DETAILED ACTION
Notice of 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 .
Priority
Regarding U.S. Patent App. No. 16/100,328 (filed 8/10/2018), Applicant’s claim for the benefit of a prior-filed application under 35 U.S.C. 120 is acknowledged.
Preliminary Amendment
The Preliminary Amendment submitted on 7/19/2024 has been considered. Claims 1-8 are cancelled and claims 19-38 are pending.
Information Disclosure Statement
The information disclosure statement submitted on 2/26/2024 has been considered.
Drawings
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:
In Fig. 3, region 346 is not mentioned in the specification. The examiner suggests amending para.0042 in the instant specification so that “region 344” reads as “region 346”.
In Fig. 5, box 524 is not mentioned in the specification.
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.
Abstract
While this is not an objection, the examiner invites Applicant to amend the Abstract to reflect the new claim set.
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 19-38 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Step 1 of the Alice/Mayo framework, Claims 19-33 are directed to a system (a process), and Claims 34-38 are directed to a method (a process), which each fall within one of the four statutory categories of inventions.
Regarding Claim 19
Step 2A, prong 1 (Is the claim directed to a law of nature, a natural phenomenon or an abstract idea).
Claim 19 recites the following mental processes, that in each case under the broadest reasonable interpretation, covers performance of the limitation in the mind (including an observation, evaluation, judgment, opinion) or with the aid of pencil and paper but for the recitation of generic computer components (e.g., “processors”, “memory”, “predictive model”, and “graphical user interface”).
generate, ... predicted resource values for the plurality of geographic regions over a time interval subsequent to the plurality of time intervals; (under the broadest reasonable interpretation, a human can mentally predict resource values for 2 or more geographic regions for a period of time, such as a salesman predicting sales prices for a widget for the U.S. and Mexico markets in January 2027)
determine, for the plurality of geographic regions, indices of resource elasticity based on a comparison of the predicted resource values for the time interval and empirical resource values for the plurality of geographic regions over the time interval (under the broadest reasonable interpretation, a human salesman can mentally determine for 2 or more geographic regions (U.S. and Mexico), indices of resource elasticity for a widget (e.g., how elastic sales will be based on price) for January 2027 based on empirical values for a previous generation widget in the same areas)
Step 2A, prong 2 (Does the claim recite additional elements that integrate the judicial exception into a practical application?).
The judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements (e.g., “processors”, “memory”, “predictive model”, and “graphical user interface”) which are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
Regarding the “A system, comprising: one or more processors, coupled with memory” limitation, such limitations are recited at a high-level of generality and amount to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional elements of processors and memories. These additional elements are recited at a high-level of generality and amount to no more than mere instructions to apply the exception using generic computer components (processors and memories). Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Regarding the “retrieve, from the memory, one or more datasets comprising vectors of features indicative of rates of changes of resources in a plurality of geographic regions over a plurality of time intervals” limitation, such additional element of a data gathering step is recited at a high level of generality and amounts to extra-solution activity of receiving data, i.e. pre-solution activity of gathering data for use in the claimed process (see MPEP 2106.05(g)).
Regarding the “construct, using a neural network comprising a plurality of connection nodes, a predictive model based on the one or more datasets” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a neural network trained as a predictive model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a neural network). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Regarding the “... using the predictive model ... ” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a generic predictive model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a generic predictive model). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Regarding the “display, via a graphical user interface, graphical indications of the plurality of geographic regions arranged in accordance with the indices of resource elasticity determined based on the comparison of the predicted resource values for the time interval and the empirical resource values” limitation, such limitation amounts to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)).
Accordingly, at Step 2A, prong two, after considering all claim elements individually and as an ordered combination, it is determined that the claims do not integrate the judicial exception into a practical application.
Step 2B (Does the claim recite additional elements that amount to significantly more than the judicial exception?)
In accordance with Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the additional elements (e.g., “processors”, “memory”, “predictive model”, and “graphical user interface”) are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using a generic computer component (See MPEP 2106.05(f)).
Regarding the “A system, comprising: one or more processors, coupled with memory” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding the “retrieve, from the memory, one or more datasets comprising vectors of features indicative of rates of changes of resources in a plurality of geographic regions over a plurality of time intervals” limitation, as discussed above, the additional element of a data gathering step is recited at a high level of generality and amounts to extra-solution activity of receiving data, i.e. pre-solution activity of gathering data for use in the claimed process. The courts have found limitations directed to obtaining information electronically, recited at a high level of generality, to be well-understood, routine, and conventional (see MPEP 2106.05(d)(II), “receiving or transmitting data over a network”, "electronic record keeping," and "storing and retrieving information in memory").
Regarding the “construct, using a neural network comprising a plurality of connection nodes, a predictive model based on the one or more datasets” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding the “... using the predictive model ... ” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding the “display, via a graphical user interface, graphical indications of the plurality of geographic regions arranged in accordance with the indices of resource elasticity determined based on the comparison of the predicted resource values for the time interval and the empirical resource values” limitation, this limitation amounts to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”)
Accordingly, at Step 2B after considering all claim elements individually and as an ordered combination, it is determined that the claims do not integrate the judicial exception into a practical application.
Regarding Claim 20
Step 2A, Prong 2
Regarding the “the neural network comprising at least 100 million connection nodes” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a neural network having a minimum number of nodes. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a neural network having a minimum number of nodes). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “the neural network comprising at least 100 million connection nodes” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 21
Step 2A, Prong 1
scrub the one or more datasets prior to formation of the predictive model based on the one or more datasets (under the broadest reasonable interpretation, a human such as a data scientist can scrub data from a data set prior to training a model, such as throwing pages of data away before they can be digitized and used to train the predictive model)
Step 2A, Prong 2
Regarding the “form the predictive model using the scrubbed one or more datasets” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a predictive model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a predictive model). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “form the predictive model using the scrubbed one or more datasets” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 22
Step 2A, Prong 1
modify or remove incomplete data to scrub the one or more datasets. (under the broadest reasonable interpretation, a human such as a data scientist can scrub data from a data set prior to training a model, such as throwing pages of incomplete data away before they can be digitized and used to train the predictive model)
Regarding Step 2A, Prong 2, the claim does not include any additional elements that integrate the judicial exception into a practical application and regarding Step 2B, there are no additional elements recited that amount to significantly more than the judicial exception.
