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 .
Detailed Action
The following action is in response to the communication(s) received on 12/04/2023.
As of the claims filed 12/04/2023:
Claims 1-20 are pending.
Claims 1, 8, and 15 are independent claims.
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 recites a method, thus a process, one of the four statutory categories of patentable subject matter (Step 1). However, Claim 1 further recites:
processing... the first data... to generate synthetic data, which is an evaluation or judgement that can be performed in the human mind;
generate anomaly scores, which is an evaluation or judgement that can be performed in the human mind;
combining... the anomaly scores with the first data, the second data, and the synthetic data to generate final data, which is an evaluation or judgement that can be performed in the human mind
performing… one or more actions based on the trained machine learning model, which is an evaluation or judgement that can be performed in the human mind.
Thus, the claim recites an abstract idea under Step 2A Prong 1. Under Step 2A Prong 2, the claim recites:
receiving, by a device, first data associated with a first class and second data associated with a second class that is different than the first class, which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
training, by the device, a variational autoencoder (VAE) model using the second data..., as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application;
training, by the device, a machine learning model with the final data, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because the activity of data gathering (MPEP 2106.05(g)) cannot provide significantly more, as storing and retrieving information in memory is well understood, routine, and conventional (MPEP 2106.05(d)(II)(iv)); implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 2, dependent on 1, further recites
processing the new data... to generate a prediction of whether the new data is associated with the first class or the second class; and performing one or more additional actions based on the prediction, which is an evaluation or judgement that can be performed in the human mind.
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites:
receiving new data, which is merely an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application;
with the trained machine learning model, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because the activity of data gathering (MPEP 2106.05(g)) cannot provide significantly more, as storing and retrieving information in memory is well understood, routine, and conventional (MPEP 2106.05(d)(II)(iv)); implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 3, dependent on 2, further recite. no additional abstract ideas. However: Under Step 2A Prong 2, the claim recites:
retraining the machine learning model based on the prediction, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 4, dependent on 2, further recites
generating a whitelist or a blacklist based on the prediction; determining fraudulent activity based on the prediction; or utilizing the prediction to make a decision associated with the new data, which is an evaluation or judgement that can be performed in the human mind.
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites: no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself; thus, the claim remains ineligible.
Claim 5, dependent on 1, further recites
the synthetic data is generated based on learning a distribution of patterns associated with the first data, which is an evaluation or judgement that can be performed in the human mind.
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites: no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself; thus, the claim remains ineligible.
Claim 6, dependent on 1, further recites
a sum of a quantity of the first data and a quantity of the synthetic data is substantially equivalent to a quantity of the second data, which is merely a detail of an abstract idea (generate synthetic data).
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites: no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself; thus, the claim remains ineligible.
Claim 7, dependent on 1, further recite. no additional abstract ideas. However: Under Step 2A Prong 2, the claim recites:
the first data is data associated with fraudulent activities and the second data is data associated with non-fraudulent activities, which is merely a detail of an insignificant extra-solution activity of data gathering, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because the activity of data gathering (MPEP 2106.05(g)) cannot provide significantly more, as storing and retrieving information in memory is well understood, routine, and conventional (MPEP 2106.05(d)(II)(iv)). The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 8 recites a device, thus a machine, one of the four statutory categories of patentable subject matter. However, Claim 8 recites comprising: one or more processors configured to perform precisely the abstract ideas and additional elements of Claim 1 and 2, combined. Additionally, Claim 8 recites: generate a risk score, which is an evaluation or judgement that can be performed in the human mind. Therefore, Step 2A Prong 1 analysis remains the same. As for Step 2A Prong 2 and Step 2B: performance on a computer cannot integrate an abstract idea into a practical application (Step 2A Prong 2) nor provide significantly more than the abstract idea itself (Step 2B) (MPEP 2106.05(f)), and thus Claim 8 is rejected as subject-matter ineligible for reasons set forth in the rejections of Claim 1 and 2, combined.
Claim 9, dependent on 8, further recite. no additional abstract ideas. However: Under Step 2A Prong 2, the claim recites:
the VAE model is an unsupervised neural network model, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 10, dependent on 8, further recites
… configured to identify a data distribution of the second data, which is an evaluation or judgement that can be performed in the human mind.
