DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . This action is responsive to claims 1-20 filed 10/22/2025.
Allowable Subject Matter
Claim(s) 8-10 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for allowance for the indication of allowable subject matter:
Svyatkovskiy (US 20200249918 A1) discloses: a method of preserving signals in data inputs with moderate to high levels of variances in data sequence lengths for artificial neural network model training (fig.4 gives overview of preserving training signal data relevancy for artificial neural network training), the method comprising:
accessing, by a processor, a data series containing a plurality of data samples that are organized in a first order within the data series (fig.4:404, 0058), wherein each of the data samples contains a plurality of data values that are sequentially arranged within that respective data sample, wherein the data samples have a plurality of sequence lengths that vary from each other based on the data values in those respective data samples (ibid);
segmenting, by the processor, the data series organized in the second order according to a window size such that the data series are segmented into a plurality of data columns sized based on the window size (fig.4:408, 0058);
constructing, by the processor, a plurality of virtual batches from the modified data series such that the virtual batches (a) each has a plurality of rows according to a batch size, (b) each is sized according to the window size and the batch size, (c) have the rows list all of the data values from the modified data series (fig.4:408, 0058: each sampling window constitutes a virtual batch).
Yoon ("A simple distortion-free method to handle variable length sequences for recurrent neural networks in text dependent speaker verification", published 2020) discloses: a method of preserving signals in data inputs with moderate to high levels of variances in data sequence lengths for artificial neural network model training (§1: removing information distortion due to variable length data for RNNs), the method comprising:
accessing, by a processor, a data series containing a plurality of data samples that are organized in a first order within the data series, wherein each of the data samples contains a plurality of data values that are sequentially arranged within that respective data sample, wherein the data samples have a plurality of sequence lengths that vary from each other based on the data values in those respective data samples (§2.4 (p.5) contemplates 3 sequences of varying length);
rearranging, by the processor, the data samples within the data series such that the data samples are organized in a second order within the data series, wherein the first order is different from the second order (§2.4 ¶2: sorting sequences by length);
segmenting, by the processor, the data series organized in the second order according to a window size such that the data series are segmented into a plurality of data columns sized based on the window size (ibid: sequences are segmented into subscript-indexed portions);
constructing, by the processor, a plurality of virtual batches from the modified data series such that the virtual batches, (b) each is sized according to the window size and the batch size (§2.4 ¶2, fig.3: PackedSequence constructed based on a concatenation based on window size), (c) have the rows list all of the data values from the modified data series (ibid: all data is represented).
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.
Claim(s) 11-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. Claim 11 recites a system comprising a server to perform the recited steps. However, a server may be purely software (e.g., a virtual server as in 0035). Hence, the entire system may be directed to a software construct, which is non-statutory. Explicit recitation of hardware components (e.g., processor) would suffice to overcome this rejection. Dependent claims 12-20 are rejected for failing to cure the deficiency of the parent claim by dependency.
Claim(s) 1-7, 11-17 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The 35 U.S.C. 101 subject matter eligibility analysis first asks whether the claim is directed to one of the four statutory categories (Step 1). It next asks whether the claim is directed to an abstract idea (Step 2A), via Prong 1, whether an abstract idea (e.g., mathematical concept, mental process, certain methods of organizing human activity) is recited, and Prong 2, whether it is integrated into a practical application. It finally asks whether the claim as a whole includes additional elements that amount to significantly more than the judicial exception (Step 2B). See MPEP 2106.
STEP 1: The claims falls within one of the four statutory categories:
Claims 1-10 recite methods. Claims 11-20 recite systems that, under BRI, could be interpreted as software only. However, in the interests of compact prosecution, the remaining steps are analyzed below.
STEP 2A PRONG 1: The claims recite a judicial exception:
The claims are directed to normalizing or preparing data of varying length for use in machine learning applications, such as RNNs. Individual serial data is grouped into minibatches and randomized, partitioned into windows, and finally densely packed into a regular minibatch for without padding. However, all these data arrangement operations for packing and concatenating data are mental processes, sequencing and arrangement operations which may be performed in the mind (2A-1). Hence, the claims are directed to a mental process. In particular:
For Claim 1: A method of preserving signals in data inputs with moderate to high levels of variances in data sequence lengths for artificial neural network model training, the method comprising:
accessing, by a processor, a data series containing a plurality of data samples that are organized in a first order within the data series, wherein each of the data samples contains a plurality of data values that are sequentially arranged within that respective data sample, wherein the data samples have a plurality of sequence lengths that vary from each other based on the data values in those respective data samples (Accessing data for arrangement is a mental process that may be performed with the mind or with the aid of pen and paper, the additional limitations place restrictions on the form of the data but do not preclude the performance of the method in the mind);
randomizing, by the processor, the data samples within the data series such that the data samples are organized in a second order within the data series, wherein the first order is different from the second order (Randomizing or rearrange may be performed in the mind or with the aid of pen and paper);
segmenting, by the processor, the data series organized in the second order according to a window size such that the data series are segmented into a plurality of data columns sized based on the window size (Segmenting or dividing the data series may be performed in the mind or with the aid of pen and paper);
removing, by the processor, all of the data values of each respective data sample that does not satisfy the window size within each respective data column such that a modified data series remains (Removing extraneous values may be performed in the mind or with the aid of pen and paper); and
constructing, by the processor, a plurality of virtual batches from the modified data series such that the virtual batches (a) each has a plurality of rows according to a batch size, (b) each is sized according to the window size and the batch size, (c) have the rows list all of the data values from the modified data series, and (d) have each of the rows sequentially contain the data values from only one of the data samples of the modified data series (Constructing virtual batches, i.e., regrouping and concatenating the data samples into new structures may be performed in the mind or with the aid of pen and paper).