Regarding Claim 23
Step 2A, Prong 1
convert, via one-hot encoding, text in the one or more datasets to numerical values (under the broadest reasonable interpretation, a human can take text in a dataset and convert it to vectors of one-hot encoding, where a 1 is placed for a binary yes, and a 0 is placed for a binary no, and such vector can be written on paper)
Step 2A, Prong 2
Regarding the “form the predictive model using the converted one or more datasets” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a predictive model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a predictive model). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “form the predictive model using the converted one or more datasets” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 24
Step 2A, Prong 1
test the predictive model based on a mean absolute error; (under the broadest reasonable interpretation, a human such as a statistician can take the outputs of a predictive model and mentally determine a mean absolute error, or alternatively, calculate such mean absolute error using pencil and paper)
determine, based on the test, that the predictive model satisfies a threshold (under the broadest reasonable interpretation, a human such as a statistician can mentally determine that the mean absolute error (and therefore the predictive model” satisfies a threshold of a small enough error rate)
Step 2A, Prong 2
Regarding the “apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a predictive model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a predictive model). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 25
Step 2A, Prong 1
test the predictive model based on a mean absolute error; (under the broadest reasonable interpretation, a human such as a statistician can take the outputs of a predictive model and mentally determine a mean absolute error, or alternatively, calculate such mean absolute error using pencil and paper)
determine, based on the test, that the predictive model does not satisfy a threshold; (under the broadest reasonable interpretation, a human such as a statistician can mentally determine that the mean absolute error (and therefore the predictive model” does not satisfy a threshold of a small enough error rate)
Step 2A, Prong 2
Regarding the “change, responsive to the determination that the predictive model does not satisfy the threshold, one or more hyperparameters used by the neural network to form the predictive model” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a neural network having tunable hyperparameters. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a neural network having tunable hyperparameters). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Regarding the “retrain the predictive model using the changed hyperparameters” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of retraining a model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (retraining a model). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “change, responsive to the determination that the predictive model does not satisfy the threshold, one or more hyperparameters used by the neural network to form the predictive model” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding the “retrain the predictive model using the changed hyperparameters” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 26
Step 2A, Prong 1
provide, based on the mean absolute error, an error rate between training data and test data (under the broadest reasonable interpretation, a human such as a statistician can mentally determine an error rate between training data and test data mean absolute error calculations)
compare the error rate with the threshold to determine that the predictive model does not satisfy the threshold. (under the broadest reasonable interpretation, a human such as a statistician can mentally compare the error rate with a threshold to determine whether the threshold is satisfied)
Regarding Step 2A, Prong 2, the claim does not include any additional elements that integrate the judicial exception into a practical application and regarding Step 2B, there are no additional elements recited that amount to significantly more than the judicial exception.
Regarding Claim 27
Step 2A, Prong 1
determine, using the mean absolute error, that the predictive model retrained using the changed hyperparameters satisfies the threshold (under the broadest reasonable interpretation, a human such as a statistician can mentally determine whether the threshold is satisfied by the retrained predictive model)
Step 2A, Prong 2
Regarding the “apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception. In particular, the claim only recites the additional element of a predictive model. This additional element is recited at a high-level of generality and amounts to no more than mere instructions to apply the exception using a generic computer component (a predictive model). Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation merely provides instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 28
Step 2A, Prong 2
Regarding the “wherein the one or more hyperparameters control a rate at which the predictive model learns patterns” limitation, this limitation merely describes a type of hyperparameter for a machine learning mode, and therefore such limitation amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use. As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception do not integrate a judicial exception into a practical application.
Step 2B
Regarding the “wherein the one or more hyperparameters control a rate at which the predictive model learns patterns” limitation, such limitation amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use as explained above, which does not amount to significantly more than the judicial exception. MPEP 2106.05(h).
Regarding Claim 29
Step 2A, Prong 1
rank order, based on the indices, the plurality of geographic regions to arrange the graphical indications of the plurality of geographic regions (under the broadest reasonable interpretation, a human can rank order data based on indices as explained in this limitation)
Step 2A, Prong 2
Regarding the “display the rank ordered arrangement of the graphical indications” limitation, such limitation amounts to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)).
Step 2B
Regarding the “display the rank ordered arrangement of the graphical indications” limitation, this limitation amounts to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”)
Regarding Claim 30
Step 2A, Prong 1
randomize selection of portions of the one or more datasets (under the broadest reasonable interpretation, a human can randomly shuffle pages of the dataset printed on paper)
Step 2A, Prong 2
Regarding the “construct the predictive model using the randomly selected portion of the one or more datasets to reduce bias in the predictive model” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation attempts to cover a solution to an identified problem with no restriction on how the result is accomplished, or provides no description of the mechanism for accomplishing the result. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (See MPEP 2106.05(f)).
Step 2B
Regarding the “construct the predictive model using the randomly selected portion of the one or more datasets to reduce bias in the predictive model” limitation, such limitation is recited at a high-level of generality and amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, because the limitation attempts to cover a solution to an identified problem with no restriction on how the result is accomplished, or provides no description of the mechanism for accomplishing the result. Accordingly, this additional element does not add significantly more than the judicial exception. (See MPEP 2106.05(f)).
Regarding Claim 31
Step 2A, Prong 2
Regarding the “wherein the display of the graphical indications arranged in accordance with the indices expands performance of an operation in a geographic region of the plurality of geographic regions” limitation, such limitation amounts to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)).
Step 2B
Regarding the “wherein the display of the graphical indications arranged in accordance with the indices expands performance of an operation in a geographic region of the plurality of geographic regions” limitation, this limitation amounts to extra solution activity because it is a mere nominal or tangential addition to the claim, amounting to mere data output (see MPEP 2106.05(g)). The courts have similarly found limitations directed to displaying a result, recited at a high level of generality, to be well-understood, routine, and conventional. See (MPEP 2106.05(d)(II), "presenting offers and gathering statistics.", “determining an estimated outcome and setting a price”)
Regarding Claim 32
Step 2A, Prong 2
Regarding the “perform a reinforcement learning technique to improve performance of the predictive model based on the empirical resource values” limitation, such limitation amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (specifically using reinforcement learning for machine learning). As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception do not integrate a judicial exception into a practical application.
Step 2B
Regarding the “perform a reinforcement learning technique to improve performance of the predictive model based on the empirical resource values” limitation, such limitation amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use as explained above, which does not amount to significantly more than the judicial exception. MPEP 2106.05(h).
Regarding Claim 33
Step 2A, Prong 2
Regarding the “wherein the reinforcement learning technique comprises Q- learning” limitation, such limitation amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (specifically using Q-learning for machine learning). As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible "simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use." Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception do not integrate a judicial exception into a practical application.
