvement to the machine-learning model itself, do not render an abstract idea patent eligible. See Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025). Because the claim recites only the idea of obtaining generated data points, invokes the generator merely as a tool, and (as discussed above) reflects no improvement to the generator or to a machine-learning model, the generating limitation does not integrate the judicial exception into a practical application. MPEP 2106.05(a), (f).
Viewed as a whole and as an ordered combination, the additional elements interact with the abstract idea only by storing the data to be analyzed, supplying additional values to be binned, and executing the abstract steps on a generic computer. The claim covers the idea of supplementing a histogram of buffered data with generated values, rather than a particular technological solution to a technological problem, and therefore does not impose a meaningful limit on the judicial exception. Accordingly, claim 21 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. As discussed above, the processor and the memory storing instructions amount to no more than mere instructions to apply the exception using generic computer components, which cannot provide an inventive concept. Alice, 573 U.S. at 223-26; MPEP 2106.05(f). In addition, these components are well-understood, routine, and conventional, as evidenced by the specification, which describes them generically as “a microprocessor, such as a central processing unit (CPU), a graphics processing unit (GPU), or other electronic circuitry capable of carrying out the instructions of a computer program” and as “random-access memory (RAM), a non-volatile memory, such as a hard disk drive, a flash memory, or the like” (¶[016]; see also ¶[038]), and which states that “[c]omputer programs based on the written description and methods of this specification are within the skill of a software developer” (¶[042]). See also Alice, 573 U.S. at 225-26; Bancorp Servs., L.L.C. v. Sun Life Assurance Co. of Canada (U.S.), 687 F.3d 1266, 1278 (Fed. Cir. 2012) (computer employed only for its most basic function).
The storage of the streaming data, re-evaluated at Step 2B, is the well-understood, routine, and conventional activity of receiving data and storing it in memory. See MPEP 2106.05(d)(II) (citing Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334 (Fed. Cir. 2015) and OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363 (Fed. Cir. 2015) (storing and retrieving information in memory), and buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355 (Fed. Cir. 2014) (receiving and sending information over a network)); see also ¶[018] (describing the incoming data stream as received through an input/output interface 104 and stored in part of system memory 102). To the extent the size determination is considered an additional element, the Step 2A, Prong 2 conclusion that it amounts to mere instructions to apply the exception carries over to Step 2B (MPEP 2106.05, subsection II), and it therefore cannot supply an inventive concept.
The synthetic data generator, re-evaluated at Step 2B, likewise does not supply an inventive concept. The Step 2A, Prong 2 conclusion that the generating limitation amounts to mere instructions to apply the exception using a generic tool carries over to Step 2B. MPEP 2106.05, subsection II; MPEP 2106.05(f). In addition, the generator is well-understood, routine, and conventional. The specification describes it only generically, as “one or more computing devices configured to generate data” (¶[019]) whose computing resources can run “SageMaker, Genesys, or similar machine learning training applications” (¶[020]), i.e., commercially available products, and states that the generator “might be using a recurrent neural network, a generative adversarial network, or any other form of machine learning suitable for data generation” (¶[036]), without describing the particulars of any such model. The specification thus describes the generator as a commercially available product and in a manner indicating that it is sufficiently well-known that the specification need not describe its particulars to satisfy 35 U.S.C. 112(a). MPEP 2106.07(a), subsection III(A); Berkheimer v. HP Inc., 881 F.3d 1360 (Fed. Cir. 2018). See also I. Goodfellow, Y. Bengio & A. Courville, Deep Learning, chs. 10, 20 (MIT Press 2016) (textbook describing recurrent neural networks and generative adversarial networks as established deep-learning techniques); A. Creswell et al., “Generative Adversarial Networks: An Overview,” IEEE Signal Processing Magazine, vol. 35, no. 1, pp. 53-65 (Jan. 2018) (review article describing GAN architectures, training, and applications); MPEP 2106.07(a), subsection III(C).
Considered as an ordered combination, the additional elements add nothing that is not already present when they are considered individually: a generic computer stores incoming data, performs the abstract binning analysis, obtains additional values from a generic data generator, and compiles a dataset, with each component performing its ordinary, conventional function in a conventional arrangement. Unlike BASCOM Global Internet Servs., Inc. v. AT&T Mobility LLC, 827 F.3d 1341, 1350 (Fed. Cir. 2016), the claim recites no non-conventional and non-generic arrangement of known components. Accordingly, claim 21 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 22
Step 1: Claim 22 depends from claim 21 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 22 incorporates all the limitations of claim 21 and is therefore directed to the same abstract idea identified in claim 21. Claim 22 additionally recites “wherein analyzing the plurality of bins comprises determining minimum and maximum values for each bin.” Determining the minimum and maximum values of the samples in each bin is a mathematical calculation (identifying the extreme values of a set of numbers) and is also an observation and evaluation that a person can practically perform by inspecting the values tallied in each range and noting the smallest and largest. See MPEP 2106.04(a)(2)(I), (III); ¶[029].
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 23
Step 1: Claim 23 depends from claim 21 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 23 incorporates all the limitations of claim 21 and is therefore directed to the same abstract idea identified in claim 21. Claim 23 additionally recites “wherein the processing is based on a window size specified by a processing threshold.” The window size is “a size of the dataset to be processed” and “affects the accuracy and the overall processor load,” and the processing threshold is used “to maintain a balance between accuracy, processing speed, and cost” (¶[022]). Basing the binning analysis on an amount of data specified by a threshold value sets a numerical parameter of the mathematical analysis and reflects a judgment about how much data to analyze at a time, which a person can practically make, e.g., by deciding to tally collected values in batches of a chosen size. See MPEP 2106.04(a)(2)(I), (III).
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 24
Step 1: Claim 24 depends from claim 21 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: No additional abstract idea limitations are introduced.
Step 2A, Prong 2: Claim 24 additionally recites “wherein the stored streaming data includes at least one of image data, video data, or audio data.” This limitation specifies the type of data that is analyzed and does not alter the abstract idea: the claim recites no image-, video-, or audio-specific processing (e.g., no feature extraction, compare ¶[027]), and creating value-range bins, assigning sample values to them, and compiling a dataset from the populated bins remain mathematical concepts irrespective of the type of data from which the sample values are taken. See MPEP 2106.04(a)(2)(I). The additional elements in claim 24 are those of claim 21 plus the limitation of the stored streaming data to image data, video data, or audio data.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 21, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator do not integrate the judicial exception into a practical application.
With respect to the new additional element, limiting the stored streaming data to image, video, or audio data merely specifies the type of data to which the abstract binning is applied. This generally links the use of the judicial exception to a particular technological environment or field of use without limiting how the analysis is performed (MPEP 2106.05(h); see Electric Power Group, 830 F.3d at 1354-55 (limiting the collection and analysis of information to power-grid data)), and selecting a particular type of data to be manipulated is insignificant extra-solution activity (MPEP 2106.05(g)). The claim does not recite any improvement to image, video, or audio processing technology; it merely bins values from such data in the same manner as any other numerical data. Accordingly, claim 24 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 21, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator are well-understood, routine, and conventional and amount to no more than mere instructions to apply the exception. With respect to the limitation to image, video, or audio data, a field-of-use limitation cannot provide significantly more. MPEP 2106.05(h). Further, to the extent the limitation requires receiving and storing image, video, or audio data streams, receiving and storing such data is well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610 (Fed. Cir. 2016) (transmitting images), buySAFE, 765 F.3d at 1355, and Versata, 793 F.3d at 1334); see also ¶¶[018], [036] (describing the incoming data stream generically as including numeric, image, audio, or video data). Considered as an ordered combination with the elements of claim 21, the data-type limitation adds nothing beyond the conventional arrangement discussed for claim 21. Accordingly, claim 24 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 25
Step 1: Claim 25 depends from claim 21 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: No additional abstract idea limitations are introduced.
Step 2A, Prong 2: Claim 25 additionally recites “wherein the stored streaming data includes at least one of salary information, age information, or tax information.” This limitation specifies the content of the information that is analyzed and does not alter the abstract idea; binning salary, age, or tax values is precisely the kind of analysis that a person can practically perform with pen and paper (see ¶[027]). See MPEP 2106.04(a)(2)(I), (III). The additional elements in claim 25 are those of claim 21 plus the limitation of the stored streaming data to salary information, age information, or tax information.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 21, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator do not integrate the judicial exception into a practical application. With respect to the new additional element, limiting the analyzed data to salary, age, or tax information merely specifies the content of the information to which the abstract idea is applied, which generally links the judicial exception to a particular field of use. MPEP 2106.05(h); see Bilski v. Kappos, 561 U.S. 593, 612 (2010) (limiting an abstract idea to one field of use does not make the concept patentable). “[M]erely selecting information, by content or source, for collection, analysis, and display does nothing significant to differentiate a process from ordinary mental processes.” Electric Power Group, 830 F.3d at 1355; see MPEP 2106.05(g). Accordingly, claim 25 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 21, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator are well-understood, routine, and conventional and amount to no more than mere instructions to apply the exception. With respect to the limitation to salary, age, or tax information, a field-of-use limitation cannot provide significantly more (MPEP 2106.05(h)), and to the extent the limitation requires receiving and storing such information, receiving and storing data is well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing buySAFE, 765 F.3d at 1355, and Versata, 793 F.3d at 1334); see also ¶[027] (describing a stream of single numbers such as salary, age, or tax as the simplest form of the data stream). Considered as an ordered combination with the elements of claim 21, the data-content limitation adds nothing beyond the conventional arrangement discussed for claim 21. Accordingly, claim 25 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 26
Step 1: Claim 26 depends from claim 21 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 26 incorporates all the limitations of claim 21 and is therefore directed to the same abstract idea identified in claim 21. Claim 26 additionally recites “wherein the plurality of synthetic data points are generated based on a generator threshold specifying one or more data points that are generated at respective iterations.” The generator threshold is a number specifying how many data points are produced at each iteration (¶[032]). The limitation does not recite a step of selecting or setting the threshold; it further specifies how the synthetic data generator of claim 21, an additional element, operates, i.e., by producing data points over respective iterations in the quantity specified by the threshold. The limitation therefore does not alter the abstract idea identified in claim 21, and it is evaluated below together with the synthetic data generator.
