DETAILED ACTION
This Office Action is sent in response to the Applicant’s Communication received on 10/26/2023 for application number 18/384,314. The Office hereby acknowledges receipt of the following and placed of record in file: Specification, Drawings, Abstract, Oath/Declaration, IDS, and Claims.
Claims 1-20 are pending.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1
Step 2A Prong 1:
Claim 1 recites:
“evaluating the test data [using a pre-trained deep neural network];” Evaluating the test is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“using compressed histograms to determine p-values for the extracted activations;” Using compressed histograms to determine p-values for the extracted activations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“evaluating the determined p-values;” Evaluating the determined p-values is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“retaining portions of the test data that are determined as being anomalous based at least in part on the evaluation of the p-values;” Retaining portions of the test data that are determined as being anomalous based at least in part on the evaluation of the p-values is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“A computer-implemented method;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“receiving a new set of test data;” Mere data gathering recited at a high level of generality, and thus are insignificant extra-solution activity (MPEP 2106.05(g)).
“using a pre-trained deep neural network;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“A computer-implemented method;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) and cannot provide an inventive concept.
“receiving a new set of test data;” Mere data gathering recited at a high level of generality, and thus are insignificant extra-solution activity. See MPEP 2106.05(g). The additional element of “receiving” does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of receiving steps amounts to no more than mere data gathering. This element amounts to receiving data over a network and are well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II (i). This cannot provide an inventive concept.
“using a pre-trained deep neural network;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 2
Step 2A Prong 1:
Claim 2 recites:
“for each of the extracted activations: determining a node of the deep neural network that a given extracted activation corresponds to;” Determining a node of the deep neural network that a given extracted activation corresponds to is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“comparing the given extracted activation to a compressed histogram correlated with the node that the given extracted activation corresponds to;” Comparing the given extracted activation to a compressed histogram correlated with the node that the given extracted activation corresponds to is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“Computing a p-value for the given extracted activation;” Computing a p-value for the given extracted activation is a claim that merely uses textual replacements for particular equations, and is therefore a mathematical concept.
Step 2A Prong Two and Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
Claim 3
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“wherein the compressed histograms are node-specific;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“wherein the compressed histograms are node-specific;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 4
Step 2A Prong 1:
Claim 4 recites:
“estimating a number of the extracted activations that are unexpected;” Estimating a number of the extracted activations that are unexpected is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“using the estimated number to determine whether at least a portion of the test data is anomalous;” Using the estimated number to determine whether at least a portion of the test data is anomalous is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data, and (ii) the corresponding extracted activations;” Discarding the portion of the test data and the corresponding extracted activations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two and Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
Claim 5
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“in response to determining that another portion of the test data is anomalous, storing (i) the anomalous portion of the test data, and (ii) the corresponding extracted activations;” This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“in response to determining that another portion of the test data is anomalous, storing (i) the anomalous portion of the test data, and (ii) the corresponding extracted activations;” These elements amount to storing… information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93; See MPEP 2106.05(d) (II)(iv). The courts have recognized the computer functions of storing as well‐understood, routine, and conventional function when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 6
Step 2A Prong 1:
Claim 6 recites:
“producing the compressed histograms by: evaluating training data using the deep neural network;” Producing the compressed histograms by: evaluating training data using the deep neural network is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“using the extracted training activations to generate the compressed histograms;” Using the extracted training activations to generate the compressed histograms is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“extracting training activations from layers of the deep neural network in response to evaluating the training data;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
“storing the compressed histograms;” This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“extracting training activations from layers of the deep neural network in response to evaluating the training data;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
“storing the compressed histograms;” These elements amount to storing… information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93; See MPEP 2106.05(d) (II)(iv). The courts have recognized the computer functions of storing as well‐understood, routine, and conventional function when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 7
Step 2A Prong 1:
Claim 7 recites:
“for each node of the deep neural network that corresponds to a subset of the extracted training activations, using the subset of the extracted training activations to create one of the compressed histograms;” Using the subset of the extracted training activations to create one of the compressed histograms is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two and Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
Claim 8
Step 2A Prong 1:
Claim 8 recites:
“sorting a respective training activation into a respective bin of the compressed histograms based on a numerical value of the respective training activation;” Sorting a respective training activation into a respective bin of the compressed histograms based on a numerical value of the respective training activation is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two and Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
Claim 9
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“wherein a number of bins of the compressed histograms is determined based on a maximum of (i) an output of a Freedman Diaconis estimator, and (ii) an output of a Sturges estimator used to evaluate the training activations;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“wherein a number of bins of the compressed histograms is determined based on a maximum of (i) an output of a Freedman Diaconis estimator, and (ii) an output of a Sturges estimator used to evaluate the training activations;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 10