Step 2B
Regarding the “wherein the reinforcement learning technique comprises Q- learning” limitation, such limitation amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use as explained above, which does not amount to significantly more than the judicial exception. MPEP 2106.05(h).
Claim 34 recites a method that corresponds to the system of claim 19 and is therefore rejected for the same reasons explained above with respect to claim 19.
Claims 35-38 each depend from claim 34 and recite a method that corresponds to the systems of claims 21 and 23-25, respectively, and are therefore each rejected for the same reasons explained above with respect to claim 34 and claims 21, 23-25 respectively.
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.
Claims 19, 29, 31, and 34 are rejected under 35 U.S.C. 103 as being unpatentable over Baughman, Matt, et al. "Predicting Amazon spot prices with LSTM networks." Proceedings of the 9th workshop on scientific cloud computing. (June 11, 2018), hereinafter referenced as BAUGHMAN, in view of US 20100106982 A1, hereinafter referenced as CASTELLI, and further in view of US 20180336638 A1, hereinafter referenced as DZIABIAK.
Regarding Claim 19
BAUGHMAN teaches:
A system, comprising: ... (BAUGHMAN, pp. 1-2, section 1: “In this paper, we again treat the spot price as a time series and use a hybrid long/short term memory (LSTM)-dense neural network architecture to predict spot prices in the future.”)
retrieve, ... one or more datasets comprising vectors of features indicative ... of resources in a plurality of geographic regions over a plurality of time intervals; (BAUGHMAN, p. 2, section 3: “In order to improve prediction accuracy and our ability to apply machine learning models, we have collected spot pricing data from multiple sources spanning 3 years, with data for all instance types and availability zones in the US regions for the past year. This data is stored as “tick” data: it only includes a point in the series when the price changes from the previous price.”;
BAUGHMAN, p. 4, section 5.1: “In order to validate the use of neural networks for spot price prediction, we arbitrarily picked a single instance type from our historical pricing data: the c3.2xlarge Linux instance type from the us-east-1b region between September 3, 2016 and September 10, 2016 (as shown in Figure 1). Note: we used historical data as, at the time of this study, there was not sufficient pricing data available from the new spot market. To make the network computationally tractable, so that we could explore different network architectures, we used a subset of 10,000 points (roughly six days) for training and held back the following 2,000 points (roughly two days) for validation. While this data represents a relatively small sample, and only considers a single instance type, availability zone, and region, we believe that it is sufficient to provide initial validation of the benefits of LSTM models.”;
Examiner’s Note: While the authors of BAUGHMAN chose a subset of the data available to train the LSTM that only covers a single availability zone and 6 day interval, as disclosed by BAUGHMAN, the entire dataset including multiple geographic regions was available for training if desirable; the data is “tick” data stored as a time series (each corresponding to a recited “vector”) where the price (corresponding to recited “feature”) is tracked over time)
construct, using a neural network comprising a plurality of connection nodes, a predictive model based on the one or more datasets; (BAUGHMAN, pp. 1-2, section 1: “In this paper, we again treat the spot price as a time series and use a hybrid long/short term memory (LSTM)-dense neural network architecture to predict spot prices in the future.”;
BAUGHMAN, p. 3, section 4.2: “Based on our previous hyperparameter selection, our final RNN incorporates elements from both LSTM and dense neural networks. As illustrated in Figure 3, we use a simple, three-layer network composed of two LSTM layers—each 32 units wide—and one dense node to consolidate input from the second LSTM layer to a final predicted value. We chose the LSTM unit to comprise the primary layers of our neural network model due to the temporal element of our data and regression problem, as well as the benefits of LSTM over other types of RNNs.”
BAUGHMAN, p. 5, section 5.3: “After exploring various network architectures, we trained our three-layer LSTM network on the 10,000 input data points for 250 epochs.”;
Examiner’s Note: the LSTM has 32 units (nodes), and is used to predict spot prices)
generate, using the predictive model, predicted resource values for the plurality of geographic regions over a time interval subsequent to the plurality of time intervals; (BAUGHMAN, pp. 1-2, section 1: “In this paper, we again treat the spot price as a time series and use a hybrid long/short term memory (LSTM)-dense neural network architecture to predict spot prices in the future.”;
BAUGHMAN, p. 6, section 6: “In particular, we are interested in exploring more complex network architectures and using larger training sets across many instance types, availability zones, and regions. Our initial experimentation, albeit with what is now old data, using larger datasets indicates that more complex network architectures might well improve performance.”;
Examiner’s Note: BAUGHMAN discloses an LSTM that generates predicted cloud resource spot values (corresponding to recited “resource values”) for 1 geographic region for the future (corresponding to recited “time interval subsequent to the plurality of time intervals”); pursuant to MPEP 2144.04 VI.B, mere duplication of parts “has no patentable significance unless and a new and unexpected result is produced”, and one of ordinary skill would have found it obvious to train a second LSTM for a second geographic region so that predicted spot prices can be determined for more than 1 geographic region, which is supported by BAUGHMAN’s express desire for further research in extending the LSTM to cover additional availability zones)
determine, for the plurality of geographic regions, indices of resource elasticity based on a comparison of the predicted resource values for the time interval and empirical resource values for the plurality of geographic regions over the time interval; (BAUGHMAN, p. 1, section 1: “These unique characteristics have made spot instances particularly attractive in scientific computing scenarios as a way of obtaining elastic resources at low cost”
BAUGHMAN, p. 4, section 5:
PNG
media_image1.png
128
308
media_image1.png
Greyscale
Examiner’s Note: the MSE is based on comparing the actual and predicted values to determine forecast accuracy; under the broadest reasonable interpretation, the MSE corresponds to the recited “indices of resource elasticity” because the MSE concerns the accuracy of the spot price forecast relates to the elasticity of Amazon spot prices (e.g., lower prices make elasticity higher), and the accuracy of the prediction is an indicia of such elasticity (e.g., higher accuracy means more confidence about how elastic the price is))
However, BAUGHMAN fails to explicitly teach:
one or more processors, coupled with memory, to:
from the memory
... of rates of changes ...
... display, via a graphical user interface, graphical indications of the plurality of geographic regions arranged in accordance with the indices of resource elasticity determined based on the comparison of the predicted resource values for the time interval and the empirical resource values.