Step 2A, Prong 2: The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 21, the additional elements do not integrate the judicial exception into a practical application. Requiring the synthetic data generator to produce a specified number of data points at each of several iterations merely repeats the generic data-generation function of claim 21 without specifying how the data points are generated; it remains mere instructions to apply the exception using a generic tool (MPEP 2106.05(f)), and it does not reflect any improvement to the generator, to a machine-learning model, or to computer functionality. Accordingly, claim 26 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 21, the additional elements, including the synthetic data generator, are well-understood, routine, and conventional. Invoking the generator repeatedly to produce a specified number of data points per iteration adds only the repetitive performance of a conventional function, which the courts have recognized as well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing Parker v. Flook, 437 U.S. 584, 594 (1978), and Bancorp, 687 F.3d at 1278 (performing repetitive calculations)); see also ¶[032] (generator threshold may be preset or set manually or automatically). Accordingly, claim 26 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 27
Step 1: Claim 27 depends from claim 26 (which depends from claim 21) and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 27 incorporates all the limitations of claims 21 and 26 and is therefore directed to the same abstract idea identified in claim 21. Claim 27 additionally recites “wherein the generator threshold is pre-set or set based on at least one of determined minimum and maximum values of the assigned samples or a number of the assigned samples.” Setting a numerical threshold in advance, or as a function of the minimum and maximum values or the number of the assigned samples, recites a mathematical relationship between the threshold and those values, and it is a judgment that a person can practically make by inspecting the tallied values and choosing how many additional values to generate; the specification confirms that the generator threshold may be “set manually” (¶[032]). See MPEP 2106.04(a)(2)(I), (III).
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 28
Step 1: Claim 28 depends from claim 21 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 28 incorporates all the limitations of claim 21 and is therefore directed to the same abstract idea identified in claim 21. Claim 28 additionally recites “wherein the plurality of bins includes edge widths defining data ranges for each of the plurality of bins.” The edge widths are the numerical boundaries and widths of the intervals that define each bin’s range, i.e., a mathematical definition of the intervals of a histogram (¶[033]), which a person can also practically set by choosing the boundaries of each range. See MPEP 2106.04(a)(2)(I), (III).
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 29
Step 1: Claim 29 depends from claim 28 (which depends from claim 21) and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 29 incorporates all the limitations of claims 21 and 28 and is therefore directed to the same abstract idea identified in claim 21. Claim 29 additionally recites “wherein the edge widths are configured to minimize error associated with approximating an original distribution.” Configuring the interval widths of a histogram to minimize the error with which the histogram approximates an underlying distribution is a mathematical optimization, i.e., a mathematical calculation and relationship expressed in words (¶[033]). See MPEP 2106.04(a)(2)(I); In re Grams, 888 F.2d 835, 837 n.1 (Fed. Cir. 1989) (words used in a claim operating on data to solve a problem can serve the same purpose as a formula); SAP America, 898 F.3d at 1163-65.
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 30
Step 1: Claim 30 recites “A system for synthetic data generation, comprising: at least one processor; and at least one non-transitory memory storing instructions,” which falls within the statutory category of a machine. See MPEP 2106.03. Accordingly, claim 30 satisfies Step 1 of the eligibility analysis.
Step 2A, Prong 1: Claim 30 is directed to an abstract idea. Specifically, claim 30 recites mathematical concepts and mental processes. The claim recites: “determining a size of stored streaming data has reached a first threshold”; “in response to the size determination, processing the stored streaming data, the processing being iteratively performed based on a window size specified by a processing threshold and comprising: creating a plurality of bins having respective data ranges; and assigning samples from the stored streaming data to the plurality of bins”; “populating the bins with synthetic data, the populating comprising: . . . assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins”; and “creating a processed dataset based on the populated bins.” For the reasons discussed with respect to claim 21, creating value-range bins and assigning the samples and the synthetic data points to the bins according to their values recite mathematical relationships and calculations, i.e., the construction and supplementation of a frequency distribution (histogram) through mathematical correlations, and creating a processed dataset from the populated bins compiles the information so organized into a new form. MPEP 2106.04(a)(2)(I); Digitech, 758 F.3d at 1350-51; SAP America, 898 F.3d at 1163-65; see ¶¶[029], [033]-[034]. These steps, together with determining that the size of the collected data has reached a threshold, are also observations, evaluations, and judgments that a person can practically perform with pen and paper at the claimed level of generality, e.g., by counting collected values against a threshold, writing down value ranges, tallying each value into its range, and writing out the resulting dataset. MPEP 2106.04(a)(2)(III); Electric Power Group, 830 F.3d at 1353-55; CyberSource, 654 F.3d at 1372-73.
Claim 30 further requires that the processing be “iteratively performed based on a window size specified by a processing threshold.” The window size is “a size of the dataset to be processed” and “affects the accuracy and the overall processor load,” and the processing threshold is used “to maintain a balance between accuracy, processing speed, and cost” (¶[022]). Performing the binning analysis repeatedly on successive subsets of data of a specified size is a repetition of the same mathematical analysis governed by a numerical parameter, which a person can also practically perform, e.g., by tallying collected values in successive batches of a chosen size. See MPEP 2106.04(a)(2)(I), (III).
Considered together (MPEP 2106.04, subsection II(B)), these limitations recite the abstract idea of collecting data until a threshold amount is reached and, for successive data subsets of a specified size, organizing the values into value-range bins, supplementing the bins with additional values assigned according to their values, and compiling a processed dataset from the supplemented bins.
Step 2A, Prong 2: The judicial exception is not integrated into a practical application. The additional elements, considered individually and as an ordered combination, do not integrate the judicial exception into a practical application.
The additional elements are: “at least one processor”; “at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations”; the recitation that the data operated on is “stored streaming data”; and “generating, by a synthetic data generator, a plurality of synthetic data points”.
As discussed with respect to claim 21 (applying Ex Parte Desjardins and MPEP 2106.04(d)(1) and 2106.05(a)), the specification describes (i) improvements in the accuracy of the statistical approximation (¶¶[003], [026], [030], [033]), which are improvements in the abstract idea itself rather than in technology (MPEP 2106.05(a), subsection II) and which, in any event, claim 30 does not reflect because it recites none of the steps credited with providing them (¶¶[030]-[031], [033]); (ii) regulatory benefits from avoiding the use of sensitive real-world data (¶¶[002]-[003]), which are not technological; and (iii) a memory-handling scheme for real-time processing of a continuous data stream in limited memory (¶¶[004], [017], [021]-[024], [034]-[035]; Fig. 2). Claim 30’s recitation that the processing is “iteratively performed based on a window size specified by a processing threshold” does not cause the claim to reflect that memory-handling scheme, even assuming the scheme would be recognized as an improvement in computer functionality. In the specification, the asserted efficiency results from accumulating new data in a second memory portion while a first portion is processed, storing the processed data in long-term memory or another system, and purging the first memory portion so that it can be reused (¶¶[023]-[024], [034]-[035]). Claim 30 recites none of these features; it recites the window size only as a parameter specifying how much data is processed in each iteration, and nothing in the claim requires that any memory be allocated as a window, released, or reused, or that the processing keep pace with the incoming stream. Nor does the specification’s statement that the window size affects “the overall processor load” (¶[022]) change this conclusion: selecting how much data a generic processor handles at a time is a generic resource tradeoff, and claim 30 recites no particular technique for managing memory or processor resources. MPEP 2106.05(f). Because claim 30 does not include the components or steps that provide the described improvement, it does not reflect an improvement in computer functionality or any other technology. MPEP 2106.05(a); Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d at 1316. Nor does the claim reflect any improvement to the synthetic data generator or to a machine-learning model: the statement in ¶[001] that the embodiments concern improvements to generative adversarial network models is a bare assertion unsupported by any technical explanation in this disclosure (MPEP 2106.04(d)(1)), and claim 30 recites the generator solely in terms of its result, so it could not reflect any improvement described in the application incorporated by reference (¶[025]). See ¶¶[019]-[020], [036].