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites:
the trained VAE model includes an encoder-decoder architecture configured to reconstruct the first data, the second data, and the synthetic data, and a Kullback-Leibler divergence loss function, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 11, dependent on 8, further recites
a range of anomaly scores associated with the first data is greater than a range of anomaly scores associated with the second data, which is merely a detail of an abstract idea (generate anomaly scores).
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites: no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself. Thus, the claim remains ineligible.
Claim 12, dependent on 8, further recite no additional abstract ideas. However: Under Step 2A Prong 2, the claim recites:
implement the trained machine learning model in a system associated with the first data and the second data, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 13, dependent on 8, further recite no additional abstract ideas. However: Under Step 2A Prong 2, the claim recites:
machine learning model is one of an XGBoost model, a multilayer perceptron model, or a support vector machine model, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 14, dependent on 8, further recites
… addresses a class imbalance issue associated with the first data and the second data, which is an evaluation or judgement that can be performed in the human mind;
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites:
the trained machine learning model…, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claim 15 recites A non-transitory computer-readable medium, thus an article of manufacture, one of the four statutory categories of patentable subject matter. However, Claim 15 recites storing a set of instructions, the set of instructions comprising: one or more instructions that, when executed by one or more processors of a device, cause the device to perform precisely the abstract ideas and additional elements of Claim 1 and 2, combined. Additionally, Claim 15 recites: process… an anomaly score prediction… to generate a prediction, which is an evaluation or judgement that can be performed in the human mind. Therefore, Step 2A Prong 1 analysis remains the same. As for Step 2A Prong 2 and Step 2B: performance on a computer cannot integrate an abstract idea into a practical application (Step 2A Prong 2) nor provide significantly more than the abstract idea itself (Step 2B) (MPEP 2106.05(f)), and thus Claim 15 is rejected as subject-matter ineligible for reasons set forth in the rejections of Claim 1 and 2, combined.
Claim 16, dependent on 15, further recites
utilize the prediction to make a decision associated with the new data, which is an evaluation or judgement that can be performed in the human mind.
Thus, the claim recites an abstract idea under Step 2A Prong 1.Under Step 2A Prong 2, the claim recites:
retrain the machine learning model based on the prediction, as the performance of an abstract idea on a computer is not more than instructions to 'apply it' on a computer, which by MPEP 2106.05(f) cannot integrate an abstract idea into a practical application;
provide the prediction for display, which is merely an insignificant extra-solution activity of displaying results, which by MPEP 2106.05(g) cannot integrate an abstract idea into a practical application.
Thus, the claim is directed towards and abstract idea.
Further, the additional element(s), alone or in combination, do not provide significantly more than the abstract idea itself, because the activity of displaying results (MPEP 2106.05(g)) cannot provide significantly more, as displaying results is well understood, routine, and conventional (Tuncel [p.17 right ¶2] To prevent the audience from getting bored while reading a scientific article, some of the data should be expressed in a visual format in graphics… Peer-reviewers frequently look at tables…) (MPEP 2106.05(d)); implementation on a computer (MPEP 2106.05(f)) cannot provide significantly more. The combination of these additional elements does not provide an inventive concept; thus, the claim remains ineligible.
Claims 17-20 recite precisely the abstract ideas and additional elements of Claims 5-7 and 10, respectively; thus, they are rejected as subject-matter ineligible for reasons set forth in the rejections of Claims 5-7 and 10, 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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 5, and 7 are rejected under 35 U.S.C. 103 as being unpatentable over Pollastro et al., “Semi-Supervised Detection of Structural Damage Using Variational Autoencoder and a One-Class Support Vector Machine” (hereinafter Pollastro), in view of Chen et al., “Variational Autoencoders and Wasserstein Generative Adversarial Networks for Improving the Anti-Money Laundering Process” (hereinafter Chen).
Regarding Claim 1, Pollastro teaches:
A method, comprising: receiving, by a device, first data associated with a first class and second data associated with a second class that is different than the first class; (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection) and a One-Class Support Vector Machines (OC-SVM) to classify data as undamaged or not, thus avoiding any user-defined decision rule.) (Note: damaged data corresponds to the first data; undamaged data corresponds to the second data)
Pollastro’s does not teach, but Chen further teaches:
processing, by the device, the first data, with a generative adversarial network model, to generate synthetic data; (Chen [p.23 left ¶3] To obtain a more balanced dataset, WGAN was used to generate more fraud transactions, which were mixed with the base dataset to produce the mixed dataset. This was then used to train the autoencoders.) (Note: fraud transactions correspond to the first class; the generated fraud transactions correspond to the generated synthetic data)
Chen and Pollastro are analogous to the present invention because both are from the same field of endeavor of anomaly detection methods using VAEs. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the synthetic data generation method from Chen into Pollastro’s anomaly detection method. The motivation would be to “To obtain a more balanced dataset” (Chen [p.23 left ¶3]).