However, in a related field of endeavor (resource spot markets, see para. 0017), CASTELLI teaches and makes obvious:
one or more processors, coupled with memory, to: (CASTELLI, para. 0035: “As shown, the computer system 304 includes a central processing unit (CPU) 312, a memory 316, a bus 320, and input/output (I/O) interfaces 324.”;
Examiner’s Note: the BAUGHMAN-CASTELLI combination now implements the LSTM of BAUGHMAN using the CPU and memory explicitly taught by CASTELLI)
... from the memory ... (CASTELLI, para. 0035: “As shown, the computer system 304 includes a central processing unit (CPU) 312, a memory 316, a bus 320, and input/output (I/O) interfaces 324.”;
Examiner’s Note: the BAUGHMAN-CASTELLI combination now implements the LSTM of BAUGHMAN using the CPU and memory explicitly taught by CASTELLI, where training data from the datasets can be stored in the memory of CASTELLI)
retrieve, from the memory, one or more datasets comprising vectors of features indicative of rates of changes of resources in a plurality of geographic regions over a plurality of time intervals; (CASTELLI, para. 0021: “Examples of feeds, information and data appropriate to function as environmental data inputs (for example at 102 of FIG. 1 above) or as power-setting indicators (for example at 108 of FIG. 1 above) include a spot price of electricity, a rate of change of price of electricity, weather conditions, time of day, a facility (e.g. a household, building or company) energy use, census data associated with a specified energy use or demand, as well as other data relevant to energy costs. In another aspect the present invention may simultaneously and efficiently cascade power consumption reduction indicators or other data to a plurality of individual computers, server farms or computer systems, on a massive or even a global scale.”;
Examiner’s Note: CASTELLI discloses tracking the rate of change of spot prices; the BAUGHMAN-CASTELLI combination now takes the Amazon spot price data of BAUGHMAN and calculates the rate of change of such spot prices as CASTELLI to utilize such rate of change information as training data)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI as explained above. As disclosed by CASTELLI, one of ordinary skill would have been motivated to do so in order to take into consideration spot electricity prices into account when predicting the price of cloud services. (para. 0003).
However, BAUGHMAN and CASTELLI fail to explicitly teach:
display, via a graphical user interface, graphical indications of the plurality of geographic regions arranged in accordance with the indices of resource elasticity determined based on the comparison of the predicted resource values for the time interval and the empirical resource values.
However, in a related field of endeavor (analyzing prices with respect to geographic areas, see para. 0027), DZIABIAK teaches and makes obvious:
display, via a graphical user interface, graphical indications of the plurality of geographic regions arranged in accordance with the indices of resource elasticity determined based on the comparison of the predicted resource values for the time interval and the empirical resource values. (DZIABIAK, para.0186: “An example of such a user interface is described below with reference to FIG. 11. In some embodiments, the insurance options may be presented in the ranked order, for example, with user interface elements that, upon being selected by the user, cause the user computing device, for example its browser, to request content from the corresponding insurance provider server 16 described above.”;
DZIABIAK, para. 0192: “FIG. 11 shows an example of a user interface 280 in which a plurality of insurance options 282 are presented in ranked order. In some embodiments, the user interface 280 may include three or more, five or more, or 10 or more insurance options in ranked order 282, for example, each including an identifier of an insurance provider 286, and a link 284 that upon being selected by a user, causes the user's web browser to navigate to the website of the corresponding insurance provider.”;
DZIABIAK, para. 0208: “ranking the attributes based on the respective amounts of effects of the respective attributes on price of insurance for the user, wherein the instructions to present the subsequent user interface with visual elements indicating the respective amounts of effects of the respective attributes on price of insurance for the user comprises: instructing the user computing device to display identifiers of at least some of the attributes in ranked order.”;
Examiner’s Note: DZIABIAK discloses a user interface that displays price information in ranked order; the BAUGHMAN-CASTELLI-DZIABIAK combination now displays the predicted spot prices and accuracy rates of BAUGHMAN using the user interface of DZIABIAK, which displays information in a ranked order)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI and DZIABIAK as explained above. As disclosed by DZIABIAK, one of ordinary skill would have been motivated to do so because “the user may benefit from context provide by a comparison” between displayed scores. (para. 0088).
Regarding Claim 29
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN and CASTELLI fail to explicitly teach:
rank order, based on the indices, the plurality of geographic regions to arrange the graphical indications of the plurality of geographic regions; and display the rank ordered arrangement of the graphical indications.
However, in a related field of endeavor (analyzing prices with respect to geographic areas, see para. 0027), DZIABIAK teaches and makes obvious:
rank order, based on the indices, the plurality of geographic regions to arrange the graphical indications of the plurality of geographic regions; and display the rank ordered arrangement of the graphical indications. (DZIABIAK, para.0186: “An example of such a user interface is described below with reference to FIG. 11. In some embodiments, the insurance options may be presented in the ranked order, for example, with user interface elements that, upon being selected by the user, cause the user computing device, for example its browser, to request content from the corresponding insurance provider server 16 described above.”;
DZIABIAK, para. 0192: “FIG. 11 shows an example of a user interface 280 in which a plurality of insurance options 282 are presented in ranked order. In some embodiments, the user interface 280 may include three or more, five or more, or 10 or more insurance options in ranked order 282, for example, each including an identifier of an insurance provider 286, and a link 284 that upon being selected by a user, causes the user's web browser to navigate to the website of the corresponding insurance provider.”;
DZIABIAK, para. 0208: “ranking the attributes based on the respective amounts of effects of the respective attributes on price of insurance for the user, wherein the instructions to present the subsequent user interface with visual elements indicating the respective amounts of effects of the respective attributes on price of insurance for the user comprises: instructing the user computing device to display identifiers of at least some of the attributes in ranked order.”;
Examiner’s Note: DZIABIAK discloses a user interface that displays price information in ranked order; the BAUGHMAN-CASTELLI-DZIABIAK combination now displays the predicted spot prices and accuracy rates of BAUGHMAN using the user interface of DZIABIAK, which displays information in a ranked order)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI and DZIABIAK as explained above. As disclosed by DZIABIAK, one of ordinary skill would have been motivated to do so because “the user may benefit from context provide by a comparison” between displayed scores. (para. 0088).
Regarding Claim 31
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN and CASTELLI fail to explicitly teach:
wherein the display of the graphical indications arranged in accordance with the indices expands performance of an operation in a geographic region of the plurality of geographic regions.