The processor and the memory storing instructions are generic computing components (¶[016]) used as tools to perform the abstract idea, which amounts to mere instructions to apply the exception on a generic computer (MPEP 2106.05(f); Alice, 573 U.S. at 223-24), and they do not constitute a particular machine (MPEP 2106.05(b)). The recitation of “stored streaming data” is mere data gathering (storing data before it is analyzed) and selection of the type or source of the data to be analyzed, i.e., insignificant extra-solution activity (MPEP 2106.05(g); Electric Power Group, 830 F.3d at 1355), and it generally links the abstract idea to the environment of data streams (MPEP 2106.05(h)). The limitation “generating, by a synthetic data generator, a plurality of synthetic data points” is recited solely in terms of its result, without any details of how the data points are generated, and invokes the generator, whether given its plain meaning or construed under 35 U.S.C. 112(f) as computing devices executing a generative model (¶¶[019]-[020], [036]), merely as a tool to supply values for the abstract binning; it is mere instructions to apply the exception (MPEP 2106.05(f); Recentive Analytics, 134 F.4th 1205). Even if the iterative processing were considered an additional element, repeating the abstract analysis on successive data subsets does not meaningfully limit the exception. Likewise, even if the size determination were treated as an additional element, monitoring how much data has been buffered and initiating processing when a threshold is reached is a generic computer function recited at a high level of generality and amounts to mere instructions to apply the exception, for the reasons discussed with respect to claim 21 (MPEP 2106.05(f)). Viewed as a whole and as an ordered combination, the additional elements do not impose a meaningful limit on the judicial exception. Accordingly, claim 30 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the reasons discussed in the Step 2B analysis of claim 21, the processor and the memory storing instructions amount to mere instructions to apply the exception using generic computer components and are well-understood, routine, and conventional (¶¶[016], [038], [042]; Alice, 573 U.S. at 225-26; Bancorp, 687 F.3d at 1278); storing the streaming data is the well-understood, routine, and conventional activity of receiving data and storing it in memory (MPEP 2106.05(d)(II), citing Versata, 793 F.3d at 1334, OIP Techs., 788 F.3d at 1363, and buySAFE, 765 F.3d at 1355; ¶[018]); and the synthetic data generator amounts to mere instructions to apply the exception, a conclusion that carries over to Step 2B (MPEP 2106.05, subsection II), and is also well-understood, routine, and conventional as evidenced by the specification’s generic description of it, including its reliance on commercially available machine learning applications and known model types described without particulars (¶¶[019]-[020], [036]; MPEP 2106.07(a), subsection III(A); see also Goodfellow, Bengio & Courville, Deep Learning, chs. 10, 20 (MIT Press 2016); Creswell et al., IEEE Signal Processing Magazine, vol. 35, no. 1, pp. 53-65 (Jan. 2018)); and, to the extent the size determination is considered an additional element, it likewise amounts to mere instructions to apply the exception for the reasons discussed with respect to claim 21. To the extent the iterative, window-based processing is considered an additional element, performing the same processing repeatedly on successive data subsets is the performance of repetitive calculations, which the courts have recognized as well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing Parker v. Flook, 437 U.S. 584, 594 (1978) and Bancorp, 687 F.3d at 1278). Considered as an ordered combination, the additional elements add nothing that is not already present when they are considered individually; each performs its ordinary, conventional function in a conventional arrangement. Accordingly, claim 30 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 31
Step 1: Claim 31 depends from claim 30 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 31 incorporates all the limitations of claim 30 and is therefore directed to the same abstract idea identified in claim 30. Claim 31 additionally recites “wherein analyzing the plurality of bins comprises determining minimum and maximum values for each bin.” Determining the minimum and maximum values of the samples in each bin is a mathematical calculation (identifying the extreme values of a set of numbers) and is also an observation and evaluation that a person can practically perform by inspecting the values tallied in each range and noting the smallest and largest. See MPEP 2106.04(a)(2)(I), (III); ¶[029]. No new additional elements are introduced.
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 32
Step 1: Claim 32 depends from claim 30 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 32 incorporates all the limitations of claim 30 and is therefore directed to the same abstract idea identified in claim 30. Claim 32 additionally recites “wherein the processing is based on a window size specified by a processing threshold.” This limitation repeats the window-size limitation already recited in claim 30. The window size is “a size of the dataset to be processed” and “affects the accuracy and the overall processor load,” and the processing threshold is used “to maintain a balance between accuracy, processing speed, and cost” (¶[022]). Basing the binning analysis on an amount of data specified by a threshold value sets a numerical parameter of the mathematical analysis and reflects a judgment about how much data to analyze at a time, which a person can practically make, e.g., by deciding to tally collected values in batches of a chosen size. See MPEP 2106.04(a)(2)(I), (III). No new additional elements are introduced.
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 33
Step 1: Claim 33 depends from claim 32 (which depends from claim 30) and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Step 2A, Prong 1: No additional abstract idea limitations are introduced.
Step 2A, Prong 2: Claim 33 additionally recites “wherein the stored streaming data includes at least one of image data, video data, or audio data.” This limitation specifies the type of data that is analyzed and does not alter the abstract idea: the claim recites no image-, video-, or audio-specific processing (e.g., no feature extraction, compare ¶[027]), and creating value-range bins, assigning sample values to them, and compiling a dataset from the populated bins remain mathematical concepts irrespective of the type of data from which the sample values are taken. See MPEP 2106.04(a)(2)(I). The additional elements in claim 33 are those of claim 30 plus the limitation of the stored streaming data to image data, video data, or audio data.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 30, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator do not integrate the judicial exception into a practical application.
With respect to the new additional element, limiting the stored streaming data to image, video, or audio data merely specifies the type of data to which the abstract binning is applied. This generally links the use of the judicial exception to a particular technological environment or field of use without limiting how the analysis is performed (MPEP 2106.05(h); see Electric Power Group, 830 F.3d at 1354-55 (limiting the collection and analysis of information to power-grid data)), and selecting a particular type of data to be manipulated is insignificant extra-solution activity (MPEP 2106.05(g)). The claim does not recite any improvement to image, video, or audio processing technology; it merely bins values from such data in the same manner as any other numerical data. Accordingly, claim 33 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 30, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator are well-understood, routine, and conventional and amount to no more than mere instructions to apply the exception. With respect to the limitation to image, video, or audio data, a field-of-use limitation cannot provide significantly more. MPEP 2106.05(h). Further, to the extent the limitation requires receiving and storing image, video, or audio data streams, receiving and storing such data is well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610 (Fed. Cir. 2016) (transmitting images), buySAFE, 765 F.3d at 1355, and Versata, 793 F.3d at 1334); see also ¶¶[018], [036] (describing the incoming data stream generically as including numeric, image, audio, or video data). Considered as an ordered combination with the elements of claim 30, the data-type limitation adds nothing beyond the conventional arrangement discussed for claim 30. Accordingly, claim 33 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 34
Step 1: Claim 34 depends from claim 30 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: No additional abstract idea limitations are introduced.
Step 2A, Prong 2: Claim 34 additionally recites “wherein the stored streaming data includes at least one of salary information, age information, or tax information.” This limitation specifies the content of the information that is analyzed and does not alter the abstract idea; binning salary, age, or tax values is precisely the kind of analysis that a person can practically perform with pen and paper (see ¶[027]). See MPEP 2106.04(a)(2)(I), (III). The additional elements in claim 34 are those of claim 30 plus the limitation of the stored streaming data to salary information, age information, or tax information.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 30, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator do not integrate the judicial exception into a practical application.
With respect to the new additional element, limiting the analyzed data to salary, age, or tax information merely specifies the content of the information to which the abstract idea is applied, which generally links the judicial exception to a particular field of use. MPEP 2106.05(h); see Bilski v. Kappos, 561 U.S. 593, 612 (2010) (limiting an abstract idea to one field of use does not make the concept patentable). “[M]erely selecting information, by content or source, for collection, analysis, and display does nothing significant to differentiate a process from ordinary mental processes.” Electric Power Group, 830 F.3d at 1355; see MPEP 2106.05(g). Accordingly, claim 34 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 30, the processor, the memory storing instructions, the stored streaming data, and the synthetic data generator are well-understood, routine, and conventional and amount to no more than mere instructions to apply the exception. With respect to the limitation to salary, age, or tax information, a field-of-use limitation cannot provide significantly more (MPEP 2106.05(h)), and to the extent the limitation requires receiving and storing such information, receiving and storing data is well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing buySAFE, 765 F.3d at 1355, and Versata, 793 F.3d at 1334); see also ¶[027] (describing a stream of single numbers such as salary, age, or tax as the simplest form of the data stream). Considered as an ordered combination with the elements of claim 30, the data-content limitation adds nothing beyond the conventional arrangement discussed for claim 30. Accordingly, claim 34 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 35
Step 1: Claim 35 depends from claim 30 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 35 incorporates all the limitations of claim 30 and is therefore directed to the same abstract idea identified in claim 30. Claim 35 additionally recites “wherein the plurality of synthetic data points are generated based on a generator threshold specifying one or more data points that are generated at respective iterations.” The generator threshold is a number specifying how many data points are produced at each iteration (¶[032]). The limitation does not recite a step of selecting or setting the threshold; it further specifies how the synthetic data generator of claim 30, an additional element, operates, i.e., by producing data points over respective iterations in the quantity specified by the threshold. The limitation therefore does not alter the abstract idea identified in claim 30, and it is evaluated below together with the synthetic data generator.
Step 2A, Prong 2: The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 30, the additional elements do not integrate the judicial exception into a practical application. Requiring the synthetic data generator to produce a specified number of data points at each of several iterations merely repeats the generic data-generation function of claim 30 without specifying how the data points are generated; it remains mere instructions to apply the exception using a generic tool (MPEP 2106.05(f)), and it does not reflect any improvement to the generator, to a machine-learning model, or to computer functionality. Accordingly, claim 35 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 30, the additional elements, including the synthetic data generator, are well-understood, routine, and conventional. Invoking the generator repeatedly to produce a specified number of data points per iteration adds only the repetitive performance of a conventional function, which the courts have recognized as well-understood, routine, and conventional. MPEP 2106.05(d)(II) (citing Parker v. Flook, 437 U.S. 584, 594 (1978), and Bancorp, 687 F.3d at 1278 (performing repetitive calculations)); see also ¶[032] (generator threshold may be preset or set manually or automatically). Accordingly, claim 35 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 36
Step 1: Claim 36 depends from claim 30 and recites a system, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 36 incorporates all the limitations of claim 30 and is therefore directed to the same abstract idea identified in claim 30. Claim 36 additionally recites “wherein the generator threshold is pre-set or set based on at least one of determined minimum and maximum values of the assigned samples or a number of the assigned samples.” Setting a numerical threshold in advance, or as a function of the minimum and maximum values or the number of the assigned samples, recites a mathematical relationship between the threshold and those values, and it is a judgment that a person can practically make by inspecting the tallied values and choosing how many additional values to generate; the specification confirms that the generator threshold may be “set manually” (¶[032]). See MPEP 2106.04(a)(2)(I), (III). No new additional elements are introduced.