Pollastro, via Pollastro/Chen, further teaches:
training, by the device, a variational autoencoder (VAE) model using the second data, to generate a trained VAE model; (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection…) and a One-Class Support Vector Machines (OC-SVM)… to classify data as undamaged or not, thus avoiding any user-defined decision rule.) (Note: damaged data corresponds to the first data; undamaged data corresponds to the second data; the VAE trained on undamaged raw data corresponds to the generated trained VAE model)
utilizing, by the device, the first data, the second data, and the synthetic data with the trained VAE model to generate anomaly scores; (Pollastro [fig.1]
PNG
media_image1.png
368
593
media_image1.png
Greyscale
[p.8 bottom right] Mean Squared Error (MSE), which measures the reconstruction error between the input acceleration signals and their reconstruction…
Original-to-Reconstructed-Signal Ratio (ORSR)… represents the ratio in decibels between the magnitudes of the original signal and its reconstruction…
[fig.9]
PNG
media_image2.png
843
466
media_image2.png
Greyscale
) (Note: MSE and ORSR measure the original signal to the reconstructed signal and thus correspond to anomaly scores; the test data which contains both the damaged (and the generated synthetic anomaly data) and undamaged data correspond to the first, second, and synthetic data)
combining, by the device, the anomaly scores with the first data, the second data, and the synthetic data to generate final data; training, by the device, a machine learning model with the final data to generate a trained machine learning model; (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection [23], [24], [25]) and a One-Class Support Vector Machines (OC-SVM) to classify data as undamaged or not, thus avoiding any user-defined decision rule.
[fig.9]
PNG
media_image2.png
843
466
media_image2.png
Greyscale
) (Note: the OC-SVM corresponds to the trained machine learning model; MSE and ORSR measure the original signal to the reconstructed signal and thus correspond to anomaly scores; the ORSR-MSE values inputted to the OC-SVM correspond to the final data generated by combining the anomaly scores with the first, second and synthetic data (as input for the OC-SVM))
and performing, by the device, one or more actions based on the trained machine learning model. (Pollastro [p.11 left ¶2] On the contrary, our method identifies all the different structural conditions.) (Note: identifying the different structural conditions corresponds to performing one or more actions)
Regarding Claim 5, Pollastro/Chen respectively teaches and incorporates the claimed limitations and rejections of Claim 1. Chen, via Pollastro/Chen further teaches:
The method of claim 1, wherein the synthetic data is generated based on learning a distribution of patterns associated with the first data. (Chen [p.23 left ¶3] To obtain a more balanced dataset, WGAN was used to generate more fraud transactions, which were mixed with the base dataset to produce the mixed dataset. This was then used to train the autoencoders.
[p.8 bottom left] …by replacing the GAN loss function based on the Wasserstein 1 or Earth-Mover distance (EMD). In our case, this is where the ‘‘critic’’ discriminator is calculating the Wasser stein distance between the real and fake samples. As the loss function decreases in the training process, the Wasserstein distance becomes smaller. Hence, the generator generates samples closer to the real ones. The intuition behind EMD is that it measures how much mass γ(x,y) should be transported by d = x − y for the probability distribution p_data to match the probability distribution p_g…. (x,y)is also known as a transport plan to reflect the strategy for transporting masses to match the two probability distributions…) (Note: the generated fraud transactions using the EMD which involves comparing probability distributions correspond to the generated synthetic data)
Regarding Claim 7, Pollastro/Chen respectively teaches and incorporates the claimed limitations and rejections of Claim 1. Chen, via Pollastro/Chen, further teaches:
The method of claim 1, wherein the first data is data associated with fraudulent activities and the second data is data associated with non-fraudulent activities. (Chen [p.9 right ¶2] The Day group has the most transactions, with a total of 2706 transactions, where 45 of them are fraud and the rest are normal with only 3 Null values under the P2 attribute and no duplicates. While the Week group has 1490 transactions, 44 of them are fraudulent ones, with three Null values under the P2 attribute and no duplicates. Finally, the Month group has 693 transactions, with 44 of them being fraud, two Null values under the P2 attribute, and no duplicates.) (Note: the data containing fraudulent transactions correspond to the first data associated with the first class; the rest of transactions which are normal (and non-fraudulent) correspond to the second data associated with the second class)
Chen and Pollastro are analogous to the present invention because both are from the same field of endeavor of anomaly detection methods. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the fraudulent data from Chen into Pollastro’s VAE-based anomaly detection method. The motivation would be to “Studies show that supervised techniques often have the limitation of not adapting well to new irregular fraud patterns when the dataset is highly imbalanced. Instead, unsupervised learning can have a better capability to find anomalous and irregular patterns in new transaction.” (Chen, abstract).