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“causing the deep neural network to be retrained using the retained portions of the test data that are determined as being anomalous;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“causing the deep neural network to be retrained using the retained portions of the test data that are determined as being anomalous;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 11
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“wherein the determination of anomalous data portions is implemented using non-parametric scan statistics;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“wherein the determination of anomalous data portions is implemented using non-parametric scan statistics;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 12
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“wherein the non-parametric scan statistics include a higher criticism statistic;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“wherein the non-parametric scan statistics include a higher criticism statistic;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 13
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“wherein the p-values determined for the extracted activations include ranges of p-values for the respective extracted activations;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“wherein the p-values determined for the extracted activations include ranges of p-values for the respective extracted activations;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 14
Step 2A Prong 1:
Claim 14 recites:
“evaluate the test data [using a pre-trained deep neural network];” Evaluating the test is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“use compressed histograms to determine p-values for the extracted activations;” Using compressed histograms to determine p-values for the extracted activations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“evaluate the determined p-values;” Evaluating the determined p-values is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“retain portions of the test data that are determined as being anomalous based at least in part on the evaluation of the p-values;” Retaining portions of the test data that are determined as being anomalous based at least in part on the evaluation of the p-values is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a processor, executable by the processor, or readable and executable by the processor;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“receiving a new set of test data;” Mere data gathering recited at a high level of generality, and thus are insignificant extra-solution activity (MPEP 2106.05(g)).
“using a pre-trained deep neural network;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a processor, executable by the processor, or readable and executable by the processor;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) and cannot provide an inventive concept.
“receiving a new set of test data;” Mere data gathering recited at a high level of generality, and thus are insignificant extra-solution activity. See MPEP 2106.05(g). The additional element of “receiving” does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of receiving steps amounts to no more than mere data gathering. This element amounts to receiving data over a network and are well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II (i). This cannot provide an inventive concept.
“using a pre-trained deep neural network;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 15-17 are computer program product claims that recite similar limitations to method claims 2-4, respectively. Therefore, claims 15-17 are rejected using the same rationale as claims 2-4 respectively.
Claim 18
Step 2A Prong 1:
Claim 18 recites:
“evaluate the test data [using a pre-trained deep neural network];” Evaluating the test is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“use compressed histograms to determine p-values for the extracted activations;” Using compressed histograms to determine p-values for the extracted activations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“evaluate the determined p-values;” Evaluating the determined p-values is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“retain portions of the test data that are determined as being anomalous based at least in part on the evaluation of the p-values;” Retaining portions of the test data that are determined as being anomalous based at least in part on the evaluation of the p-values is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“A system, comprising: a processor; and logic executable by the processor;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“receiving a new set of test data;” Mere data gathering recited at a high level of generality, and thus are insignificant extra-solution activity (MPEP 2106.05(g)).
“using a pre-trained deep neural network;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“A system, comprising: a processor; and logic executable by the processor;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) and cannot provide an inventive concept.
“receiving a new set of test data;” Mere data gathering recited at a high level of generality, and thus are insignificant extra-solution activity. See MPEP 2106.05(g). The additional element of “receiving” does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of receiving steps amounts to no more than mere data gathering. This element amounts to receiving data over a network and are well-understood, routine, conventional activity. See MPEP 2106.05(d), subsection II (i). This cannot provide an inventive concept.
“using a pre-trained deep neural network;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 19
Step 2A Prong 1:
Claim 19 recites:
“estimating a number of the extracted activations that are unexpected;” Estimating a number of the extracted activations that are unexpected is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“using the estimated number to determine whether at least a portion of the test data is anomalous;” Using the estimated number to determine whether at least a portion of the test data is anomalous is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data, and (ii) the corresponding extracted activations;” Discarding the portion of the test data and the corresponding extracted activations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“in response to determining that another portion of the test data is anomalous, storing (i) the anomalous portion of the test data, and (ii) the corresponding extracted activations;” This limitation is merely a post-solution step of storing the data—a nominal addition to the claim that does not meaningfully limit the claim. The method storing is recited at a high level of generality. Simply implementing the abstract idea in a generic method is not a practical application of the abstract idea. Therefore, storing step is an insignificant extra-solution activity. See MPEP 2106.05(g).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“in response to determining that another portion of the test data is anomalous, storing (i) the anomalous portion of the test data, and (ii) the corresponding extracted activations;” These elements amount to storing… information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93; See MPEP 2106.05(d) (II)(iv). The courts have recognized the computer functions of storing as well‐understood, routine, and conventional function when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 20 is a system claim that recites similar limitations to method claim 2. Therefore, claim 20 is rejected using the same rationale as claim 2.
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.