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 37
Step 1: Claim 37 recites “A method for synthetic data generation” comprising a series of steps, which falls within the statutory category of a process. See MPEP 2106.03. Accordingly, claim 37 satisfies Step 1 of the eligibility analysis.
Step 2A, Prong 1: Claim 37 is directed to an abstract idea. Specifically, claim 37 recites mathematical concepts and mental processes. The claim recites: “determining a size of stored streaming data has reached a first threshold”; “in response to the size determination, processing the stored streaming data, the processing comprising: creating a plurality of bins, having respective data ranges; and assigning samples from the stored streaming data to the plurality of bins”; “populating the bins with synthetic data, the populating comprising: . . . assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins”; and “creating a processed dataset based on the populated bins.” These limitations are substantively identical to the limitations of claim 21 discussed above and, for the same reasons, recite mathematical concepts, namely the construction and supplementation of a frequency distribution (histogram) through mathematical correlations between values and value ranges (MPEP 2106.04(a)(2)(I); Digitech, 758 F.3d at 1350-51; SAP America, 898 F.3d at 1163-65; see ¶¶[029], [033]-[034]), and mental processes that a person can practically perform with pen and paper at the claimed level of generality, i.e., counting collected values against a threshold, writing down value ranges, tallying each value into its range, and writing out the resulting dataset (MPEP 2106.04(a)(2)(III); Electric Power Group, 830 F.3d at 1353-55; CyberSource, 654 F.3d at 1372-73). Indeed, claim 37 recites no processor or memory; apart from the synthetic data generator, every step may be performed by a person.
Considered together (MPEP 2106.04, subsection II(B)), these limitations recite the abstract idea of collecting data until a threshold amount is reached, organizing the collected values into value-range bins, supplementing the bins with additional values assigned according to their values, and compiling a processed dataset from the supplemented bins.
Step 2A, Prong 2: The judicial exception is not integrated into a practical application. The additional elements, considered individually and as an ordered combination, do not integrate the judicial exception into a practical application.
The additional elements are the recitation that the data operated on is “stored streaming data” and “generating, by a synthetic data generator, a plurality of synthetic data points,” which, as discussed with respect to claim 21, is evaluated as an additional element rather than as part of the mental process (¶[036]).
As discussed with respect to claim 21 (applying Ex Parte Desjardins and MPEP 2106.04(d)(1) and 2106.05(a)), the benefits described in the specification are improvements in the accuracy of the statistical approximation (¶¶[003], [026], [030], [033]), which are improvements in the abstract idea itself (MPEP 2106.05(a), subsection II) and which, in any event, claim 37 does not reflect because it recites none of the steps credited with providing them (¶¶[030]-[031], [033]); regulatory benefits (¶¶[002]-[003]), which are not technological; and a memory-handling scheme for real-time processing of a continuous data stream in limited memory (¶¶[004], [017], [021]-[024], [034]-[035]; Fig. 2). Even assuming that the memory-handling scheme would be recognized as an improvement in computer functionality, claim 37 does not reflect it: the claim does not recite allocating memory portions or windows, accumulating new data in a second memory portion while stored data is processed, storing the processed dataset in a second storage device, or purging or releasing memory, and it recites no memory or processor at all. See MPEP 2106.05(a); Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d at 1316. Nor does the claim reflect any improvement to the synthetic data generator or to a machine-learning model: the statement in ¶[001] that the embodiments concern improvements to generative adversarial network models is a bare assertion unsupported by any technical explanation in this disclosure (MPEP 2106.04(d)(1)), and claim 37 recites the generator solely in terms of its result, so it could not reflect any improvement described in the application incorporated by reference (¶[025]). See ¶¶[019]-[020], [036].
The recitation of “stored streaming data” is mere data gathering (storing data before it is analyzed) and selection of the type or source of the data to be analyzed, i.e., insignificant extra-solution activity (MPEP 2106.05(g); Electric Power Group, 830 F.3d at 1355), and it generally links the abstract idea to the environment of data streams (MPEP 2106.05(h)). The limitation “generating, by a synthetic data generator, a plurality of synthetic data points” is recited solely in terms of its result, without any details of how the data points are generated, and invokes the generator, whether given its plain meaning or construed under 35 U.S.C. 112(f) as computing devices executing a generative model (¶¶[019]-[020], [036]), merely as a tool to supply values for the abstract binning. It is mere instructions to apply the exception (MPEP 2106.05(f); Recentive Analytics, 134 F.4th 1205). Likewise, the analysis would not change if the size determination were treated as an additional element rather than as part of the mental process, for the reasons discussed with respect to claim 21 (MPEP 2106.05(f)). Viewed as a whole and as an ordered combination, the additional elements interact with the abstract idea only by supplying the data to be binned, and they do not impose a meaningful limit on the judicial exception. Accordingly, claim 37 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the reasons discussed in the Step 2B analysis of claim 21, storing the streaming data is the well-understood, routine, and conventional activity of receiving data and storing it in memory (MPEP 2106.05(d)(II), citing Versata, 793 F.3d at 1334, OIP Techs., 788 F.3d at 1363, and buySAFE, 765 F.3d at 1355; ¶[018]), and the synthetic data generator amounts to mere instructions to apply the exception, a conclusion that carries over to Step 2B (MPEP 2106.05, subsection II), and is also well-understood, routine, and conventional as evidenced by the specification’s generic description of it, including its reliance on commercially available machine learning applications and known model types described without particulars (¶¶[019]-[020], [036]; MPEP 2106.07(a), subsection III(A); see also Goodfellow, Bengio & Courville, Deep Learning, chs. 10, 20 (MIT Press 2016); Creswell et al., IEEE Signal Processing Magazine, vol. 35, no. 1, pp. 53-65 (Jan. 2018)). Considered as an ordered combination, the additional elements add nothing that is not already present when they are considered individually: data is stored, a generic data generator supplies additional values, and the abstract binning and compilation are performed on the result. Accordingly, claim 37 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 38
Step 1: Claim 38 depends from claim 37 and recites a method, which falls within the statutory category of a process. See MPEP 2106.03.
Step 2A, Prong 1: Claim 38 incorporates all the limitations of claim 37 and is therefore directed to the same abstract idea identified in claim 37. Claim 38 additionally recites “wherein the plurality of bins includes edge widths defining data ranges for each of the plurality of bins.” The edge widths are the numerical boundaries and widths of the intervals that define each bin’s range, i.e., a mathematical definition of the intervals of a histogram (¶[033]), which a person can also practically set by choosing the boundaries of each range. See MPEP 2106.04(a)(2)(I), (III).
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 39
Step 1: Claim 39 depends from claim 38 (which depends from claim 37) and recites a method, which falls within the statutory category of a process. See MPEP 2106.03.
Step 2A, Prong 1: Claim 39 incorporates all the limitations of claims 37 and 38 and is therefore directed to the same abstract idea identified in claim 37. Claim 39 additionally recites “wherein the edge widths are configured to minimize error associated with approximating an original distribution.” Configuring the interval widths of a histogram to minimize the error with which the histogram approximates an underlying distribution is a mathematical optimization, i.e., a mathematical calculation and relationship expressed in words (¶[033]). See MPEP 2106.04(a)(2)(I); In re Grams, 888 F.2d 835, 837 n.1 (Fed. Cir. 1989) (words used in a claim operating on data to solve a problem can serve the same purpose as a formula); SAP America, 898 F.3d at 1163-65.
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
CLAIM 40
Step 1: Claim 40 depends from claim 37 and recites a method, which falls within the statutory category of a process. See MPEP 2106.03.
Step 2A, Prong 1: Claim 40 incorporates all the limitations of claim 37 and is therefore directed to the same abstract idea identified in claim 37. Claim 40 additionally recites “wherein each of the plurality of bins has a different edge width.” Requiring each bin to have a different edge width is a mathematical constraint on the widths of the intervals of the frequency distribution (i.e., non-uniform binning, ¶[033]), and choosing unequal value ranges is a judgment that a person can practically make with pen and paper. See MPEP 2106.04(a)(2)(I), (III).
Step 2A, Prong 2 and Step 2B: No new additional elements are introduced. The analysis for the parent claim is maintained.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 21-23, 25-28, 30-32 and 34-38 are rejected under 35 U.S.C. 103 as being unpatentable over US Pat. Pub. No. 2013/0179585A1 to Branson et al. (hereinafter Branson) in view of DataSynthesizer: Privacy-Preserving Synthetic Datasets to Ping et al. (hereinafter Ping).