However, in a related field of endeavor (analyzing prices with respect to geographic areas, see para. 0027), DZIABIAK teaches and makes obvious:
wherein the display of the graphical indications arranged in accordance with the indices expands performance of an operation in a geographic region of the plurality of geographic regions. (DZIABIAK, para.0186: “An example of such a user interface is described below with reference to FIG. 11. In some embodiments, the insurance options may be presented in the ranked order, for example, with user interface elements that, upon being selected by the user, cause the user computing device, for example its browser, to request content from the corresponding insurance provider server 16 described above.”;
DZIABIAK, para. 0192: “FIG. 11 shows an example of a user interface 280 in which a plurality of insurance options 282 are presented in ranked order. In some embodiments, the user interface 280 may include three or more, five or more, or 10 or more insurance options in ranked order 282, for example, each including an identifier of an insurance provider 286, and a link 284 that upon being selected by a user, causes the user's web browser to navigate to the website of the corresponding insurance provider.”;
DZIABIAK, para. 0208: “ranking the attributes based on the respective amounts of effects of the respective attributes on price of insurance for the user, wherein the instructions to present the subsequent user interface with visual elements indicating the respective amounts of effects of the respective attributes on price of insurance for the user comprises: instructing the user computing device to display identifiers of at least some of the attributes in ranked order.”;
Examiner’s Note: DZIABIAK discloses a user interface that displays price information in ranked order; the BAUGHMAN-CASTELLI-DZIABIAK combination now displays the predicted spot prices and accuracy rates of BAUGHMAN using the user interface of DZIABIAK, which displays information in a ranked order to enable a user to determine where spot prices are lower so that expanded cloud operational services can be ordered in particular geographical regions)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI and DZIABIAK as explained above. As disclosed by DZIABIAK, one of ordinary skill would have been motivated to do so because “the user may benefit from context provide by a comparison” between displayed scores. (para. 0088)
Claim 34 recites a method that corresponds to the system of claim 19 is and therefore rejected for the same reasons explained above with respect to claim 19.
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI and DZIABIAK and further in view of US 20190378007 A1, hereinafter referenced as MARKRAM.
Regarding Claim 20
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
the neural network comprising at least 100 million connection nodes.
However, in a related field of endeavor (artificial neural networks, see para. 0005), MARKRAM teaches and makes obvious:
the neural network comprising at least 100 million connection nodes. (MARKRAM, para. 0036: “For example, in some implementations, neural network devices can include hundreds of thousands, millions, or even billions of nodes. Thus, recurrent neural network device 100 can be a fraction of a larger recurrent artificial neural network (i.e., a subnetwork).”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK combination now modifies the LSTM of BAUGHMAN so that it includes over 100 million nodes as in MARKRAM)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and MARKRAM as explained above. One of ordinary skill would have been motivated to do so in order to increase the complexity of an LSTM in order to make more powerful predictions.
Claims 21-22 and 35 are rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI and DZIABIAK and further in view of US 20170236441 A1, hereinafter referenced as CHUANG.
Regarding Claim 21
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
scrub the one or more datasets prior to formation of the predictive model based on the one or more datasets; and
form the predictive model using the scrubbed one or more datasets.
However, in a related field of endeavor (machine learning algorithms, see paras. 0059-0060), CHUANG teaches and makes obvious:
scrub the one or more datasets prior to formation of the predictive model based on the one or more datasets; and (CHUANG, para. 0054: “Pre-processing: Data in the data sets may include errors, data loss and incomplete data which should be removed so as to cancel the influence from the interference and inconsistent data. Pre-processing also processes data of different data formats to make the processed data have a consistent data format.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-CHUANG combination now scrubs the Amazon spot price datasets of BAUGHMAN using the techniques of CHUANG)
form the predictive model using the scrubbed one or more datasets. (CHUANG, para. 0054: “Pre-processing: Data in the data sets may include errors, data loss and incomplete data which should be removed so as to cancel the influence from the interference and inconsistent data. Pre-processing also processes data of different data formats to make the processed data have a consistent data format.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-CHUANG combination now scrubs the Amazon spot price datasets of BAUGHMAN using the techniques of CHUANG and then trains the LSTM of BAUGHMAN using the pre-processed training data)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and CHUANG as explained above. As disclosed by CHUANG, one of ordinary skill would have been motivated to do so in order to ensure that processed data has a consistent data format. (para. 0054). As disclosed by CHUANG, one of ordinary skill would further have been motivated to do so in order to remove incomplete data’s “influence from the interference and inconsistent data.” (para. 0054).
Regarding Claim 22
BAUGHMAN, CASTELLI, DZIABIAK, and CHUANG disclose the system of claim 21 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
modify or remove incomplete data to scrub the one or more datasets.
However, in a related field of endeavor (machine learning algorithms, see paras. 0059-0060), CHUANG teaches and makes obvious:
modify or remove incomplete data to scrub the one or more datasets. (CHUANG, para. 0054: “Pre-processing: Data in the data sets may include errors, data loss and incomplete data which should be removed so as to cancel the influence from the interference and inconsistent data. Pre-processing also processes data of different data formats to make the processed data have a consistent data format.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-CHUANG combination now scrubs the Amazon spot price datasets of BAUGHMAN using the techniques of CHUANG to remove incomplete data)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and CHUANG as explained above. As disclosed by CHUANG, one of ordinary skill would have been motivated to do so in order to ensure that processed data has a consistent data format. (para. 0054). As disclosed by CHUANG, one of ordinary skill would further have been motivated to do so in order to remove incomplete data’s “influence from the interference and inconsistent data.” (para. 0054).
Claim 35 depends from claim 34 and recites a method that corresponds to the system of claim 21, and is therefore rejected for the same reasons explained with respect to claims 21 and 34.
Claims 23 and 36 are rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI and DZIABIAK and further in view of US 20190392257 A1, hereinafter referenced as FOLEY.
Regarding Claim 23
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
convert, via one-hot encoding, text in the one or more datasets to numerical values; and form the predictive model using the converted one or more datasets.
However, in a related field of endeavor (machine learning, including neural networks, see para. 0031), FOLEY teaches and makes obvious:
convert, via one-hot encoding, text in the one or more datasets to numerical values; and form the predictive model using the converted one or more datasets. (FOLEY, para. 0031: “For example, the training data collected from the interface may be used to construct multiple binary machine learning classifier models, such as one per item of a given checklist, for example. In one such example implementation, the text from the implicitly captured pages on view when checklist ticks were made, and/or the more fine-grained feedback in the form of user-selected text may be used an input for such modules. To make the textual data more amenable to a machine learning model, a data preprocessing pipeline may be constructed, to convert the text data into matrices of numbers, for example, or other data formats. For example, a one-hot encoding approach, a word embedding approach, and/or other textual analysis tools may be used to convert the relatively unstructured, natural language text into a data format more readily processed by a machine learning model.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-FOLEY combination now uses the teachings of FOLEY to encode the Amazon spot price training data of BAUGHMAN using one-hot encoding, and then uses such training data to train the LSTM of BAUGHMAN)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and FOLEY as explained above. As disclosed by FOLEY, one of ordinary skill would have been motivated to do so in order to “make the textual data more amenable to a machine learning model.” (para. 0031).
Claim 36 depends from claim 34 and recites a method that corresponds to the system of claim 23, and is therefore rejected for the same reasons explained with respect to claims 23 and 34.
Claims 24-25, 27, and 37-38 are rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI and DZIABIAK and further in view of US 11182691 B1, hereinafter referenced as ZHANG.
Regarding Claim 24
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
test the predictive model based on a mean absolute error; determine, based on the test, that the predictive model satisfies a threshold; and
apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals.
However, in a related field of endeavor (machine learning services, see col. 4, lines 40-43), ZHANG teaches and makes obvious:
test the predictive model based on a mean absolute error; determine, based on the test, that the predictive model satisfies a threshold; and (ZHANG, col. 62, lines 40-58: “If the maximum limit on the number of iterations performed has been reached (as detected in element 4616), no more iterations may be performed (element 4619) and the training may be concluded. If the maximum limit on iterations has not been reached, and if the error metrics (e.g., mean absolute error computed for the training set) resulting from training using the new sample has improved relative to the previous iteration (as detected in element 4618), a new sampling and training iteration may be implemented. For example, the error values from the just-completed iteration may be normalized and used as weights for the next iteration, and operations corresponding to element 4607 onwards may be repeated. If the improvement in the error metrics is below a threshold, or if there is no improvement in the error metrics (as also detected in element 4618), no more sampling iterations may be needed (element 4619), and the training of the model may also be concluded.”;
Examiner’s Note: ZHANG discloses training a machine learning model using mean absolute error (MAE) as an error metric, and comparing such error metric to a threshold; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG combination now modifies BAUGHMAN to use MAE instead of MSE, and compares the MAE to a threshold to determine when to stop iteratively training the LSTM of BAUGHMAN)
apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals. (ZHANG, col. 62, lines 40-58: “If the maximum limit on the number of iterations performed has been reached (as detected in element 4616), no more iterations may be performed (element 4619) and the training may be concluded. If the maximum limit on iterations has not been reached, and if the error metrics (e.g., mean absolute error computed for the training set) resulting from training using the new sample has improved relative to the previous iteration (as detected in element 4618), a new sampling and training iteration may be implemented. For example, the error values from the just-completed iteration may be normalized and used as weights for the next iteration, and operations corresponding to element 4607 onwards may be repeated. If the improvement in the error metrics is below a threshold, or if there is no improvement in the error metrics (as also detected in element 4618), no more sampling iterations may be needed (element 4619), and the training of the model may also be concluded.”;
Examiner’s Note: ZHANG discloses training a machine learning model using mean absolute error (MAE) as an error metric, and comparing such error metric to a threshold; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG combination now modifies BAUGHMAN to use MAE instead of MSE, and compares the MAE to a threshold to determine when to stop iteratively training the LSTM of BAUGHMAN, and once the LSTM of BAUGHMAN is trained, using the LSTM of BAUGHMAN to predict Amazon spot prices at future intervals)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and ZHANG as explained above. As disclosed by ZHANG, one of ordinary skill would have been motivated to do so in order to determine when the model has been sufficiently trained, which will conserve resources as opposed to training for a predetermined number of epochs. (col. 62, lines 53-57).
Regarding Claim 25
BAUGHMAN, CASTELLI, and DZIABIAK disclose the system of claim 19 as explained above. BAUGHMAN further teaches:
change, ... one or more hyperparameters used by the neural network to form the predictive model; and (BAUGHMAN, p. 3, section 4.2: “Based on our previous hyperparameter selection, our final RNN incorporates elements from both LSTM and dense neural networks”;
Examiner’s Note: BAUGHMAN teaches that hyperparameter selection is determined for the neural network)
retrain the predictive model using the changed hyperparameters. (BAUGHMAN, p. 3, section 4.2: “Based on our previous hyperparameter selection, our final RNN incorporates elements from both LSTM and dense neural networks”;
Examiner’s Note: BAUGHMAN teaches that hyperparameter selection is determined for the neural network and based on such hyperparameter selection (such as the number of layers), the LSTM is iteratively retrained)
However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
test the predictive model based on a mean absolute error; determine, based on the test, that the predictive model satisfies a threshold; and
... responsive to the determination that the predictive model satisfies the threshold ...
However, in a related field of endeavor (machine learning services, see col. 4, lines 40-43), ZHANG teaches and makes obvious:
test the predictive model based on a mean absolute error; determine, based on the test, that the predictive model satisfies a threshold; and (ZHANG, col. 62, lines 40-58: “If the maximum limit on the number of iterations performed has been reached (as detected in element 4616), no more iterations may be performed (element 4619) and the training may be concluded. If the maximum limit on iterations has not been reached, and if the error metrics (e.g., mean absolute error computed for the training set) resulting from training using the new sample has improved relative to the previous iteration (as detected in element 4618), a new sampling and training iteration may be implemented. For example, the error values from the just-completed iteration may be normalized and used as weights for the next iteration, and operations corresponding to element 4607 onwards may be repeated. If the improvement in the error metrics is below a threshold, or if there is no improvement in the error metrics (as also detected in element 4618), no more sampling iterations may be needed (element 4619), and the training of the model may also be concluded.”;
Examiner’s Note: ZHANG discloses training a machine learning model using mean absolute error (MAE) as an error metric, and comparing such error metric to a threshold; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG combination now modifies BAUGHMAN to use MAE instead of MSE, and compares the MAE to a threshold to determine when to stop iteratively training the LSTM of BAUGHMAN)
... responsive to the determination that the predictive model satisfies the threshold ... (ZHANG, col. 62, lines 40-58: “If the maximum limit on the number of iterations performed has been reached (as detected in element 4616), no more iterations may be performed (element 4619) and the training may be concluded. If the maximum limit on iterations has not been reached, and if the error metrics (e.g., mean absolute error computed for the training set) resulting from training using the new sample has improved relative to the previous iteration (as detected in element 4618), a new sampling and training iteration may be implemented. For example, the error values from the just-completed iteration may be normalized and used as weights for the next iteration, and operations corresponding to element 4607 onwards may be repeated. If the improvement in the error metrics is below a threshold, or if there is no improvement in the error metrics (as also detected in element 4618), no more sampling iterations may be needed (element 4619), and the training of the model may also be concluded.”;
Examiner’s Note: ZHANG discloses training a machine learning model using mean absolute error (MAE) as an error metric, and comparing such error metric to a threshold; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG combination now modifies BAUGHMAN to use MAE instead of MSE, and compares the MAE to a threshold to determine when to stop iteratively training the LSTM of BAUGHMAN)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and ZHANG as explained above. As disclosed by ZHANG, one of ordinary skill would have been motivated to do so in order to determine when the model has been sufficiently trained, which will conserve resources as opposed to training for a predetermined number of epochs. (col. 62, lines 53-57).
Regarding Claim 27
BAUGHMAN, CASTELLI, DZIABIAK, and ZHANG disclose the system of claim 25 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
determine, using the mean absolute error, that the predictive model retrained using the changed hyperparameters satisfies the threshold; and (ZHANG, col. 62, lines 40-58: “If the maximum limit on the number of iterations performed has been reached (as detected in element 4616), no more iterations may be performed (element 4619) and the training may be concluded. If the maximum limit on iterations has not been reached, and if the error metrics (e.g., mean absolute error computed for the training set) resulting from training using the new sample has improved relative to the previous iteration (as detected in element 4618), a new sampling and training iteration may be implemented. For example, the error values from the just-completed iteration may be normalized and used as weights for the next iteration, and operations corresponding to element 4607 onwards may be repeated. If the improvement in the error metrics is below a threshold, or if there is no improvement in the error metrics (as also detected in element 4618), no more sampling iterations may be needed (element 4619), and the training of the model may also be concluded.”;
Examiner’s Note: ZHANG discloses training a machine learning model using mean absolute error (MAE) as an error metric, and comparing such error metric to a threshold; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG combination now modifies BAUGHMAN to use MAE instead of MSE, and compares the MAE to a threshold after the LSTM of BAUGHMAN has been re-trained to determine when to stop iteratively re-training the LSTM of BAUGHMAN)
apply, responsive to the determination that the predictive model satisfies the threshold, the predictive model to the time interval subsequent to the plurality of time intervals. (ZHANG, col. 62, lines 40-58: “If the maximum limit on the number of iterations performed has been reached (as detected in element 4616), no more iterations may be performed (element 4619) and the training may be concluded. If the maximum limit on iterations has not been reached, and if the error metrics (e.g., mean absolute error computed for the training set) resulting from training using the new sample has improved relative to the previous iteration (as detected in element 4618), a new sampling and training iteration may be implemented. For example, the error values from the just-completed iteration may be normalized and used as weights for the next iteration, and operations corresponding to element 4607 onwards may be repeated. If the improvement in the error metrics is below a threshold, or if there is no improvement in the error metrics (as also detected in element 4618), no more sampling iterations may be needed (element 4619), and the training of the model may also be concluded.”;
Examiner’s Note: ZHANG discloses training a machine learning model using mean absolute error (MAE) as an error metric, and comparing such error metric to a threshold; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG combination now modifies BAUGHMAN to use MAE instead of MSE, and uses the re-trained LSTM of BAUGHMAN to predict spot prices for future time intervals)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and ZHANG as explained above. As disclosed by ZHANG, one of ordinary skill would have been motivated to do so in order to determine when the model has been sufficiently trained, which will conserve resources as opposed to training for a predetermined number of epochs. (col. 62, lines 53-57).
Claim 37 depends from claim 34 and recites a method that corresponds to the system of claim 24, and is therefore rejected for the same reasons explained with respect to claims 24 and 34.
Claim 38 depends from claim 34 and recites a method that corresponds to the system of claim 25, and is therefore rejected for the same reasons explained with respect to claims 25 and 34.
Claim 26 is rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI, DZIABIAK, and ZHANG and further in view of US 20180172667 A1, hereinafter referenced as NOSKOV.
Regarding Claim 26
BAUGHMAN, CASTELLI, DZIABIAK, and ZHANG disclose the system of claim 25 as explained above. However, BAUGHMAN, CASTELLI, DZIABIAK, and ZHANG fail to explicitly teach:
provide, based on the mean absolute error, an error rate between training data and test data; and compare the error rate with the threshold to determine that the predictive model does not satisfy the threshold.
However, in a related field of endeavor (machine learning, see para. 0005), NOSKOV teaches and makes obvious:
provide, based on the mean absolute error, an error rate between training data and test data; and compare the error rate with the threshold to determine that the predictive model does not satisfy the threshold. (NOSKOV, para. 0222: “The error-bars in the plots were the calculated using standard deviations, expecting lower RMSE values for the training set (Train) as compared to the validation/test sets (Test).”;
Examiner’s Note: NOSKOV teaches comparing the error rates between a training set and a test/validation set; the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG-NOSKOV combination now compares the error rates of the training and test sets of BAUGHMAN, using the MAE error metric of ZHANG, and compares the results to a threshold as in ZHANG)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, ZHANG, and NOSKOV as explained above. One of ordinary skill would have been motivated to do so in order to determine if the training has resulted in overfitting. (para. 0222).
Claim 28 is rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI, DZIABIAK, and ZHANG and further in view of US 20190377984 A1 hereinafter referenced as GHANTA.
Regarding Claim 28
BAUGHMAN, CASTELLI, DZIABIAK, and ZHANG disclose the system of claim 25 as explained above. However, BAUGHMAN, CASTELLI, DZIABIAK, and ZHANG fail to explicitly teach:
wherein the one or more hyperparameters control a rate at which the predictive model learns patterns.
However, in a related field of endeavor (machine learning, see para. 0001), GHANTA teaches and makes obvious:
wherein the one or more hyperparameters control a rate at which the predictive model learns patterns. (GHANTA, para. 0072: “As used herein, a hyper-parameter search, optimization, or tuning is the problem of choosing a set of optimal hyper-parameters for a learning algorithm. In certain embodiments, the same kind of machine learning model can require different constraints, weights, or learning rates to generalize different data patterns. These measures may be called hyper-parameters, and may be tuned so that the model can optimally solve the machine learning problem.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-ZHANG-GHANTA combination now modifies BAUGHMAN to search for different learning rate parameters as in GHANTA).