Claim(s) 1, 10, 11, 13, 14, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Rahmes et al. (US 20260179352 A1), hereinafter Rahmes, in view of Cintas et al. (US 20220138584 A1), hereinafter Cintas, and Saxena et al. (Feature Extraction using Symbolic Dynamic Filtering for Fault Analysis in Distribution Systems, published 2020), hereinafter Saxena.
Regarding claim 1, Rahmes teaches,
A computer-implemented method [Para 0002, the present disclosure relates to implementing systems and methods], comprising: receiving a new set of test data [Para 0157, the object detection device 255 receives inputs 258 (e.g., pixel intensities, high maps, CAD model data, etc.) and labeled truth data features];
evaluating (Para 0214, determine the optimal waveform parameters (Block 426) and provide a real or near real-time success assessment) the test data using a pre-trained deep neural network (Para 0152, fully convolutional-deconvolutional network trained end-to-end with semantic segmentation) [Para 0108, the machine-learning based approach uses expert systems and/or deep learning to facilitate signal recognition and classification. The deep learning can be implemented by one or more neural networks (e.g., … Convolutional Neural Network(s) (CNN(s))); Para 0152, The present approach may apply a fully convolutional-deconvolutional network trained end-to-end with semantic segmentation to classify land use/land cover features, for example; Para 0157, the object detection device 255 receives inputs 258 (e.g., pixel intensities, high maps, CAD model data, etc.) and labeled truth data features 259 which are provided to GAN and testing modules 260, 261; Para 0214, raw I/Q data 420 and labeled truth data features 421 are provided to a GAN training module 422 and a testing module 423, similar to those described above. The output of the testing module 423 is provided to the quantum computing circuit 417, which in the illustrated example performs model fusion with the reward matrix (Block 424) and subset summing optimization (Block 425). The output of the quantum computing circuit 417 is used to determine the optimal waveform parameters (Block 426) and provide a real or near real-time success assessment];
using histograms to determine p-values for the extracted activations [Para 0159, Each bin's histogram values are added from left to right, and observation values are compared to determine a P-value for the observation];
evaluating the determined p-values [Para 0159, One may use superposition to simultaneously calculate and compare a P-value with each cluster].
Rahmes teaches the above limitations of claim 1 including the test data (Para 0157); the deep neural network (Para 0108), evaluating the test data (Para 0214), and the evaluation of the p-values (Para 0159).
Rahmes does not teach extracting activations from layers of neural network in response to evaluating data; histograms being compressed histograms; and retaining portions of data that are determined as being anomalous based at least in part on evaluation.
Cintas teaches,
extracting activations from layers of neural network in response to evaluating data [Para 0062, In analyzing data for anomalousness, discriminator 116 first obtains samples 122 for evaluation. Discriminator 116 then extracts activations 124 from the samples to compute empirical p-value values 134; Para 0111, Table 2 shows precision and recall metrics for an example of group subset scanning over three convolutional layers of a discriminator using a GAN improvement pipeline 430 in accordance with the present disclosure]; and
retaining portions of data that are determined as being anomalous based at least in part on evaluation [Para 0064, Anomalous nodes, samples, and evaluation metrics are extracted 142. Anomalous nodes, samples, and metrics may be displayed in the output 150 with one or more indicators to identify anomalous activity].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide extracting activations and retaining anomalous data in order to [Cintas, para 0058] to better evaluate and classify data.
Rahmes-Cintas does not teach histograms being compressed histograms.
Saxena teaches,
histograms being compressed histograms [Abstract, Feature extraction for fault detection and classification using Symbolic Dynamic Filtering (SDF) is explored in this paper. It provides an edge over existing methodologies by compressing voluminous waveform data into probability histograms].
Saxena is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Saxena and provide compressed histograms [Saxena, Sect III, para 2] reduce the computational burden on the classifier.
Regarding claim 10, Rahmes-Cintas-Saxena teach the limitations of claim 1 including the deep neural network (Rahmes, Para 0108).
Cintas further teaches,
causing neural network to be retrained using the retained portions of the test data that are determined as being anomalous [Para 0105, A gradient mask may be generated and a partial update of the GAN completed 438. A gradient mask developed using the normal distribution may efficiently retrain filters containing anomalous nodes. The gradient mask may be used to retrain the model. By retraining the model with a gradient mask based on the normal distribution, the system 400 may perform partial updates to the GAN using detected anomalous nodes].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide retraining with anomalous data in order to [Cintas, para 0105] better generate artificial samples and to better classify samples.
Regarding claim 11, Rahmes-Cintas-Saxena teach the limitations of claim 1.
Cintas further teaches,
wherein the determination of anomalous data portions is implemented using non-parametric scan statistics [Para 0062, In analyzing data for anomalousness, discriminator 116 first obtains samples 122 for evaluation. Discriminator 116 then extracts activations 124 from the samples to compute empirical p-value values 134. Nonparametric scan statistics (NPSS) are maximized 140. Anomalous nodes, samples, and evaluation metrics are extracted 142 from the maximized NPSS 140].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide using non-parametric scan statistics for outlier resistance when processing anomalous data.
Regarding claim 13, Rahmes-Cintas-Saxena teach the limitations of claim 1.
Cintas further teaches,
wherein the p-values determined for the extracted activations include ranges of p-values for the respective extracted activations [Para 103, Group-based subset scanning is then applied 434. Group-based subset scanning 434 provides anomalous nodes via unsupervised machine learning. The matrix of activations may be converted into a matrix of empirical p-values or p-value ranges corresponding to the proportion of activations from a background set of activations that are larger than activations from an evaluation set or input].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide ranges of p-values for activations in order to create a simple summary of data that easily understandable to human users.
Regarding claim 14, Rahmes teaches,
A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a processor, executable by the processor, or readable and executable by the processor [Para 0106, The implementing systems of method 900 may comprise a circuit (e.g., quantum registers, quantum adder circuits, and/or quantum comparator circuits), and/or a non-transitory computer-readable storage medium comprising programming instructions that are configured to cause the quantum processor to implement method 900; ], to cause the processor to:
receive a new set of test data [Para 0157, the object detection device 255 receives inputs 258 (e.g., pixel intensities, high maps, CAD model data, etc.) and labeled truth data features];
evaluate (Para 0214, determine the optimal waveform parameters (Block 426) and provide a real or near real-time success assessment) the test data using a pre-trained deep neural network (Para 0152, fully convolutional-deconvolutional network trained end-to-end with semantic segmentation) [Para 0108, the machine-learning based approach uses expert systems and/or deep learning to facilitate signal recognition and classification. The deep learning can be implemented by one or more neural networks (e.g., … Convolutional Neural Network(s) (CNN(s))); Para 0152, The present approach may apply a fully convolutional-deconvolutional network trained end-to-end with semantic segmentation to classify land use/land cover features, for example; Para 0157, the object detection device 255 receives inputs 258 (e.g., pixel intensities, high maps, CAD model data, etc.) and labeled truth data features 259 which are provided to GAN and testing modules 260, 261; Para 0214, raw I/Q data 420 and labeled truth data features 421 are provided to a GAN training module 422 and a testing module 423, similar to those described above. The output of the testing module 423 is provided to the quantum computing circuit 417, which in the illustrated example performs model fusion with the reward matrix (Block 424) and subset summing optimization (Block 425). The output of the quantum computing circuit 417 is used to determine the optimal waveform parameters (Block 426) and provide a real or near real-time success assessment];
use histograms to determine p-values for the extracted activations [Para 0159, Each bin's histogram values are added from left to right, and observation values are compared to determine a P-value for the observation];
evaluate the determined p-values [Para 0159, One may use superposition to simultaneously calculate and compare a P-value with each cluster]; and
Rahmes teaches the above limitations of claim 1 including the test data (Para 0157); the deep neural network (Para 0108), evaluating the test data (Para 0214), and the evaluation of the p-values (Para 0159).
Rahmes does not teach extract activations from layers of neural network in response to evaluating data; histograms being compressed histograms; and retain portions of data that are determined as being anomalous based at least in part on evaluation.
Cintas teaches,
extract activations from layers of neural network in response to evaluating data [Para 0062, In analyzing data for anomalousness, discriminator 116 first obtains samples 122 for evaluation. Discriminator 116 then extracts activations 124 from the samples to compute empirical p-value values 134; Para 0111, Table 2 shows precision and recall metrics for an example of group subset scanning over three convolutional layers of a discriminator using a GAN improvement pipeline 430 in accordance with the present disclosure]; and
retain portions of data that are determined as being anomalous based at least in part on evaluation [Para 0064, Anomalous nodes, samples, and evaluation metrics are extracted 142. Anomalous nodes, samples, and metrics may be displayed in the output 150 with one or more indicators to identify anomalous activity].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide extracting activations and retaining anomalous data in order to [Cintas, para 0058] to better evaluate and classify data.
Rahmes-Cintas does not teach histograms being compressed histograms.
Saxena teaches,
histograms being compressed histograms [Abstract, Feature extraction for fault detection and classification using Symbolic Dynamic Filtering (SDF) is explored in this paper. It provides an edge over existing methodologies by compressing voluminous waveform data into probability histograms].
Saxena is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Saxena and provide compressed histograms [Saxena, Sect III, para 2] reduce the computational burden on the classifier.
Regarding claim 18, Rahmes teaches,
A system [Para 0002, The present disclosure relates generally to quantum computing systems and associated algorithms], comprising: a processor [Para 0020, FIG. 9 is a flow diagram of an illustrative method for operating a quantum processor]; and logic executable by the processor [Para 0012, A related non-transitory computer-readable medium may have computer-executable instructions for causing a controller to cooperate with and at least one quantum computing circuit] to cause the processor to:
receive a new set of test data [Para 0157, the object detection device 255 receives inputs 258 (e.g., pixel intensities, high maps, CAD model data, etc.) and labeled truth data features];
evaluate (Para 0214, determine the optimal waveform parameters (Block 426) and provide a real or near real-time success assessment) the test data using a pre-trained deep neural network (Para 0152, fully convolutional-deconvolutional network trained end-to-end with semantic segmentation) [Para 0108, the machine-learning based approach uses expert systems and/or deep learning to facilitate signal recognition and classification. The deep learning can be implemented by one or more neural networks (e.g., … Convolutional Neural Network(s) (CNN(s))); Para 0152, The present approach may apply a fully convolutional-deconvolutional network trained end-to-end with semantic segmentation to classify land use/land cover features, for example; Para 0157, the object detection device 255 receives inputs 258 (e.g., pixel intensities, high maps, CAD model data, etc.) and labeled truth data features 259 which are provided to GAN and testing modules 260, 261; Para 0214, raw I/Q data 420 and labeled truth data features 421 are provided to a GAN training module 422 and a testing module 423, similar to those described above. The output of the testing module 423 is provided to the quantum computing circuit 417, which in the illustrated example performs model fusion with the reward matrix (Block 424) and subset summing optimization (Block 425). The output of the quantum computing circuit 417 is used to determine the optimal waveform parameters (Block 426) and provide a real or near real-time success assessment];
use histograms to determine p-values for the extracted activations [Para 0159, Each bin's histogram values are added from left to right, and observation values are compared to determine a P-value for the observation];
evaluate the determined p-values [Para 0159, One may use superposition to simultaneously calculate and compare a P-value with each cluster]; and
Rahmes teaches the above limitations of claim 1 including the test data (Para 0157); the deep neural network (Para 0108), evaluating the test data (Para 0214), and the evaluation of the p-values (Para 0159).
Rahmes does not teach extract activations from layers of neural network in response to evaluating data; histograms being compressed histograms; and retain portions of data that are determined as being anomalous based at least in part on evaluation.
Cintas teaches,
extract activations from layers of neural network in response to evaluating data [Para 0062, In analyzing data for anomalousness, discriminator 116 first obtains samples 122 for evaluation. Discriminator 116 then extracts activations 124 from the samples to compute empirical p-value values 134; Para 0111, Table 2 shows precision and recall metrics for an example of group subset scanning over three convolutional layers of a discriminator using a GAN improvement pipeline 430 in accordance with the present disclosure]; and
retain portions of data that are determined as being anomalous based at least in part on evaluation [Para 0064, Anomalous nodes, samples, and evaluation metrics are extracted 142. Anomalous nodes, samples, and metrics may be displayed in the output 150 with one or more indicators to identify anomalous activity].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide extracting activations and retaining anomalous data in order to [Cintas, para 0058] to better evaluate and classify data.
Rahmes-Cintas does not teach histograms being compressed histograms.
Saxena teaches,
histograms being compressed histograms [Abstract, Feature extraction for fault detection and classification using Symbolic Dynamic Filtering (SDF) is explored in this paper. It provides an edge over existing methodologies by compressing voluminous waveform data into probability histograms].
Saxena is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Saxena and provide compressed histograms [Saxena, Sect III, para 2] reduce the computational burden on the classifier.
Claim(s) 2, 15, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas and Saxena, and in further view of Leobandung (US 10741611 B1), hereinafter Leobandung and Choi (WO 2023003432 A1, see attached translation), hereinafter Choi.
Regarding claim 2, Rahmes-Cintas-Saxena teach the limitations of claim 1 including using histograms to determine the p-values for the extracted activations (Rahmes, Para 0159), compressed histograms (Saxena, Abstract), the deep neural network (Rahmes, Para 0108).
Cintas further teaches,
computing a p-value for the given extracted activation [Para 0062, Discriminator 116 then extracts activations 124 from the samples to compute empirical p-value values 134].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide computing p-values from activations in order to [Cintas, para 0049] determine anomalous nodes.
Rahmes-Cintas-Saxena do not teach for each of activations: determining a node of neural network that a given activation corresponds to; and comparing the given activation to a histogram correlated with the node that the given activation corresponds to.