Per claim 21, Branson discloses A system… (Branson: ¶[0007]…Branson discloses a stream computing system of compute nodes, each having a computer processor, which constitutes the system limitation under BRI, "In another embodiment, the system comprises at least two compute nodes, each compute node comprising at least one computer processor"), comprising:
at least one processor (Branson: ¶[0043]…Branson's CPU 205 retrieves and executes programming instructions, which constitutes the at least one processor under BRI, "Each CPU 205 retrieves and executes programming instructions stored in the memory 225"); and
at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations (Branson: ¶[0043]…Branson's memory 225 is a DRAM or Flash device that holds the programming instructions the CPU executes, which constitutes the at least one non-transitory memory storing instructions under BRI, "Each CPU 205 retrieves and executes programming instructions stored in the memory 225…The memory 225 is generally included to be representative of a random access memory (e.g., DRAM or Flash)"; ¶[0044]…the operators held in memory 225 are executable code that carries out the stream-processing operations, which constitutes the instructions that cause the system to perform operations under BRI, "As noted above, each operator 240 may provide a small chunk of executable code configured to evaluate data flowing into a processing element (e.g., PE 235) and to emit data to other operators 240 in that PE or to other PEs in the stream computing application") comprising:
determining a size of stored streaming data has reached a first threshold (Branson: ¶[0046]…Branson stores the tuples of the incoming data stream in buffer 260, which constitutes the stored streaming data under BRI, "As shown, storage 230 contains a buffer 260 which stores data tuples 265"; ¶[0052]…Branson's window activator checks whether the buffer has reached a configured size of four tuples, which constitutes the determining a size of stored streaming data has reached a first threshold under BRI, "In one embodiment, the windowing parameters 257 associated with the operator instructs the window activator 255 to trigger a window once the buffer 260 reaches a size of four tuples");
in response to the size determination, processing the stored streaming data, the processing comprising (Branson: ¶[0052]…once the buffer-size parameter is satisfied, Branson triggers a window over the buffered tuples, "Using either parameter, at time 5, the window activator 255 determines that the parameter is satisfied and triggers a window"; ¶[0045]…the operator then processes the buffered tuples contained in the triggered window, which constitutes the in response to the size determination, processing the stored streaming data under BRI, "In one embodiment, an operator 240 may only evaluate received tuples after a window is triggered, and even then, the operator 240 processes only the tuples contained within the window"):
Branson does not expressly disclose, but Ping does teach:
…for synthetic data generation… (Ping: Abstract…Ping discloses a tool that takes an input dataset and generates a statistically similar synthetic dataset, which constitutes the for synthetic data generation under BRI, "To facilitate collaboration over sensitive data, we present DataSynthesizer, a tool that takes a sensitive dataset as input and generates a structurally and statistically similar synthetic dataset with strong privacy guarantees");
creating a plurality of bins having respective data ranges (Ping: § 3…Ping's DataDescriber builds an equi-width histogram over each numerical attribute, i.e., a plurality of bins each spanning its own range of attribute values, which constitutes the creating a plurality of bins having respective data ranges under BRI, "For non-categorical numerical and datetime attributes, DataDescriber derives an equi-width histogram to represent the distribution"; § 3.1.3…the histogram consists of a configurable number of bins, "Precision of the histograms can be refined by a parameter named histogram size, which represents the number of bins and is set to 20 by default"; § 3.1.4…each bin covers a stated value interval, "For example, the age attribute has intervals {[10, 20),[20, 30),[30, 40)} in its unconditioned distribution"); and
assigning samples…to the plurality of bins (Ping: § 3.1.3…Ping's frequency-based histogram estimation counts how many input tuples fall within each bin, which necessarily allocates each input sample to the bin whose interval contains its value and constitutes the assigning samples to the plurality of bins under BRI, "When invoked in independent attribute mode, DataDescriber performs frequency-based estimation of the unconditioned probability distributions of the attributes"; Fig. 2…Fig. 2 plots the resulting number of input tuples per age bin, "Histogram on age: Adult Income [6]");
populating the bins with synthetic data, the populating comprising (Ping: § 3…Ping's DataGenerator draws synthetic values out of the bins of the learned histogram, filling those bins with synthetic data, which constitutes the populating the bins with synthetic data under BRI, "DataGenerator draws samples from this histogram during data generation using uniform sampling"):
generating, by a synthetic data generator, a plurality of synthetic data points (Ping: § 3.2…Ping's DataGenerator module samples synthetic tuples, which constitutes the generating, by a synthetic data generator, a plurality of synthetic data points under BRI, "Given a dataset description file generated by DataDescriber, DataGenerator samples synthetic data from the distributions in this file"); and
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins (Ping: § 3…Ping's DataGenerator draws each synthetic value from a bin of the histogram by uniform sampling over that bin's interval, so that every synthetic value is placed in the bin whose data range it was drawn from, which constitutes the assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins under BRI, "DataGenerator draws samples from this histogram during data generation using uniform sampling"; § 2…Ping then bins the synthetic output into the same histogram form as the input, "DataDescriber can be invoked on any CSV file. In particular, we can invoke it on both the input file and the output file, and compare the resulting summaries"; Fig. 3…Fig. 3 plots the synthetic tuples per age bin, "Histogram on age: synthetic"); and
creating a processed dataset based on the populated bins (Ping: Abstract…Ping's DataGenerator outputs the synthetic dataset sampled from the histogram bins, which constitutes the creating a processed dataset based on the populated bins limitation under BRI, "DataGenerator samples from the summary computed by DataDescriber and outputs synthetic data").
Branson and Ping are analogous art because they are from the same field of endeavor, specifically computer processing of tuple-based data sets to derive statistical summaries of that data: Branson's stream operators evaluate windows of tuples to compute statistics such as a moving average (Branson: ¶[0055]), and Ping derives a histogram summary of each attribute of an input dataset (Ping: § 3). Ping is, moreover, reasonably pertinent to the problem at hand of generating synthetic data having statistical characteristics similar to those of the original data (Specification ¶[003]), because Ping generates "a structurally and statistically similar synthetic dataset" (Ping: Abstract).
Before the effective filing date of the claimed invention, it would have been obvious to a PHOSITA to use Ping's DataDescriber and DataGenerator as the processing function that Branson's operator performs on each triggered window of buffered stream tuples, so that, once Branson's buffer reaches its configured size, the tuples in the window are binned into a histogram, synthetic data points are sampled from and binned into that histogram, and the resulting synthetic dataset is output.
The suggestion/motivation for doing so would have been provided by Ping itself, which teaches that its synthesizer is intended to run as a component inside larger data platforms, "We see DataSynthesizer being used in a variety of application contexts, both as a stand-alone library and as a component of more comprehensive data sharing platforms" (Ping: § 5). Branson, in turn, teaches that a stream operator applies its processing function to the group of tuples held in a triggered window rather than to tuples one at a time, "Instead of processing the tuples as they are received, the operator may wait to evaluate a group of tuples—i.e., a window" (Branson: ¶[0019]). Furthermore, this is the combination of prior art elements according to known methods to yield predictable results, the rationale of MPEP § 2143(I)(A): Ping's histogram-based synthesizer operates on a set of tuples, Branson's triggered window supplies exactly such a set, and each element performs the same function in combination that it performs separately.
Per claim 22, Branson combined with Ping discloses claim 21. Branson does not expressly disclose, but Ping does teach: wherein analyzing the plurality of bins comprises determining minimum and maximum values for each bin (§3.1.4…Ping's histogram records for each bin the interval running from that bin's lowest to its highest value, which constitutes the determining minimum and maximum values for each bin under BRI, "For example, the age attribute has intervals {[10, 20),[20, 30),[30, 40)} in its unconditioned distribution"). The rationale to combine Ping with Branson is the same as the parent claim.
Per claim 23, Branson combined with Ping discloses claim 21. Branson further teaches wherein the processing is based on a window size specified by a processing threshold (Branson: ¶[0045]…Branson's windowing parameters fix the number of tuples a window holds, and the operator processes only the tuples in that window, which constitutes the processing is based on a window size specified by a processing threshold under BRI, "The windowing parameters 257 may include, for example, a predefined number of tuples in a window, a predefined period of time, a threshold for evaluating the tuple rate, a predefined number of exceptions, a threshold for evaluating an exception rate, and any combination thereof").
Per claim 25, Branson combined with Ping discloses claim 21. Branson does not expressly disclose, but Ping does teach: wherein…includes at least one of…age information… (Ping: § 3.3…Ping's input data set includes the age attribute of the Adult Income dataset, which constitutes the age information under BRI, "Figures 2 and 3 present such summaries for the attribute age from the Adult Income dataset [6]"). The rationale to combine Ping with Branson is the same as the parent claim.
Per claim 26, Branson combined with Ping discloses claim 21. Branson does not expressly disclose, but Ping does teach: wherein the plurality of synthetic data points are generated based on a generator threshold specifying one or more data points that are generated at respective iterations (Ping: Algorithm 2…Ping's DataGenerator takes as an input parameter the number of tuples n to generate, "Require: number of tuples n to generate"; § 3.2…that number is fixed by the user or by default before generation; in the Branson and Ping combination applied to claim 21, the generator runs once for each window Branson triggers, so each of those successive runs generates the specified number of synthetic data points, which constitutes the generator threshold specifying one or more data points that are generated at respective iterations under BRI, "The size of the output dataset is specified by the user, and defaults to n, the size of the input dataset"). The rationale to combine Ping with Branson is the same as the parent claim.
Per claim 27, Branson combined with Ping discloses claim 26. Branson does not expressly disclose, but Ping does teach: wherein the generator threshold is pre-set or set based on at least one of…a number of the assigned samples (Ping: § 3.2…Ping's output size is either specified in advance by the user or defaults to n, the number of tuples in the input dataset, which constitutes the generator threshold is pre-set or set based on a number of the assigned samples under BRI, "The size of the output dataset is specified by the user, and defaults to n, the size of the input dataset"). The rationale to combine Ping with Branson is the same as the parent claim.
Per claim 28, Branson combined with Ping discloses claim 21. Branson does not expressly disclose, but Ping does teach: wherein the plurality of bins includes edge widths defining data ranges for each of the plurality of bins (Ping: § 3…Ping's equi-width histogram gives every bin a width that bounds its range of values, which constitutes the edge widths defining data ranges for each of the plurality of bins under BRI, "For non-categorical numerical and datetime attributes, DataDescriber derives an equi-width histogram to represent the distribution"; § 3.1.4…for example, each age bin is ten years wide, "For example, the age attribute has intervals {[10, 20),[20, 30),[30, 40)} in its unconditioned distribution"). The rationale to combine Ping with Branson is the same as the parent claim.