Claims 2, 3, 4, and 8-17, 19, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Pollastro/Chen further in view of Hemati et al., “Continual Learning for Unsupervised Anomaly Detection in Continuous Auditing of Financial Accounting Data” (hereinafter Hemati).
Regarding Claim 2, Pollastro/Chen respectively teaches and incorporates the claimed limitations and rejections of Claim 1. Pollastro/Chen does not teach, but Hemati further teaches:
The method of claim 1, further comprising: receiving new data; (Hemati [p.3, fig.2, bottom left]
PNG
media_image3.png
408
1002
media_image3.png
Greyscale
[abstract] Ignoring situations where audit relevant distribution changes are not evident in the training data or become incrementally available over time. In contrast, in continuous auditing, deep-learning models are continually trained on a stream of recorded journal entries, e.g., of the last hour. Resulting in situations where previous knowledge interferes with new information and will be entirely overwritten. This work proposes a continual anomaly detection framework to overcome both challenges and designed to learn from a stream of journal entry data experiences) (Note: new information from the stream of entries corresponds to new data)
Hemati and Pollastro/Chen are analogous to the present invention because both are from the same field of endeavor of anomaly detection methods. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the method of continual learning via new data streams from Hemati into Pollastro/Chen’s method of detecting anomalies. The motivation would be to “to learn from a stream of journal entry data experiences” (Hemati, abstract).
Pollastro, via Pollastro/Chen/Hemati, further teaches:
processing the new data, with the trained machine learning model, to generate a prediction of whether the new data is associated with the first class or the second class; (Pollastro [p.7 top left] Accelerations were recorded in the absence (Case 1) and in the presence of structural damage. Eight cases of damage were simulated. Table 1 and Figure 6 summarize the various damage scenarios in which the intensity gradually increases from Case 2 to Case 9. The simulated structural damage consists in the removal of diagonal stiffening elements in Cases 2 to 7, while the loosening of the connecting bolts is added in Cases 8 and 9.
[p.9 left ¶2] Finally, the overall structure score for each case Si was computed by averaging the PoD values of each sensor:
PNG
media_image4.png
57
366
media_image4.png
Greyscale
…a low value of PoDij indicates a low probability that the signal i recorded by the j -th sensor belongs to an undamaged state. On the other hand, a high value indicates a high probability of belonging to damaged state. Same observations are valid for the PoDavg,i value.) (Note: PoD score for each case corresponds to the prediction of whether the new data is associated with the first or second class)
and performing one or more additional actions based on the prediction. (Pollastro [p.8 bottom left] MLP Neural Networks were adopted as architecture to model both the probabilistic encoder and probabilistic decoder. Search spaces for hyperparameters were established during a preliminary manual analysis with the aim of minimizing the computational time needed for the overall model selection stage.) (Note: performing the model selection based on the prediction performance of the resulting model corresponds to performing additional actions based on the prediction)
Regarding Claim 3, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 2. Pollastro, via Pollastro/Chen/Hemati further teaches:
The method of claim 2, wherein performing the one or more additional actions comprises: retraining the machine learning model based on the prediction. (Pollastro [p.8 bottom left] MLP Neural Networks were adopted as architecture to model both the probabilistic encoder and probabilistic decoder. Search spaces for hyperparameters were established during a preliminary manual analysis with the aim of minimizing the computational time needed for the overall model selection stage.) (Note: performing the model selection based on the prediction performance of the resulting model involves retraining the model that includes the OC-SVM (the machine learning model), thus corresponding to performing additional actions based on the prediction; each model in the search space corresponds to the retrained machine learning model)
Regarding Claim 4, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 2. Chen, via Pollastro/Chen/Hemati, further teaches:
The method of claim 2, wherein performing the one or more additional actions comprises one or more of: generating a whitelist or a blacklist based on the prediction; determining fraudulent activity based on the prediction; or utilizing the prediction to make a decision associated with the new data. (Chen [abstract] Studies show that supervised techniques often have the limitation of not adapting well to new irregular fraud patterns when the dataset is highly imbalanced. Instead, unsupervised learning can have a better capability to find anomalous and irregular patterns in new transaction.)