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, ZHANG, and GHANTA as explained above. As disclosed by GHANTA, one of ordinary skill would have been motivated to do so in order to tune the model so that it “can optimally solve the machine learning problem.” (para. 0072).
Claim 30 is rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI, and DZIABIAK and further in view of US 20140379619 A1, hereinafter referenced as PERMEH.
Regarding Claim 30
BAUGHMAN, CASTELLI, DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
randomize selection of portions of the one or more datasets; and construct the predictive model using the randomly selected portion of the one or more datasets to reduce bias in the predictive model.
However, in a related field of endeavor (machine learning, see para. 0002), PERMEH teaches and makes obvious:
randomize selection of portions of the one or more datasets; and construct the predictive model using the randomly selected portion of the one or more datasets to reduce bias in the predictive model. (PERMEH, para. 0126: “The training set generator can be configured to attempt to reduce any specific bias in the selection of samples for the training set from the general sample population. The training set generator can do this by randomly selecting training samples from the population.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-PERMEH combination now modifies the training of the LSTM of BAUGHMAN to randomly sample from the Amazon spot price historical data in order to reduce bias as in PERMEH)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and PERMEH as explained above. As disclosed by PERMEH, one of ordinary skill would have been motivated to do so in order to “reduce any specific bias in the selection of samples for the training set from the general sample population.” (para. 0126).
Claims 32-33 are rejected under 35 U.S.C. 103 as being unpatentable over BAUGHMAN, in view of CASTELLI, and DZIABIAK and further in view of US 20190283745 A1, hereinafter referenced as NAGEL.
Regarding Claim 32
BAUGHMAN, CASTELLI, DZIABIAK disclose the system of claim 19 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
perform a reinforcement learning technique to improve performance of the predictive model based on the empirical resource values.
However, in a related field of endeavor (machine learning, see para. 0001), NAGEL teaches and makes obvious:
perform a reinforcement learning technique to improve performance of the predictive model based on the empirical resource values. (NAGEL, para. 0045: “The learning unit can perform various types of machine learning, such as supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, transduction, multitask learning, etc. In a preferred embodiment, the learning unit performs reinforcement learning using Q-learning. The learning unit for performing reinforcement learning can includes a reward computing unit for computing a reward based on at least one type of the measured condition parameters, which are monitored by the state observing unit, and a function updating unit (artificial intelligence) for updating a function, for example, an action-value function (action-value table) for deciding, from the measured risk value at present, based on the reward computed by the reward computing unit, at least one of an future measured risk probability parameter, wherein the function updating unit may update other functions.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-NAGEL combination now modifies the training of the LSTM of BAUGHMAN to use reinforcement learning based on measured values as in NAGEL)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and NAGEL as explained above. As disclosed by NAGEL, one of ordinary skill would have been motivated to do so in order to maximize a reward amount. (para. 0045).
Regarding Claim 33
BAUGHMAN, CASTELLI, DZIABIAK, and NAGEL disclose the system of claim 32 as explained above. However, BAUGHMAN, CASTELLI, and DZIABIAK fail to explicitly teach:
wherein the reinforcement learning technique comprises Q- learning.
However, in a related field of endeavor (machine learning, see para. 0001), NAGEL teaches and makes obvious:
wherein the reinforcement learning technique comprises Q- learning. (NAGEL, para. 0045: “The learning unit can perform various types of machine learning, such as supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, transduction, multitask learning, etc. In a preferred embodiment, the learning unit performs reinforcement learning using Q-learning. The learning unit for performing reinforcement learning can includes a reward computing unit for computing a reward based on at least one type of the measured condition parameters, which are monitored by the state observing unit, and a function updating unit (artificial intelligence) for updating a function, for example, an action-value function (action-value table) for deciding, from the measured risk value at present, based on the reward computed by the reward computing unit, at least one of an future measured risk probability parameter, wherein the function updating unit may update other functions.”;
Examiner’s Note: the BAUGHMAN-CASTELLI-DZIABIAK-NAGEL combination now modifies the training of the LSTM of BAUGHMAN to use Q-learning based on measured values as in NAGEL)
Before the effective filing date of the present application, it would have been obvious for one of ordinary skill in the art to combine the teachings of BAUGHMAN with CASTELLI, DZIABIAK, and NAGEL as explained above. As disclosed by NAGEL, one of ordinary skill would have been motivated to do so in order to maximize a reward amount. (para. 0045).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 20200005204 A1 (Kumar). “The report 302 includes numbers and graphical representation of the evolution of the professionals, the number of job posts identified in this period for machine learning, a hiring difficulty index, and the median compensation (together with respective growth indicators over the previous year).” (para. 0054).
US 20190266544 A1 (Mandala). “The chart 402 may be represented by a percentage of applicants by location (e.g., country, state, hemisphere, continent, region). The chart 402 may alternatively show a percentage hired from a total number of applicants per department, where the percentage hired may indicate how popular a department 348 is for applicants to apply to. The screen 400 may include a chart 404 to present results related to a hiring amount (e.g., a total number, a count) of applicants by location. In addition, the screen 400 may include a chart 406 to present a predictive model of hiring for the enterprise. The chart 406 may aggregate results from the data analytics routine of the server system 218 to showcase a prediction related to hiring of the enterprise. As shown, the chart 406 may include actual (e.g., past, previous, historic) hiring data and predictions for future hiring data. It is noted that the server system 218 runs a data analytics routine at a timeframe A and displays the results at a timeframe B, so that the prediction for future hiring is based on data analytics routines performed at the timeframe A and thus the prediction for future hiring is inclusive of the predictions related to timeframe B. In other words, the server system 218 may refresh and/or perform the data analytics routine on pre-determined time intervals, in response to a trigger, in response to a change in an update time (e.g., update time 376), and the like to cause the predictions for future hiring to update. The chart 406 may include predictions on hiring for the enterprise, hiring for respective departments, hiring based on aspects of profiles of applicants (e.g., predictions based on skills listed in the profiles), and the like. The predictions may be used to predict and/or estimate an amount of new hires for future hiring.” (para. 0110).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MICHAEL C LEE whose telephone number is (571)272-4933. The examiner can normally be reached M-F 12:00 pm - 8:00 pm ET.
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, Omar Fernandez Rivas can be reached at 571-272-2589. 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.
/MICHAEL C. LEE/Examiner, Art Unit 2128
Prosecution Insights