Per claim 30, Branson discloses A system… (Branson: ¶[0007]…Branson discloses a stream computing system of compute nodes, each having a computer processor, which constitutes the system under BRI, "In another embodiment, the system comprises at least two compute nodes, each compute node comprising at least one computer processor"), comprising:
at least one processor (Branson: ¶[0043]…Branson's CPU 205 retrieves and executes programming instructions, which constitutes the at least one processor under BRI, "Each CPU 205 retrieves and executes programming instructions stored in the memory 225"); and
at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations comprising (Branson: ¶[0043]…Branson's memory 225 is a DRAM or Flash device that holds the programming instructions the CPU executes, which constitutes the at least one non-transitory memory storing instructions under BRI, "Each CPU 205 retrieves and executes programming instructions stored in the memory 225…The memory 225 is generally included to be representative of a random access memory (e.g., DRAM or Flash)"; ¶[0044]…the operators held in memory 225 are executable code that carries out the stream-processing operations, which constitutes the instructions that cause the system to perform operations under BRI, "As noted above, each operator 240 may provide a small chunk of executable code configured to evaluate data flowing into a processing element (e.g., PE 235) and to emit data to other operators 240 in that PE or to other PEs in the stream computing application"):
determining a size of stored streaming data has reached a first threshold (Branson: ¶[0046]…Branson stores the tuples of the incoming data stream in buffer 260, which constitutes the stored streaming data under BRI, "As shown, storage 230 contains a buffer 260 which stores data tuples 265"; ¶[0052]…Branson's window activator checks whether the buffer has reached a configured size of four tuples, which constitutes the determining a size of stored streaming data has reached a first threshold under BRI, "In one embodiment, the windowing parameters 257 associated with the operator instructs the window activator 255 to trigger a window once the buffer 260 reaches a size of four tuples");
in response to the size determination, processing the stored streaming data, the processing being iteratively performed based on a window size specified by a processing threshold and comprising (Branson: ¶[0052]…once the buffer-size parameter is satisfied, Branson triggers a window over the buffered tuples, "Using either parameter, at time 5, the window activator 255 determines that the parameter is satisfied and triggers a window"; ¶[0045]…Branson's windowing parameters fix the number of tuples a window holds, which constitutes the window size specified by a processing threshold under BRI, "The windowing parameters 257 may include, for example, a predefined number of tuples in a window, a predefined period of time, a threshold for evaluating the tuple rate, a predefined number of exceptions, a threshold for evaluating an exception rate, and any combination thereof"; ¶[0054]…after each window is processed Branson waits for the buffer to refill to that same number of tuples and triggers the next window, which constitutes the processing being iteratively performed under BRI, "For a new window to be triggered, the window activator 255 waits until the buffer 260 again contains the requisite number of tuples or until the predefined period of time has expired"):
Branson does not expressly disclose, but Ping does teach:
…for synthetic data generation… (Ping: Abstract…Ping discloses a tool that takes an input dataset and generates a statistically similar synthetic dataset, which constitutes the for synthetic data generation under BRI, "To facilitate collaboration over sensitive data, we present DataSynthesizer, a tool that takes a sensitive dataset as input and generates a structurally and statistically similar synthetic dataset with strong privacy guarantees").
creating a plurality of bins having respective data ranges (Ping: § 3…Ping's DataDescriber builds an equi-width histogram over each numerical attribute, i.e., a plurality of bins each spanning its own range of attribute values, which constitutes the creating a plurality of bins having respective data ranges under BRI, "For non-categorical numerical and datetime attributes, DataDescriber derives an equi-width histogram to represent the distribution"; § 3.1.3…the histogram consists of a configurable number of bins, "Precision of the histograms can be refined by a parameter named histogram size, which represents the number of bins and is set to 20 by default"; § 3.1.4…each bin covers a stated value interval, "For example, the age attribute has intervals {[10, 20),[20, 30),[30, 40)} in its unconditioned distribution"); and
assigning samples…to the plurality of bins (Ping: § 3.1.3…Ping's frequency-based histogram estimation counts how many input tuples fall within each bin, which necessarily allocates each input sample to the bin whose interval contains its value and constitutes the assigning samples to the plurality of bins under BRI, "When invoked in independent attribute mode, DataDescriber performs frequency-based estimation of the unconditioned probability distributions of the attributes"; Fig. 2…Fig. 2 plots the resulting number of input tuples per age bin, "Histogram on age: Adult Income [6]"); and
populating the bins with synthetic data, the populating comprising (Ping: § 3…Ping's DataGenerator draws synthetic values out of the bins of the learned histogram, filling those bins with synthetic data, which constitutes the populating the bins with synthetic data under BRI, "DataGenerator draws samples from this histogram during data generation using uniform sampling"):
generating, by a synthetic data generator, a plurality of synthetic data points (Ping: § 3.2…Ping's DataGenerator module samples synthetic tuples, which constitutes the generating, by a synthetic data generator, a plurality of synthetic data points under BRI, "Given a dataset description file generated by DataDescriber, DataGenerator samples synthetic data from the distributions in this file"); and
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins (Ping: § 3…Ping's DataGenerator draws each synthetic value from a bin of the histogram by uniform sampling over that bin's interval, so that every synthetic value is placed in the bin whose data range it was drawn from, which constitutes the assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins under BRI, "DataGenerator draws samples from this histogram during data generation using uniform sampling"; § 2…Ping then bins the synthetic output into the same histogram form as the input, "DataDescriber can be invoked on any CSV file. In particular, we can invoke it on both the input file and the output file, and compare the resulting summaries"; Fig. 3…Fig. 3 plots the synthetic tuples per age bin, "Histogram on age: synthetic"); and
creating a processed dataset based on the populated bins (Ping: Abstract…Ping's DataGenerator outputs the synthetic dataset sampled from the histogram bins, which constitutes the creating a processed dataset based on the populated bins under BRI, "DataGenerator samples from the summary computed by DataDescriber and outputs synthetic data").
Branson and Ping are analogous art because they are from the same field of endeavor, specifically computer processing of tuple-based data sets to derive statistical summaries of that data: Branson's stream operators evaluate windows of tuples to compute statistics such as a moving average (Branson: ¶[0055]), and Ping derives a histogram summary of each attribute of an input dataset (Ping: § 3). Ping is, moreover, reasonably pertinent to the problem at hand, that of generating synthetic data having statistical characteristics similar to those of the original data (Specification ¶[003]), because Ping generates "a structurally and statistically similar synthetic dataset" (Ping: Abstract).
Before the effective filing date of the claimed invention, it would have been obvious to a PHOSITA to use Ping's DataDescriber and DataGenerator as the processing function that Branson's operator performs on each triggered window of buffered stream tuples, so that, once Branson's buffer reaches its configured size, the tuples in the window are binned into a histogram, synthetic data points are sampled from and binned into that histogram, and the resulting synthetic dataset is output.
The suggestion/motivation for doing so would have been provided by Ping itself, which teaches that its synthesizer is intended to run as a component inside larger data platforms, "We see DataSynthesizer being used in a variety of application contexts, both as a stand-alone library and as a component of more comprehensive data sharing platforms" (Ping: § 5). Branson, in turn, teaches that a stream operator applies its processing function to the group of tuples held in a triggered window rather than to tuples one at a time, "Instead of processing the tuples as they are received, the operator may wait to evaluate a group of tuples—i.e., a window" (Branson: ¶[0019]). Furthermore, this is the combination of prior art elements according to known methods to yield predictable results, the rationale of MPEP § 2143(I)(A): Ping's histogram-based synthesizer operates on a set of tuples, Branson's triggered window supplies exactly such a set, and each element performs the same function in combination that it performs separately.
Claims 31, 32, 34, 35 and 36 are substantially similar in scope and spirit as claims 22, 23, 25, 26 and 27. Therefore the rejections of claims 22, 23, 25, 26 and 27 are applied accordingly.
Claims 37 and 38 are substantially similar in scope and spirit as claims 21 and 28. Therefore the rejections of claims 21 and 28 are applied accordingly.
Claims 24 and 33 are rejected under 35 U.S.C. 103 as being unpatentable over Branson in view of Ping, as applied in the rejection of claims 21 and 32 above, and further in view of US Pat. Pub. No. 2016/0328450A1 to Branson et al. (hereinafter Branson '450).
Per claim 24, Branson combined with Ping discloses claim 21. Branson combined with Ping does not expressly disclose, but Branson '450 does teach: wherein the stored streaming data includes at least one of image data… (Branson '450: ¶[0058]…Branson '450 teaches that the data entering a stream computing application, and carried through it as tuples, may be image data, which constitutes the stored streaming data includes image data under BRI, "Information enters the stream computing application through the two sources 602 and 604. The information may be any form of data, whether structured or unstructured, e.g., sensor data or image data").
Per claim 33, Branson combined with Ping discloses claim 32. Branson combined with Ping does not expressly disclose, but Branson '450 does teach: wherein the stored streaming data includes at least one of image data… (Branson '450: ¶[0058]…Branson '450 teaches that the data entering a stream computing application, and carried through it as tuples, may be image data, which constitutes the stored streaming data includes image data limitation under BRI, "Information enters the stream computing application through the two sources 602 and 604. The information may be any form of data, whether structured or unstructured, e.g., sensor data or image data").
As it pertains to claims 24 and 33, Branson, Ping and Branson '450 are analogous art because they are from the same field of endeavor, specifically computer processing of tuple-based data sets to derive statistical summaries of that data; Branson and Branson '450 in particular both address windowed processing of tuples in a stream computing application (Branson: ¶[0045]; Branson '450: ¶[0028]), and Ping derives a histogram summary of each attribute of its input tuples (Ping: § 3).
Before the effective filing date of the claimed invention, it would have been obvious to a PHOSITA to apply the windowed stream processing and histogram-based synthesis of Branson and Ping to a data stream whose tuples carry image data, as taught by Branson '450, so that the stored streaming data includes image data.
The suggestion/motivation for doing so would have been provided by Branson '450 itself, which teaches that a stream computing application is not limited to a particular kind of data and may carry image data, "Information enters the stream computing application through the two sources 602 and 604. The information may be any form of data, whether structured or unstructured, e.g., sensor data or image data" (Branson '450: ¶[0058]). Furthermore, this is the application of a known technique to a known system ready for improvement to yield predictable results, the rationale of MPEP § 2143(I)(D): the Branson and Ping combination bins and synthesizes whatever tuple values its window holds, and image-data tuples are simply another kind of value it would process in the same way.
Claims 29, 39 and 40 are rejected under 35 U.S.C. 103 as being unpatentable over Branson in view of Ping, as applied in the rejection of claims 28, 37 and 38 above, and further in view of Improved Histograms for Selectivity Estimation of Range Predicates to Poosala et al. (hereinafter Poosala).
Per claim 29, Branson combined with Ping discloses claim 28. Branson combined with Ping does not expressly disclose, but Poosala does teach: wherein the edge widths are configured to minimize error associated with approximating an original distribution (Poosala: p. 298, § 6.1…Poosala's v-optimal(V,F) histograms place the boundaries of buckets of contiguous attribute values so as to minimize the variance, i.e., the error, of the histogram's approximation of the underlying data distribution, which constitutes the edge widths are configured to minimize error associated with approximating an original distribution under BRI, "The v-optimal(V,F) and v-optimal(V,A) histograms minimize the variances in frequencies and areas respectively, while grouping contiguous attribute values"; p. 297, § 4.5…the v-optimal criterion chooses the buckets that minimize the error of the overall approximation, "V-optimal(F,F) histograms group contiguous sets of frequencies into buckets so as to minimize the variance of the overall frequency approximation").
Per claim 39, Branson combined with Ping discloses claim 38. Branson combined with Ping does not expressly disclose, but Poosala does teach: wherein the edge widths are configured to minimize error associated with approximating an original distribution (Poosala: p. 298, § 6.1…Poosala's v-optimal(V,F) histograms place the boundaries of buckets of contiguous attribute values so as to minimize the variance, i.e., the error, of the histogram's approximation of the underlying data distribution, which constitutes the edge widths are configured to minimize error associated with approximating an original distribution under BRI, "The v-optimal(V,F) and v-optimal(V,A) histograms minimize the variances in frequencies and areas respectively, while grouping contiguous attribute values"; p. 297, § 4.5…the v-optimal criterion chooses the buckets that minimize the error of the overall approximation, "V-optimal(F,F) histograms group contiguous sets of frequencies into buckets so as to minimize the variance of the overall frequency approximation").
Per claim 40, Branson combined with Ping discloses claim 37. Branson combined with Ping does not expressly disclose, but Poosala does teach: wherein each of the plurality of bins has a different edge width (Poosala: p. 297, § 4.3…Poosala's equi-depth histograms size each bucket to hold an equal share of the frequencies rather than an equal spread of values, so that each bucket's width is set by the data falling in it rather than by a common interval, "Equi-sum(V,F) histograms are like equi-width histograms but have the sum of the frequencies in each bucket be equal rather than the sum of the spreads [Koo80, PSC84]"; p. 299, § 6.3…Poosala's maxdiff histograms likewise place each bucket boundary at a large gap between adjacent values, so that no two boundaries are constrained to a common spacing and each bucket takes its own width, which constitutes the each of the plurality of bins has a different edge width under BRI, "The maxdiff histograms try to achieve this goal by inserting bucket boundaries between adjacent source values (in sort parameter order) that differ by large amounts").
As it pertains to claims 29, 39 and 40, Branson, Ping and Poosala are analogous art because they are from the same field of endeavor, specifically computer processing of tuple-based data sets to derive statistical summaries of that data; Ping and Poosala in particular both construct histograms over attribute values to approximate a data distribution (Ping: § 3; Poosala: p. 297, § 4.2), and Branson's stream operators evaluate windows of tuples to compute statistics such as a moving average (Branson: ¶[0055]).
Before the effective filing date of the claimed invention, it would have been obvious to a PHOSITA to replace the equal-width bins of Ping's histogram in the Branson and Ping combination with Poosala's v-optimal or equi-depth bucket boundaries, so that the bin edge widths are chosen to minimize the error of approximating the original distribution and differ from bin to bin, as claimed.
The suggestion/motivation for doing so would have been provided by Poosala itself, which teaches that the conventional equal-width and equal-depth histograms are not accurate enough and must be improved upon, "None of the histograms described above are sufficiently accurate for general use in range query result size estimation" (Poosala: p. 297, § 5). Ping, in turn, teaches that the fidelity of the synthetic distribution to the original distribution is the very quantity its tool measures, "The system also computes the correlation coefficient of an appropriate kind, depending on attribute type, and the KL-divergence value, which measure how much the 'after' probability distribution diverges from the 'before' probability distribution for a given attribute" (Ping: § 3.3), so that a PHOSITA would have sought the more accurate bucket boundaries Poosala provides. Furthermore, this is the simple substitution of one known element for another to obtain predictable results, the rationale of MPEP § 2143(I)(B): Poosala's error-minimizing bucket boundaries replace Ping's equal-width boundaries, and each histogram is built and sampled in the same way.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claims 21-28 and 30-38 are rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1, 12-14 and 16-17 of U.S. Patent No. 10,860,892. Although the claims at issue are not identical, they are not patentably distinct from each other because the instant claims are either generic to, and therefore anticipated by, or obvious variants of the Patent claims. Instant independent claims 21 and 37 recite every limitation of Patent claims 1 and 17, respectively, but omit limitations recited in the Patent claims (receiving the data stream from an outside source and storing it in a first storage device; monitoring its size; determining a total number of samples and the minimum and maximum sample values; generating the synthetic data points using a neural network; determining a total number of real and synthetic data points and pausing the synthetic data generator when it reaches a second threshold; storing the processed dataset on a second storage device; and releasing the first storage device capacity). Patent claims 1 and 17 therefore each claim a species of the genus claimed in instant claims 21 and 37, the entire scope of each Patent claim falls within the scope of the corresponding instant claim, and a later patent to the genus would extend the right to exclude already granted for the species. See MPEP 804, subsection II.B.2; In re Goodman, 11 F.3d 1046 (Fed. Cir. 1993). Instant claims 22, 28 and 38 are likewise anticipated, because non-overlapping bins having data ranges, into which samples and synthetic data points are assigned based on the data ranges of the bins, necessarily have minimum and maximum values bounding each bin and edge widths defining each bin's data range. Instant claims 23, 30 and 32 further recite that the processing is (iteratively) performed based on a window size specified by a processing threshold. Patent claim 1 recites processing the stored data of a received data stream in response to its size reaching a first threshold and then releasing the first storage device capacity occupied by that stored data, which frees the capacity to accumulate the next portion of the stream. The first threshold of the Patent claim thus specifies the size of the window of stream data that is processed, and because streaming data continues to arrive, repeating the processing for each successive window of that size would have been an obvious variation to one of ordinary skill in the art. Instant claims 24 and 33 (image, video or audio data) correspond to Patent claims 13 and 14, which recite the same synthetic data generation system operating on image data and video data, and applying the system of Patent claim 1 to such data would have been obvious. Instant claims 25 and 34 recite only the informational content of the stored data (salary, age or tax information); that content does not change any claimed operation, and Patent claims 12 and 16 recite numerical data and sensitive data, of which salary, age and tax information are obvious examples. Instant claims 26, 27, 35 and 36 recite generating the synthetic data points based on a generator threshold specifying the number of points generated at respective iterations, the threshold being pre-set or set based on the minimum and maximum values or the number of the assigned samples. Patent claim 1 recites generating synthetic data points, counting the total number of data points, and pausing the synthetic data generator when that total reaches a second threshold, i.e., a generate-and-check loop controlled by a threshold; fixing in advance (pre-setting) the number of points generated in each pass of that loop is an obvious design choice that one of ordinary skill in the art would have made to control how often the second-threshold check is performed. Instant claims 31 and 38 correspond to instant claims 22 and 28, respectively, and are not patentably distinct for the same reasons. Instant claim 30 otherwise recites the same limitations as instant claim 21 and is not patentably distinct from Patent claim 1 for the reasons given for claims 21 and 23.
Instant Claim 21 ↔ U.S. Pat. No. 10,860,892 Claim 1
Instant Application Claims
U.S. Pat. No. 10,860,892 Claims
A system for synthetic data generation, comprising:
A system for synthetic data generation, comprising:
at least one processor; and
at least one processor; and
at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations comprising:
at least one non-transitory memory storing instructions that, when executed by the at least one processor cause the system to perform operations comprising:
determining a size of stored streaming data has reached a first threshold;
receiving a data stream from an outside source; storing data from the received data stream in a first storage device; monitoring a size of the stored data in the first storage device; determining whether the size of the stored data has reached a first threshold;
in response to the size determination, processing the stored streaming data, the processing comprising:
in response to the size determination, processing the stored data, the processing comprising: determining a total number of samples in the stored data, determining a minimum value and a maximum value of the samples,
creating a plurality of bins having respective data ranges; and
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping,
assigning samples from the stored streaming data to the plurality of bins;
assigning the samples to the bins, based on values of the samples and data ranges of the bins, and determining numbers of samples within the bins;
populating the bins with synthetic data, the populating comprising:
populating the bins with synthetic data, the populating comprising:
generating, by a synthetic data generator, a plurality of synthetic data points; and
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network,
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins; and
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator;
creating a processed dataset based on the populated bins.
creating a processed dataset based on the bins; storing the processed dataset on a second storage device; and releasing the first storage device capacity occupied by the stored data.