Chen and Pollastro/Chen/Hemati are analogous to the present invention because both are from the same field of endeavor of anomaly detection methods. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the fraudulent data for detecting fraudulent activity from Chen into Pollastro/Chen/Hemati’s VAE-based anomaly detection method. The motivation would be that “Studies show that supervised techniques often have the limitation of not adapting well to new irregular fraud patterns when the dataset is highly imbalanced. Instead, unsupervised learning can have a better capability to find anomalous and irregular patterns in new transaction.” (Chen, abstract).
Independent Claim 8 recites A device, comprising: one or more processors configured to (Pollastro [p.2 middle right] Our proposal consists in using a VAE, trained on undamaged raw data…) (Note: training a VAE requires a processor) to perform precisely the methods of Claim 1 and 2, combined. Thus, Claim 8 is rejected for reasons set forth in Claim 1 and 2, combined.
Additionally, Pollastro, via Pollastro/Chen/Hemati, further teaches:
generate a risk score… process the new data and the risk score (Pollastro [p.7 top left] Accelerations were recorded in the absence (Case 1) and in the presence of structural damage. Eight cases of damage were simulated. Table 1 and Figure 6 summarize the various damage scenarios in which the intensity gradually increases from Case 2 to Case 9. The simulated structural damage consists in the removal of diagonal stiffening elements in Cases 2 to 7, while the loosening of the connecting bolts is added in Cases 8 and 9.
[p.9 left ¶2] Finally, the overall structure score for each case Si was computed by averaging the PoD values of each sensor:
PNG
media_image4.png
57
366
media_image4.png
Greyscale
…a low value of PoDij indicates a low probability that the signal i recorded by the j -th sensor belongs to an undamaged state. On the other hand, a high value indicates a high probability of belonging to damaged state. Same observations are valid for the PoDavg,i value.) (Note: each PoDij corresponds to each risk score; PoD score for each case corresponds to the prediction of whether the new data (and risk score) is associated with the first or second class)
Regarding Claim 9, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 8. Pollastro, via Pollastro/Chen/Hemati, further teaches:
The device of claim 8, wherein the VAE model is an unsupervised neural network model. (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection [23], [24], [25]) and a One-Class Support Vector Machines (OC-SVM) [26] to classify data as undamaged or not, thus avoiding any user-defined decision rule.) (Note: avoiding any user-defined decision rule when training the VAE corresponds to the VAE being an unsupervised neural network model)
Regarding Claim 10, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 8. Pollastro, via Pollastro/Chen/Hemati, further teaches:
The device of claim 8, wherein the trained VAE model includes an encoder-decoder architecture configured to reconstruct the first data, the second data, and the synthetic data, and a Kullback-Leibler divergence loss function configured to identify a data distribution of the second data. (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection…) and a One-Class Support Vector Machines (OC-SVM)… to classify data as undamaged or not, thus avoiding any user-defined decision rule.)
[p.5 left ¶2] To admit inference, VAE training simultaneously optimizes both the parameters θ and ϕ while learning the marginal likelihood of the data in the following generative process:
PNG
media_image5.png
40
200
media_image5.png
Greyscale
where logpθ(x|z) can be defined as:
PNG
media_image6.png
32
393
media_image6.png
Greyscale
where DKL(⋅) stands for the Kullback–Leibler (KL) divergence and p(z) is the prior distribution over the latent variables z. Notice that KL divergence quantifies the difference between two probability distributions q and p.) (Note: VAE used to classify data as damaged or not (during inference) corresponds to reconstructing the first, second, and synthetic data; the prior distribution over the latent variables z corresponds to the identified data distribution of the second data)
Regarding Claim 11, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 8. Pollastro, via Pollastro/Chen/Hemati, further teaches:
The device of claim 8, wherein a range of anomaly scores associated with the first data is greater than a range of anomaly scores associated with the second data. (Pollastro [p.1 bottom right] In a nutshell, in anomaly detection tasks AEs are trained to minimize reconstruction error only on normal data instances, thus involving high reconstruction error on anomalous data. Then, the reconstruction error is considered as an anomaly score to classify the input data as anomalous or not, using a user-defined decision rule.