Leobandung teaches,
for each of activations: determining a node of neural network that a given activation corresponds to [Col 4, lines 36-37, Each hidden node value is determined by applying an activation function].
Leobandung is analogous to the claimed invention as they both relate to utilizing activation functions in neural networks. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Leobandung and provide determining nodes from activations in order to determine the activeness of particular neurons.
Rahmes-Cintas-Saxena-Leobandung do not teach comparing the given activation to a histogram correlated with the node that the given activation corresponds to.
Choi teaches,
Comparing (Para 0028, analyzes) the given activation to a histogram correlated with the node that the given activation corresponds to [Para 0028, The conventional method for determining the quantization range analyzes the distribution of activations through histogram generation and determines the quantization range based on the distribution of activations; Para 0004, An artificial neural network (ANN) has a structure in which nodes representing artificial neurons are connected through synapses; Para 0024, Activations can be produced through the activation function of the nodes included in the neural network].
Choi is analogous to the claimed invention as they both relate to utilizing histograms correlated with neural network nodes. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Choi and provide comparing activations to histograms in order to create an efficient visualization for analyzing neuronal activity.
Claims 15 and 20 are computer program product and system claims, respectively, that recite similar limitations to method claim 2. Therefore, claims 15 and 20 are rejected using the same rationale as claim 2.
Claim(s) 3 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas and Saxena, and in further view of Song (KR 20190015160 A, see attached translation), hereinafter Song.
Regarding claim 3, Rahmes-Cintas-Saxena teach the limitations of claim 1 including compressed histograms (Saxena, Abstract).
Rahmes-Cintas-Saxena do not teach wherein histograms are node-specific.
Song teaches,
wherein histograms are node-specific [Para 0019, The computer program comprises:… a command to histogram feature values output from each of one or more hidden nodes of a hidden layer of the neural network].
Song is analogous to the claimed invention as they both relate to histograms for nodes/neurons. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Song and provide node-specific histograms in order to provide an efficient visualization for analyzing neuronal activity.
Claim 16 is computer program product claim that recites similar limitations to method claim 3. Therefore, claim 16 is rejected using the same rationale as claim 3.
Claim(s) 4, 17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas and Saxena, and in further view of Speakman et al. (Subset Scanning Over Neural Network Activations, published 2018), hereinafter Speakman, Fields et al. (US 20190087906 A1), hereinafter Fields, and Kim et al. (US 20210325957 A1), hereinafter Kim.
Regarding claim 4, Rahmes-Cintas-Saxena teach the limitations of claim 1 including evaluating the determined p-values (Rahmes, Para 0159), retaining portions of data that are determined as being anomalous based at least in part on evaluation (Cintas, Para 0064), and the corresponding extracted activations (Cintas, Para 0062).
Rahmes-Cintas-Saxena do not teach estimating a number of activations that are unexpected; using the estimated number to determine whether at least a portion of the test data is anomalous; and in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data, and (ii) the activations.
Speakman teaches,
estimating a number of activations that are unexpected [Sect 2, para 13The p-value ranges from an evaluation input are processed by a nonparametric scan statistic in order to identify the sub set of node activations that maximizes the scoring function…F(S), as this is the subset with the most statistical evidence for having been effected by an anomalous pattern];
using the estimated number to determine whether at least a portion of the test data is anomalous [Sect 2, para 17, Therefore, we assume an anomalous process will result in some S where the observed significance is higher than the expected].
Speakman is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Speakman and provide estimating a portion of anomalous data in order to [Speakman, Sect 1, para 8] to detect the presence of subtle patterns in high dimensional spaces.
Rahmes-Cintas-Saxena-Speakman do not teach in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data, and (ii) the activations.
Fields teaches,
in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data [Para 0027, the software application may discard or otherwise ignore the non-anomalous data if, for example, it is not deemed sufficiently valuable to transmit].
Fields is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Fields and provide discarding the non-anomalous data to enhance memory efficiency when processing anomalous data.
Rahmes-Cintas-Saxena-Speakman-Fields do not teach discarding (ii) the activations.
Kim teaches,
discarding (ii) the activations [Para 0075, if the activation data is less than a threshold value, then power control component 314 may instruct activation component 312 to discard the activation data].
Kim is analogous to the claimed invention as they both relate to processing activation data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Kim and provide discarding activations in order to fine tune a neural network discarding unnecessary neurons/nodes.
Claim 17 is computer program product claim that recites similar limitations to method claim 4. Therefore, claim 17 is rejected using the same rationale as claim 4.
Regarding claim 19, Rahmes-Cintas-Saxena teach the limitations of claim 18 including the evaluation of the determined p-values (Rahmes, Para 0159), retaining portions of data that are determined as being anomalous based at least in part on evaluation (Cintas, Para 0064), and the corresponding extracted activations (Cintas, Para 0062).