Instant Claim 22 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 21, wherein analyzing the plurality of bins comprises determining minimum and maximum values for each bin.
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping, assigning the samples to the bins, based on values of the samples and data ranges of the bins, and determining numbers of samples within the bins;
Instant Claim 23 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 21, wherein the processing is based on a window size specified by a processing threshold.
receiving a data stream from an outside source; storing data from the received data stream in a first storage device; monitoring a size of the stored data in the first storage device; determining whether the size of the stored data has reached a first threshold; ... creating a processed dataset based on the bins; storing the processed dataset on a second storage device; and releasing the first storage device capacity occupied by the stored data. [Window/iteration: No express counterpart in the reference claim; obvious variation (see rationale in rejection).]
Instant Claim 24 ↔ U.S. Pat. No. 10,860,892 Claim 13
The system of claim 21, wherein the stored streaming data includes at least one of image data, video data, or audio data.
[Claim 13] wherein the data stream comprises image data. [Claim 14] wherein the data stream comprises video data.
Instant Claim 25 ↔ U.S. Pat. No. 10,860,892 Claim 12
The system of claim 21, wherein the stored streaming data includes at least one of salary information, age information, or tax information.
[Claim 12] wherein the data stream comprises numerical data. [Claim 16] wherein the data stream contains sensitive data.
Instant Claim 26 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 21, wherein the plurality of synthetic data points are generated based on a generator threshold specifying one or more data points that are generated at respective iterations.
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network, ... assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator; [Per-iteration generator threshold: No express counterpart in the reference claim; obvious variation (see rationale in rejection).]
Instant Claim 27 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 26, wherein the generator threshold is pre-set or set based on at least one of determined minimum and maximum values of the assigned samples or a number of the assigned samples.
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network, ... assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator; [Pre-set threshold: No express counterpart in the reference claim; obvious variation (see rationale in rejection).]
Instant Claim 28 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 21, wherein the plurality of bins includes edge widths defining data ranges for each of the plurality of bins.
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping,
Instant Claim 30 ↔ U.S. Pat. No. 10,860,892 Claim 1
A system for synthetic data generation, comprising:
A system for synthetic data generation, comprising:
at least one processor; and
at least one processor; and
at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to perform operations comprising:
at least one non-transitory memory storing instructions that, when executed by the at least one processor cause the system to perform operations comprising:
determining a size of stored streaming data has reached a first threshold;
receiving a data stream from an outside source; storing data from the received data stream in a first storage device; monitoring a size of the stored data in the first storage device; determining whether the size of the stored data has reached a first threshold;
in response to the size determination, processing the stored streaming data, the processing being iteratively performed based on a window size specified by a processing threshold and comprising:
in response to the size determination, processing the stored data, the processing comprising: determining a total number of samples in the stored data, determining a minimum value and a maximum value of the samples,
creating a plurality of bins having respective data ranges; and
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping,
assigning samples from the stored streaming data to the plurality of bins; and
assigning the samples to the bins, based on values of the samples and data ranges of the bins, and determining numbers of samples within the bins;
populating the bins with synthetic data, the populating comprising:
populating the bins with synthetic data, the populating comprising:
generating, by a synthetic data generator, a plurality of synthetic data points; and
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network,
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins; and
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator;
creating a processed dataset based on the populated bins.
creating a processed dataset based on the bins; storing the processed dataset on a second storage device; and releasing the first storage device capacity occupied by the stored data.
Instant Claim 31 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 30, wherein analyzing the plurality of bins comprises determining minimum and maximum values for each bin.
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping, assigning the samples to the bins, based on values of the samples and data ranges of the bins, and determining numbers of samples within the bins;
Instant Claim 32 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 30, wherein the processing is based on a window size specified by a processing threshold.
receiving a data stream from an outside source; storing data from the received data stream in a first storage device; monitoring a size of the stored data in the first storage device; determining whether the size of the stored data has reached a first threshold; ... creating a processed dataset based on the bins; storing the processed dataset on a second storage device; and releasing the first storage device capacity occupied by the stored data. [Window/iteration: No express counterpart in the reference claim; obvious variation (see rationale in rejection).]
Instant Claim 33 ↔ U.S. Pat. No. 10,860,892 Claim 13
The system of claim 32, wherein the stored streaming data includes at least one of image data, video data, or audio data.
[Claim 13] wherein the data stream comprises image data. [Claim 14] wherein the data stream comprises video data.
Instant Claim 34 ↔ U.S. Pat. No. 10,860,892 Claim 12
The system of claim 30, wherein the stored streaming data includes at least one of salary information, age information, or tax information.
[Claim 12] wherein the data stream comprises numerical data. [Claim 16] wherein the data stream contains sensitive data.
Instant Claim 35 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 30, wherein the plurality of synthetic data points are generated based on a generator threshold specifying one or more data points that are generated at respective iterations.
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network, ... assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator; [Per-iteration generator threshold: No express counterpart in the reference claim; obvious variation (see rationale in rejection).]
Instant Claim 36 ↔ U.S. Pat. No. 10,860,892 Claim 1
The system of claim 30, wherein the generator threshold is pre-set or set based on at least one of determined minimum and maximum values of the assigned samples or a number of the assigned samples.
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network, ... assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator; [Pre-set threshold: No express counterpart in the reference claim; obvious variation (see rationale in rejection).]
Instant Claim 37 ↔ U.S. Pat. No. 10,860,892 Claim 17
A method for synthetic data generation comprising:
A method for synthetic data generation comprising:
determining a size of stored streaming data has reached a first threshold;
receiving a data stream from an outside source; storing data from the received data stream in a first storage device; monitoring a size of the stored data in the first storage device; determining whether the size of the stored data has reached a first threshold;
in response to the size determination, processing the stored streaming data, the processing comprising:
in response to the size determination, processing the stored data, the processing comprising: determining a total number of samples in the stored data, determining a minimum value and a maximum value of the samples,
creating a plurality of bins, having respective data ranges; and
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping,
assigning samples from the stored streaming data to the plurality of bins; and
assigning the samples to the bins, based on values of the samples and data ranges of the bins, and determining numbers of samples within the bins;
populating the bins with synthetic data, the populating comprising:
populating the bins with synthetic data, the populating comprising:
generating, by a synthetic data generator, a plurality of synthetic data points; and
generating, by a synthetic data generator, a plurality of synthetic data points using a neural network,
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins; and
assigning the synthetic data points to the bins based on values of the synthetic data points and data ranges of the bins, determining a total number of the samples within the data range, the total number comprising real and synthetic data points, and determining whether the total number has reached a second threshold; in response to the determination of whether the total number has reached a second threshold, pausing the synthetic data generator;
creating a processed dataset based on the populated bins.
creating a processed dataset based on the bins; storing the processed dataset on a second storage device; and releasing the first storage device capacity occupied by the stored data.
Instant Claim 38 ↔ U.S. Pat. No. 10,860,892 Claim 17
The method of claim 37, wherein the plurality of bins includes edge widths defining data ranges for each of the plurality of bins.
creating a plurality of bins, the bins having data ranges between the minimum and maximum sample values without overlapping,
Claims 29 and 39-40 are rejected on the ground of nonstatutory obviousness-type double patenting as being unpatentable over claims 1, 3 and 17 of U.S. Patent No. 10,860,892 in view of Improved Histograms for Selectivity Estimation of Range Predicates to Poosala as previously cited. Patent claims 1 and 17 recite creating a plurality of bins having data ranges, assigning samples and synthetic data points to the bins based on the data ranges of the bins, and creating a processed dataset based on the bins (Patent claim 3: the processed dataset is a histogram), which renders obvious instant claims 28, 37 and 38 as set forth above. The Patent claims do not recite that the edge widths of the bins are configured to minimize error associated with approximating an original distribution (instant claims 29 and 39) or that each bin has a different edge width (instant claim 40). Poosala teaches v-optimal histograms whose bucket boundaries over contiguous attribute values are placed so as to minimize the variance, i.e., the error, of the histogram's approximation of the underlying data distribution, "The v-optimal(V,F) and v-optimal(V,A) histograms minimize the variances in frequencies and areas respectively, while grouping contiguous attribute values" (Poosala: p. 298, Section 6.1; see also p. 297, Section 4.5), and teaches equi-depth and v-optimal buckets whose boundaries are not placed at equal intervals, so that the buckets span value ranges of differing widths (Poosala: p. 297, Section 4.3; p. 298, Section 6.1). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to configure the bins of the Patent claims with Poosala's error-minimizing, variable-width bucket boundaries, because Poosala teaches that conventional histograms are not sufficiently accurate, "None of the histograms described above are sufficiently accurate for general use in range query result size estimation" (Poosala: p. 297, Section 5), and a histogram that more accurately approximates the distribution of the stored data yields synthetic data that better preserves that distribution. This is the simple substitution of one known bin-boundary scheme for another to obtain predictable results. MPEP 2143(I)(B). Poosala was published in June 1996 and is prior art under 35 U.S.C. 102(a)(1).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALAN CHEN whose telephone number is (571)272-4143. The examiner can normally be reached M-F 10-7.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kamran Afshar can be reached at (571) 272-7796. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ALAN CHEN/Primary Examiner, Art Unit 2125
Prosecution Insights