[p.10 bottom left] Assuming that generating distributions of damaged data are different from that of undamaged data, our proposal aims to learn the latent distribution of undamaged data in order to induce the probabilistic encoder to encode damaged data with different generating distributions. As a consequence, the probabilistic decoder will hardly decode data coming from distributions diverse from those learned during the training stage, thus resulting in high reconstruction error. In order to verify how much generating distributions of damaged data diverge from that of undamaged data, KL divergences were computed for each sensor and reported in Table 4.)
Regarding Claim 12, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 8. Pollastro, via Pollastro/Chen/Hemati further teaches:
The device of claim 8, wherein the one or more processors, to perform the one or more actions based on the trained machine learning model, are configured to: implement the trained machine learning model in a system associated with the first data and the second data. (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection) and a One-Class Support Vector Machines (OC-SVM) to classify data as undamaged or not, thus avoiding any user-defined decision rule.)
Regarding Claim 13, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 8. Pollastro, via Pollastro/Chen/Hemati further teaches:
The device of claim 8, wherein the machine learning model is one of an XGBoost model, a multilayer perceptron model, or a support vector machine model. (Pollastro [p.2 right ¶2] In this work, we propose a semi-supervised data-driven DL-based framework to detect damages in an SHM system. Our proposal consists in using a VAE, trained on undamaged raw data, to represent input data through damage-sensitive features (typically involved in structural damage detection [23], [24], [25]) and a One-Class Support Vector Machines (OC-SVM) to classify data as undamaged or not, thus avoiding any user-defined decision rule.
[fig.9]
PNG
media_image2.png
843
466
media_image2.png
Greyscale
) (Note: the OC-SVM corresponds to the trained machine learning model)
Regarding Claim 14, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 8. Chen, via Pollastro/Chen/Hemati, further teaches:
The device of claim 8, wherein the trained machine learning model addresses a class imbalance issue associated with the first data and the second data. (Chen [p.23 left ¶3] To obtain a more balanced dataset, WGAN was used to generate more fraud transactions, which were mixed with the base dataset to produce the mixed dataset. This was then used to train the autoencoders.) (Note: fraud transactions correspond to the first data; the generated fraud transactions correspond to the generated synthetic data; using the mixed (more balanced) dataset instead of the regular base dataset corresponds to addressing the class imbalance issue.)
Independent Claim 15 recites A device, comprising: one or more processors configured to (Pollastro [p.2 middle right] Our proposal consists in using a VAE, trained on undamaged raw data…) (Note: training a VAE requires a processor) to perform precisely the methods of Claim 1 and 2, combined. Thus, Claim 15 is rejected for reasons set forth in Claim 1 and 2, combined.
Additionally, Pollastro, via Pollastro/Chen/Hemati, further teaches:
process the new data and an anomaly score prediction for the new data (Pollastro [p.7 top left] Accelerations were recorded in the absence (Case 1) and in the presence of structural damage. Eight cases of damage were simulated. Table 1 and Figure 6 summarize the various damage scenarios in which the intensity gradually increases from Case 2 to Case 9. The simulated structural damage consists in the removal of diagonal stiffening elements in Cases 2 to 7, while the loosening of the connecting bolts is added in Cases 8 and 9.