Cintas further teaches,
in response to determining that another portion of the test data is anomalous, store (i) the anomalous portion of the test data, and (ii) the corresponding extracted activations [Para 0101, extracting activations from a GAN 432, applying group-based subset scanning 434 to the extracted activations, obtaining anomalous samples and nodes 436, generating a gradient mask and partially updating the GAN therewith 438; Para 0029, the network structure is updated. Updating the network structure may include extracting activations from a layer of the discriminator. Group-based subset scanning over said activations may be applied to obtain anomalous nodes. A gradient mask may be generated with the anomalous nodes and applied to the discriminator. The process of extracting activations, applying group-based subset scanning, generating gradient masks, and applying the gradient masks to the discriminator may be repeated until a threshold is met].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide storing anomalous data and their activations in order to [Cintas, para 0101] improve generative adversarial networks.
Rahmes-Cintas-Saxena do not teach estimating a number of activations that are unexpected; using the estimated number to determine whether at least a portion of the test data is anomalous; and in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data, and (ii) the activations.
Speakman teaches,
estimating a number of activations that are unexpected [Sect 2, para 13The p-value ranges from an evaluation input are processed by a nonparametric scan statistic in order to identify the sub set of node activations that maximizes the scoring function…F(S), as this is the subset with the most statistical evidence for having been effected by an anomalous pattern];
using the estimated number to determine whether at least a portion of the test data is anomalous [Sect 2, para 17, Therefore, we assume an anomalous process will result in some S where the observed significance is higher than the expected].
Speakman is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Speakman and provide estimating a portion of anomalous data in order to [Speakman, Sect 1, para 8] to detect the presence of subtle patterns in high dimensional spaces.
Rahmes-Cintas-Saxena-Speakman do not teach in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data, and (ii) the activations.
Fields teaches,
in response to determining that a portion of the test data is not anomalous, discarding (i) the portion of the test data [Para 0027, the software application may discard or otherwise ignore the non-anomalous data if, for example, it is not deemed sufficiently valuable to transmit].
Fields is analogous to the claimed invention as they both relate to anomaly detection. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Fields and provide discarding the non-anomalous data to enhance memory efficiency when processing anomalous data.
Rahmes-Cintas-Saxena-Speakman-Fields do not teach discarding (ii) the activations.
Kim teaches,
discarding (ii) the activations [Para 0075, if the activation data is less than a threshold value, then power control component 314 may instruct activation component 312 to discard the activation data].
Kim is analogous to the claimed invention as they both relate to processing activation data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Kim and provide discarding activations in order to fine tune a neural network discarding unnecessary neurons/nodes.
Claim(s) 5 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas and Saxena, and in further view of Speakman et al. (Subset Scanning Over Neural Network Activations, published 2018), hereinafter Speakman, Fields et al. (US 20190087906 A1), hereinafter Fields, and Kim et al. (US 20210325957 A1), hereinafter Kim.
Regarding claim 5, Rahmes-Cintas-Saxena-Speakman-Fields-Kim teach the limitations of claim 4.
Cintas further teaches,
in response to determining that another portion of the test data is anomalous, storing (i) the anomalous portion of the test data, and (ii) the corresponding extracted activations [Para 0101, extracting activations from a GAN 432, applying group-based subset scanning 434 to the extracted activations, obtaining anomalous samples and nodes 436, generating a gradient mask and partially updating the GAN therewith 438; Para 0029, the network structure is updated. Updating the network structure may include extracting activations from a layer of the discriminator. Group-based subset scanning over said activations may be applied to obtain anomalous nodes. A gradient mask may be generated with the anomalous nodes and applied to the discriminator. The process of extracting activations, applying group-based subset scanning, generating gradient masks, and applying the gradient masks to the discriminator may be repeated until a threshold is met].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide storing anomalous data and their activations in order to [Cintas, para 0101] improve generative adversarial networks.
Claim(s) 6 is rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas and Saxena, and in further view of Dolan et al. (US 20200051327 A1), hereinafter Dolan.
Regarding claim 6, Rahmes-Cintas-Saxena teach the limitations of claim 1 including producing the compressed histograms [Saxena, Abstract].
Rahmes further teaches,
evaluating training data using the deep neural network [Para 0108, the machine-learning based approach uses expert systems and/or deep learning to facilitate signal recognition and classification. The deep learning can be implemented by one or more neural networks (e.g., … Convolutional Neural Network(s) (CNN(s))); Para 0154, a generative adversarial network (GAN) may be used to generate additional image data. Furthermore, discriminative algorithms may be used to classify input data, i.e., given the features of a data instance, they predict a label or category to which that data belongs]; and
storing the histograms [Para 0159, Each bil1 n's histogram values are added from left to right, and observation values are compared to determine a P-value for the observation; Note: further processing of the histogram implies that the histogram is stored].
Rahmes-Cintas-Saxena teaches the above limitations of claim 6 including the deep neural network (Claim 1: Rahmes, para 0108) and the training data (Claim 6: Rahmes, para 0154).
Rahmes does not teach extracting training activations from layers of neural network in response to evaluating data; and using the extracted training activations to generate histograms.
Cintas further teaches,
extracting training activations from layers of neural network in response to evaluating data [Para 0062, In analyzing data for anomalousness, discriminator 116 first obtains samples 122 for evaluation. Discriminator 116 then extracts activations 124 from the samples to compute empirical p-value values 134; Para 0111, Table 2 shows precision and recall metrics for an example of group subset scanning over three convolutional layers of a discriminator using a GAN improvement pipeline 430 in accordance with the present disclosure].
Cintas is analogous to the claimed invention as they both relate to extracting activations and detecting anomalous data. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Cintas and provide extracting activations and retaining anomalous data in order to [Cintas, para 0058] to better evaluate and classify data.
Rahmes-Cintas-Saxena teach the above limitations of claim 6 including the extracted training activations (Claim 6: Cintas, para 0062).
Rahmes-Cintas-Saxena do not teach using activations to generate histograms.
Dolan teaches,
using activations to generate histograms [Para 0052, such activations can be compared by discretizing a region of an input space into corresponding grids and building a histogram of activations in the associated grids for input data and comparison data].
Dolan is analogous to the claimed invention as they both relate to building histograms for neural networks. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Dolan and provide generating histograms from activations in order to create an efficient visualization for analyzing neuronal activity.
Claim(s) 7 is rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas, Saxena, and Dolan, and in further view of Choi.
Regarding claim 7, Rahmes-Cintas-Saxena-Dolan teach the limitations of claim 6 including using activations to generate the histograms (Claim 6: Dolan, Para 0052), the compressed histograms (Claim 1: Saxena, Abstract), and the extracted training activations (Claim 6: Cintas, para 0062).
Choi teaches,
for each node of neural network that corresponds to a subset of activations [Para 0004, An artificial neural network (ANN) has a structure in which nodes representing artificial neurons are connected through synapses. Nodes can process signals received through synapses and transmit the processed signals to other nodes. Signals from each node are transmitted to other nodes through weights associated with the node and weights associated with the synapse. When a signal processed at one node is transmitted to the next node, its influence varies depending on the weight; Para 0005, Here, the weight associated with the node is referred to as the bias, and the node's output is referred to as the activation], using the subset of activations to create one of the histograms [Para 0024, Activations can be produced through the activation function of the nodes included in the neural network; Para 0025, the calculated activations are classified, or a histogram is generated from the activations].
Choi is analogous to the claimed invention as they both relate to utilizing histograms correlated with neural network nodes. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Choi and provide using nodal activations to create histograms in order to create an efficient visualization for analyzing neuronal activity of a neural network.
Claim(s) 8 is rejected under 35 U.S.C. 103 as being unpatentable over Rahmes in view of Cintas, Saxena, and Dolan, and in further view of Zhang et al. (US 20240223762 A1), hereinafter Zhang.
Regarding claim 8, Rahmes-Cintas-Saxena-Dolan teach the limitations of claim 6 including using activations to generate the histograms (Claim 6: Dolan, Para 0052), the compressed histograms (Claim 1: Saxena, Abstract), and the extracted training activations (Claim 6: Cintas, para 0062).
Zhang teaches
sorting a respective training activation into a respective bin of the histograms based on a numerical value of the respective training activation [Para 0182, The channel grouping for dependency modes 1 to 6 may be determined via a learning approach. According to an embodiment, the channel grouping may be learned via an attention mechanism. The encoder is first pre-trained using a training dataset. The importance of each channel of the quantized latent representation is determined in an unsupervised manner, for example, by measuring the excitation (e.g. using absolute activation values) of neurons using a set of evaluation data samples. The channel groups are then derived from the histogram of n bins from the channel importance map].
Zhang is analogous to the claimed invention as they both relate to utilizing histograms correlated with neural network nodes. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Rahmes’ teachings to incorporate the teachings of Zhang and provide sorting activations into bins of a histogram in order to create an efficient visualization for analyzing neuronal activity.
Allowable Subject Matter
Claims 9 and 12 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.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SYED RAYHAN AHMED whose telephone number is (571)270-0286. The examiner can normally be reached Mon-Fri 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, David Yi can be reached at (571) 270-7519. 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.
/SYED RAYHAN AHMED/Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126
Prosecution Insights