For Claim 2: The method of claim 1, wherein the virtual batches are consecutive, wherein at least two of the rows of the virtual batches contain the data values from the one of the data samples from at least two of the data columns when the virtual batches are viewed as if positioned immediately adjacent to each other and the rows are consecutive (This places limitations on the arrangement configuration but does not preclude performance of the technique in the mind).
For Claim 3: The method of claim 1, wherein at least one of the virtual batches that is not an initial virtual batch is constructed at least by determining whether the data sample in the modified data series is exhausted for all of the data columns other than an initial data column (Determining an exhaustion state of a data column in order to construct a batch may be performed in the mind or with the aid of pen and paper).
For Claim 5: The method of claim 4, wherein the at least one of the virtual batches that is not the initial virtual batch has an initial row that is populated with the data values from the next unused row in the initial data column (This limitation on batch construction does not preclude performance of the technique in the mind).
For Claim 6: The method of claim 5, wherein the at least one of the virtual batches that is not the initial virtual batch is constructed at least by population from the data values of the data sample other than from the initial data column (This limitation on batch construction does not preclude performance of the technique in the mind).
For Claim 7: The method of claim 6, wherein the at least one of the virtual batches that is not the initial virtual batch is constructed at least by population from the data values from one of the data columns that immediately follows the initial data column (This limitation on batch construction does not preclude performance of the technique in the mind).
The remaining claims 11-17 recite analogous systems and are rejected for the same reasons.
STEP 2A PRONG 2: The claims do not integrate the exception into a practical application:
The additional elements (underlined above) are implementation on a computer processor. However, this constitutes mere instructions to implement an abstract idea on a computer. It does not meaningfully limit the practice of the abstract idea and hence does not constitute an integration into a practical application.
STEP 2B: The claim as a whole do not include additional elements that amount to significantly more than the abstract idea:
The use of computer processors is well-understood, routine and conventional (WURC) and hence does not constitute significantly more.
Response to Arguments
In the remarks, Applicant argued:
1. (p.7) The express language requires machine implementation, and cannot be performed practically in the human mind, particularly at he scale and complexity required for neural network model training.
Examiner respectfully disagrees. Under BRI, the claims merely recite the ordering and rearranging of blocks of information and do not make reference to processes that are not practical to be performed in the mind.
2. Claims 1, 11 provide a technical solution to a recognized problem and are integrated into a practical application, and result in improved accuracy and efficiency.
Examiner respectfully disagrees. Under BRI, the claims merely recite the ordering and rearranging of data, implemented on a processor, and do not integrate this data into neural network applications. For example, the mere performance of mental processes on a processor do not constitute an application into practical application and do not function to improve accuracy and efficiency.
3. These technical steps constitute significantly more.
Examiner respectfully disagrees. Under BRI, the claims merely recite the ordering and rearranging of data, implemented on a processor, and do not integrate this data into neural network applications. For example, the mere performance of mental processes on a processor do does not constitute something that is not well known in the art of machine learning.
4. The steps in claim 1 are not generic data manipulation but rather constitute a technical solution to a technical problem.
Examiner respectfully disagrees. The claims at present merely recite the mentally-practical rearranging and manipulation of data, and do not make reference to solving a technical problem.
5. Similar to Example 47, claims 1, 11 recite, technical steps of randomizing, segmenting, removing, and constructing virtual batches for data series with variable lengths.
Examiner respectfully disagrees. Unlike Example 47, the current limitations merely recite data shaping or manipulation of a nature that is performable in the mind without any reference to practical applications of this data.
6. Similar to Example 48, claims 1, 11 recite, technical steps of randomizing, segmenting, removing, and constructing virtual batches for data series with variable lengths.
Examiner respectfully disagrees. Unlike Example 47, the current limitations merely recite data shaping or manipulation of a nature that is performable in the mind without any reference to practical applications of this data.
7. Similar to Example 49, claims 1, 11 recite, technical steps of randomizing, segmenting, removing, and constructing virtual batches for data series with variable lengths.
Examiner respectfully disagrees. Unlike Example 47, the current limitations merely recite data shaping or manipulation of a nature that is performable in the mind without any reference to practical applications of this data.
8. (§4) Federal register update of July 2024 clarifies that reciting specific technical solutions to problems in AI/ML data handling, such as bathing, windowing, and state management, are not abstract ideas when the steps are technologically rooted and cannot be practically performed in the human mind.
Examiner respectfully disagrees. Unlike the examples given in the federal register, the current limitations merely recite data shaping or manipulation of a nature that is performable in the mind without any reference to practical applications of this data, such as in AI / ML.
9. (§5) The specifications and drawings express technical problem solved by the present invention, as addressing issues of information loss, undefined behavior, and poor performance. These technical steps are not generic or abstract, but are expressly claimed and supported by the Specifications as a solution to a recognized technical problem in the field.
Examiner respectfully disagrees. The currently recited limitations are merely directed to mentally-practical data shaping, without any reference to integration into AI or ML applications. Although such data shaping may have potential applications as outlined in the Specifications, as claimed, they are not integrated into any AI / ML process and hence are do not constitute practical applications.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Brueckner (US 10318882 B2) fig.18, 20 disclose chunking ingested data for machine learning applications.
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LIANG LI whose telephone number is (303)297-4263. The examiner can normally be reached Mon-Fri 9-12p, 3-11p MT (11-2p, 5-1a ET).
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor Jennifer Welch can be reached on (571)272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center and the Private Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from Patent Center or Private PAIR. Status information for unpublished applications is available through Patent Center or Private PAIR to authorized users only. Should you have questions about access to Patent Center or the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
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. The examiner is available for interviews Mon-Fri 6-11a, 2-7p MT (8-1p, 4-9p ET).
/LIANG LI/
Primary examiner AU 2143