[p.9 left ¶2] Finally, the overall structure score for each case Si was computed by averaging the PoD values of each sensor:
PNG
media_image4.png
57
366
media_image4.png
Greyscale
…a low value of PoDij indicates a low probability that the signal i recorded by the j -th sensor belongs to an undamaged state. On the other hand, a high value indicates a high probability of belonging to damaged state. Same observations are valid for the PoDavg,i value.) (Note: each PoDij corresponds to each anomaly score for the new data; PoD score for each case corresponds to the prediction of whether the new data (and anomaly score) is associated with the first or second class)
Regarding Claim 16, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 15. Chen, via Pollastro/Chen/Hemati, further teaches:
The non-transitory computer-readable medium of claim 15, wherein the one or more instructions, that cause the device to perform the one or more actions, cause the device to one or more of: retrain the machine learning model based on the prediction; provide the prediction for display; or utilize the prediction to make a decision associated with the new data. (Chen [abstract] Studies show that supervised techniques often have the limitation of not adapting well to new irregular fraud patterns when the dataset is highly imbalanced. Instead, unsupervised learning can have a better capability to find anomalous and irregular patterns in new transaction.) (Note: finding anomalous patterns in new transaction correspond to utilizing the prediction to make a decision associated with the new data)
Chen into Pollastro are analogous to the present invention because both are from the same field of endeavor of anomaly detection methods. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the fraudulent data for detecting fraudulent activity from Chen into Pollastro’s VAE-based anomaly detection method. The motivation would be to “Studies show that supervised techniques often have the limitation of not adapting well to new irregular fraud patterns when the dataset is highly imbalanced. Instead, unsupervised learning can have a better capability to find anomalous and irregular patterns in new transaction.” (Chen, abstract).
Claim(s) 17, 19, and 20, dependent on Claim 15, also recite the device configured to perform precisely the methods of Claim(s) 5, 7, and 10, respectively, and thus are rejected for reasons set forth in these claim(s).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Pollastro/Chen further in view of Englemann et al., “Conditional Wasserstein GAN-based Oversampling of Tabular Data for Imbalanced Learning” (hereinafter Englemann).
Regarding Claim 6, Pollastro/Chen respectively teaches and incorporates the claimed limitations and rejections of Claim 1. Pollastro/Chen does not teach, but Englemann further teaches:
The method of claim 1, wherein a sum of a quantity of the first data and a quantity of the synthetic data is substantially equivalent to a quantity of the second data. (Englemann [p.21 ¶2] Having more training data in general can help to reduce the extent of overfitting for Random Oversampling and improve the quality of the generated data for cWGAN-based oversampling. Secondly, we apply the oversampling technique to the training set to generate new synthetic minority examples, which are added to the training set to balance it. In the present work, we only consider resampling to parity.) (Note: resampling to parity corresponds to having substantially equivalent sum of first and synthetic data to second data)
Englemann and Pollastro/Chen are analogous to the present invention because both are from the same field of endeavor of oversampling tabular data with WGAN methods. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the resampling to parity method from Englemann to Pollastro/Chen’s fraudulent activity detection method. The motivation would be to “Having more training data in general can help to reduce the extent of overfitting for Random Oversampling and improve the quality of the generated data for cWGAN-based oversampling. Secondly, we apply the oversampling technique to the training set to generate new synthetic minority examples, which are added to the training set to balance it.” (Englemann, p.21 ¶2).
Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over Pollastro/Chen/Hemati further in view of Englemann.
Regarding Claim 18, Pollastro/Chen/Hemati respectively teaches and incorporates the claimed limitations and rejections of Claim 15. Pollastro/Chen/Hemati does not teach, but Englemann further teaches:
The non-transitory computer-readable medium of claim 15, wherein a sum of a quantity of the first data and a quantity of the synthetic data is substantially equivalent to a quantity of the second data. (Englemann [p.21 ¶2] Having more training data in general can help to reduce the extent of overfitting for Random Oversampling and improve the quality of the generated data for cWGAN-based oversampling. Secondly, we apply the oversampling technique to the training set to generate new synthetic minority examples, which are added to the training set to balance it. In the present work, we only consider resampling to parity.) (Note: resampling to parity corresponds to having substantially equivalent sum of first and synthetic data to second data)
Englemann and Pollastro/Chen/Hemati are analogous to the present invention because both are from the same field of endeavor of oversampling tabular data with WGAN methods. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the resampling to parity method from Englemann to Pollastro/Chen/Hemati’s fraudulent activity detection method. The motivation would be to “Having more training data in general can help to reduce the extent of overfitting for Random Oversampling and improve the quality of the generated data for cWGAN-based oversampling. Secondly, we apply the oversampling technique to the training set to generate new synthetic minority examples, which are added to the training set to balance it.” (Englemann, p.21 ¶2).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOSEP HAN whose telephone number is (703)756-1346. The examiner can normally be reached Mon-Fri 9am-5pm.
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, Kakali Chaki can be reached on (571) 272-3719. 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.
/J.H./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights