Notice of Pre-AIA or AIA Status
This Non-Final communication is in response to application No. 17/554,630 filed on 12/17/2021 which claims priority to 63/127,378 filed on 12/18/2020. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The amendment filed 12/23/2025 has been entered which makes amendments to claims 1, 4, 8, 10, 17, cancels claim 13 and adds claim 21. Claims 1-12 and 14-21 remain pending in the application.
Response to Arguments
Applicant's arguments with respect to 35 U.S.C § 101 filed 12/23/2025 have been fully considered but they are not persuasive.
Applicant argues that the claimed invention does not recite a mental process. The examiner respectfully disagrees. For reasons stated in the previous office action and down below the invention does recite a mental process. Specifically, the limitations “set a threshold to an initial value x…” and “determine a final value of the threshold…”. Looking at the reconstruction error of a set of data and creating a threshold based on that could be performed in the human mind. Comparing two sets of data and making a judgement for the final threshold could also be performed in the human mind. Using a training controller, a generic computer component, to execute these operation does not make these any less than a mental process.
Applicant argues that the claimed invention is an improvement to technology. The applicant cites paragraph 21 and 54 of the specification describing the improvement. The examiner respectfully disagrees. According to MPEP guidance, the claim must be evaluated to ensure that the claim itself reflects the alleged improvement. Paragraph 54 of the specification cites improving the ability to detect when a machine needs maintenance. The claims focus on the reconstruction error of data to create a threshold. It is not clear how this threshold creation relates to the improvement of detecting machine maintenance in the claims. The additional elements are directed to training the autoencoder and testing it.
Therefore, the 101 rejection is maintained.
Applicant’s arguments with respect to 35 U.S.C § 103 filed 12/23/2025 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
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 1-12, 14-16 and 21 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
101 Subject Matter Eligibility analysis
Step 1: Claims 1-12, 14-16 and 21 are within the four statutory categories (a process, machine, manufacture or composition of matter.)
-With respect to claim 1:
Step 2A Prong 1: The claim recites an abstract idea enumerated in the 2019 PEG.
set a threshold to an initial value x at a percentile threshold of a first empirical distribution of reconstruction errors of the first portion of the first set of sensor data after decoding by a decoder layer of the autoencoder; (This is an abstract idea of a “Mental Process.” The “set” under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind aside from the recitation of generic computer components.)
determine a final value of the threshold based on a comparison of a second empirical distribution of reconstruction errors of a second portion of the first set of sensor data to the first empirical distribution of reconstruction errors of the first portion of the first set of sensor data, wherein a value is determined based on a percentage of the reconstruction errors of the second portion of the first set of sensor data that are greater than the value of an initial threshold from a set value, and if the value is greater than a percentile threshold of the second empirical distribution of reconstruction errors of the second portion of the first set of sensor data, then the final value of the threshold is set equal to the value of the initial value x, and if the value is less than or equal to the percentile threshold, then the final value of the threshold is equal to a value at an inverse percentile of the percentile threshold; (This is an abstract idea of a "Mental Process." The "determining" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind. The determination could be made manually by an individual.)
determine that data drift is not present when the second set of sensor data is less than the final value of the threshold (This is an abstract idea of a "Mental Process." The "determining" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind. The determination could be made manually by an individual.)
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
Additional elements:
an autoencoder configured to receive a first set of sensor data and a second set of sensor data to detect data drift for an aircraft system; (The “receive” step is mere data gathering and thus this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
train the autoencoder based on a first portion of the first set of sensor data; (This amounts to no more than mere instructions to “apply” the exception using a generic computer component.)
a testing controller configured to test the autoencoder with the second set of sensor data; (This amounts to no more than mere instructions to “apply” the exception using a generic computer component.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As explained above the additional elements “train the autoencoder” and “a testing controller” are recited in a generic level and they represent generic computer components to apply the abstract idea. Mere instructions to apply an exception cannot provide an inventive concept.
The additional element of “receive …” adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
When considered in combination, these additional elements represent mere instructions to apply an exception and insignificant extra-solution activity, which do not provide an inventive concept.
Therefore, claim 1 is ineligible.
With respect to claim 2:
Step 2A Prong 1: claim 2, which incorporates the rejection of claim 1, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the autoencoder includes an input layer, an encoder layer, and the decoder layer; (this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 2 is ineligible.
With respect to claim 3:
Step 2A Prong 1: claim 3, which incorporates the rejection of claim 2, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the autoencoder is a three-layer autoencoder; (this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 3 is ineligible.
With respect to claim 4:
Step 2A Prong 1: claim 4, which incorporates the rejection of claim 1, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the first set of sensor data and the second set of sensor data are captured by one or more sensors that monitor operation of the aircraft system; (this amounts to merely indicating a field of use or technological environment in which to apply the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds field of use to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 4 is ineligible.
With respect to claim 5:
Step 2A Prong 1: claim 5, which incorporates the rejection of claim 1, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the percentile threshold is 90-99.7; (this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 5 is ineligible.
With respect to claim 6:
Step 2A Prong 1: claim 6, which incorporates the rejection of claim 1, recites further limitations:
determine a first reconstruction error at the percentile threshold; (This is an abstract idea of a "Mental Process." The "determining" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind. The determination could be made manually by an individual.)
compare the first reconstruction error of the second set of sensor data with the final value of the threshold; (This is an abstract idea of a "Mental Process." The "compare" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind.)
determine that data drift is not present when the first reconstruction error is less than the final value of the threshold; (This is an abstract idea of a "Mental Process." The "determining" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind. The determination could be made manually by an individual.)
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
test the autoencoder with the second set of sensor data the testing controller; (This amounts to no more than mere instructions to “apply” the exception using a generic computer component.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element is recited in a generic level and they represent generic computer components to apply the abstract idea. Mere instructions to apply an exception cannot provide an inventive concept.
Therefore, claim 6 is ineligible.
With respect to claim 7:
Step 2A Prong 1: claim 7, which incorporates the rejection of claim 6, recites further limitations:
determine that data drift is present when the first reconstruction error is equal to or greater than the final value of the threshold; (This is an abstract idea of a "Mental Process." The "determining" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind. The determination could be made manually by an individual.)
Step 2A Prong 2: The claim does not include any additional elements and thus cannot be integrated into a practical application.
Step 2B: The claim does not include any additional elements.
Therefore, claim 7 is ineligible.
With respect to claim 8:
Step 2A Prong 1: The claim recites similar limitations as corresponding to claims 1 and 6. Therefore, the same analysis that was utilizes under step 2A prong 1 for claims 1 and 6, as described above, is equally applicable to claim 8. The claim recites one addition limitation:
calculating a deviation output for at least one of the one or more sensors; (This is an abstract idea of a "Mental Process." The "calculating" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind.)
Step 2A Prong 2 & Step 2B: The claim recites similar additional elements as corresponding to claims 1, 2 and 6. Therefore, the same analysis that was utilizes under step 2A prong 1 and step 2B for claims 1, 2 and 6, as described above, is equally applicable to claim 8.
Therefore, claim 8 is ineligible.
With respect to claim 9:
The claim recites similar limitations as corresponding to claims 4. Therefore, the same subject matter analysis that was utilized for claim 4, as described above, is equally applicable to claim 9.
Therefore, claim 9 is ineligible.
With respect to claim 10:
The claim recites similar limitations as corresponding to claims 5. Therefore, the same subject matter analysis that was utilized for claim 5, as described above, is equally applicable to claim 10.
Therefore, claim 10 is ineligible.
With respect to claim 11:
Step 2A Prong 1: claim 11, which incorporates the rejection of claim 8, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the percentile threshold is equivalent to a mean of the first empirical distribution of reconstruction errors plus three standard deviations away from the mean; (this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 11 is ineligible.
With respect to claim 12:
The claim recites similar limitations as corresponding to claims 7. Therefore, the same subject matter analysis that was utilized for claim 7, as described above, is equally applicable to claim 12.
Therefore, claim 12 is ineligible.
With respect to claim 14:
Step 2A Prong 1: claim 14, which incorporates the rejection of claim 8, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the input layer has a first quantity of neurons, and the decoder layer has a second quantity of neurons, wherein the first quantity is equal to the second quantity; (this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 14 is ineligible.
With respect to claim 15:
Step 2A Prong 1: claim 15, which incorporates the rejection of claim 14, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the encoder layer has a third quantity of neurons, wherein the third quantity is 60-75% of the first quantity; (this limitation amounts to adding insignificant extra-solution activity to the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The additional element adds insignificant extra-solution activity to the judicial exception and also cannot provide an inventive concept.
Therefore, claim 15 is ineligible.
With respect to claim 16:
Step 2A Prong 1: claim 16, which incorporates the rejection of claim 8, recites further limitations:
the first set of sensor data was randomized prior to receipt by the input layer; (This is an abstract idea of a "Mental Process." The "randomizing" step under its broadest reasonable interpretation, covers concepts that can be practically performed in the human mind. The determination could be made manually by an individual.)
Step 2A Prong 2: The claim does not include any additional elements and thus cannot be integrated into a practical application.
Step 2B: The claim does not include any additional elements.
Therefore, claim 16 is ineligible.
With respect to claim 21:
Step 2A Prong 1: claim 21, which incorporates the rejection of claim 9, does not recite an abstract idea.
Step 2A Prong 2: The judicial exception is not integrated into a practical application.
the aircraft system is an air compressor. (this amounts to merely indicating a field of use or technological environment in which to apply the judicial exception.)
Step 2B: the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception
The additional element merely limits the judicial exception to a particular field of use and cannot provide an inventive concept (MPEP 2106.05(h)).
Therefore, claim 21 is ineligible.
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.
Claims 8, 10-12, 14, 16 are rejected under 35 U.S.C. 103 as being unpatentable over Stocco (NPL: ‘Misbehavior Prediction for Autonomous Driving Systems’) in view of Clark et al. (NPL ‘Adaptive Threshold for Outlier Detection on Data Streams’ (2018)).
Regarding claim 8, Stocco teaches
A method for training and testing an autoencoder to detect data drift, the method comprising (Abstract)
initializing the autoencoder, the autoencoder including an input layer, an encoder layer and a decoder layer; (2 Background subsection Autoencoders “The simplest form of autoencoder (SAE) is a three-layer DNN: the input layer, the hidden layer, and the output layer.”)
training the autoencoder based on a first portion of a first set of sensor data; (5.4.4 SelfOracle’s Configurations “In the paper, authors reduced their training set to 600 images, which were resized to 120x90. During input validation, the three dimensional representation of an online input image is compared to the nominal images by measuring the average of the top-100 minimum distances from the training set.”)
setting a threshold to an initial value x at a percentile threshold of a first empirical distribution of reconstruction errors of the first portion of the first set of sensor data after decoding by the decoder layer; (Example of Threshold Estimation “Let us assume that we are willing to accept a false alarm rate ϵ = 10−2. The threshold θ with a probability mass above the threshold equal to 10−2 can be easily obtained as the inverse of the cumulative gamma distribution F (x): θ = F−1(1 − ϵ). This ensures that the cumulative probability of values ≤ θ is 1 − ϵ, leaving only a probability of ϵ to the tail following θ.”)
determining a final value of the threshold based on a comparison of a second empirical distribution of reconstruction errors of a second portion of the first set of sensor data to the first empirical distribution of reconstruction errors of the first portion of the first set of sensor data, wherein a value is determined based on a percentage of the reconstruction errors of the second portion of the first set of sensor data that are greater than the value of an initial threshold from a set value, and if the value is greater than a percentile threshold of the second empirical distribution of reconstruction errors of the second portion of the first set of sensor data, then the final value of the threshold is set equal to the value of the initial value x, and if the value is less than or equal to the percentile threshold, then the final value of the threshold is equal to a value at an inverse percentile of the percentile threshold; (Example of Threshold Estimation “Let us assume that we are willing to accept a false alarm rate ϵ = 10−2. The threshold θ with a probability mass above the threshold equal to 10−2 can be easily obtained as the inverse of the cumulative gamma distribution F (x): θ = F−1(1 − ϵ). This ensures that the cumulative probability of values ≤ θ is 1 − ϵ, leaving only a probability of ϵ to the tail following θ.” Stocco does not update their threshold but does teach creating a threshold based on reconstruction error of multiple datasets.)
testing the autoencoder with a second set of sensor data detected by one or more sensors (5.6 Results Effectiveness (RQ1) shows their testing.)
determining that data drift is not present when the first reconstruction error of the second set of sensor data is less than the final value of the threshold; and (Example of Threshold Estimation “We use the estimated θ as threshold to distinguish anomalous conditions (reconstruction error ≥ θ) from normal ones (reconstruction error < θ).”).
Stocco does not teach:
updating the initial threshold to get a final threshold
for a third empirical distribution of reconstruction errors of the second set of sensor data after decoding by the decoder layer, determining a first reconstruction error at the percentile threshold;
calculating a deviation output for at least one of the one or more sensors
However, Clark does:
updating the initial threshold to get a final threshold (Page 4 first column 3rd paragraph describes updating the threshold based on the previous iterations of data)
for a third empirical distribution of reconstruction errors of the second set of sensor data after decoding by the decoder layer, determining a first reconstruction error at the percentile threshold; (page 2 left column 2nd paragraph “In more detail, our method uses one class learning algorithms as base algorithms for outlier detection and considers the scores such methods output when presented with new instances. The method applies a statistical test to detect significant changes to the mean of these scores. Significant changes in mean scores suggest that the output of the outlier detection algorithm changed and that the threshold needs an update in order to function in the new environment. No significant change means that the system can continue proceeding as is.”)
calculating a deviation output for at least one of the one or more sensors (Page 4 2nd paragraph “This threshold is set to the mean + 2 standard deviations of the outlier scores output by the model on the validation data.”)
Stocco and Clark are considered analogous art to the claimed invention because they are in the same field of endeavor of data drift/anomaly detection. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to use the system and threshold of Stocco with the threshold updating of Clark. One of ordinary skill in the art would have been motivated to do this to minimize the error of the autoencoders (Clark abstract).
Regarding claim 10, Stocco in view of Clark teach claim 8 as outlined above. Clark further teaches:
wherein setting the initial value of the threshold includes setting the percentile threshold to 90-99.7 (Page 4 first column 2nd paragraph “This threshold is set to the mean + 2 standard deviations of the outlier scores output by the model on the validation data.”)
Regarding claim 11, Stocco in view of Clark teach claim 8 as outlined above. Clark further teaches:
wherein the percentile threshold is equivalent to a mean of the first empirical distribution of reconstruction errors plus three standard deviations away from the mean (Page 4 2nd paragraph “This threshold is set to the mean + 2 standard deviations of the outlier scores output by the model on the validation data.”)
Regarding claim 12, Stocco in view of Clark teach claim 8 as outline above. Stocco further teaches:
determining that data drift is present when the first reconstruction error of the second set of sensor data is equal to or greater than the final value of the threshold (Example of Threshold Estimation “We use the estimated θ as threshold to distinguish anomalous conditions (reconstruction error ≥ θ) from normal ones (reconstruction error < θ).”).
Regarding claim 14, Stocco in view of Clark teach claim 8 as outlined above. Stocco further teaches
wherein the input layer has a first quantity of neurons, and the decoder layer has a second quantity of neurons, further the first quantity is equal to the second. (2 Background subsection Autoencoders “The input and output layers of autoencoders have the same number of nodes.”)
Regarding claim 16, Stocco in view of Clark teaches claim 8 as outlined above. Stocco further teaches:
the first set of sensor data was randomized prior to receipt by the input layer. (5.4 Procedure “Our own implementation relaxed the restrictions above by considering a training set consisting of 3,000 randomly sampled images”)
Claims 1-7, 9 and 17-21 are rejected under 35 U.S.C. 103 as being unpatentable over Stocco in view of Clark and Lore (US 20200234179 A1)
Regarding claim 1, Stocco teaches the following:
A system comprising (Abstract “Our approach SelfOracle is based on a novel concept of self-assessment oracle, which monitors the DNN confidence at runtime, to predict unsupported driving scenarios in advance. SelfOracle uses autoencoder and time-series-based anomaly detection to reconstruct the driving scenarios seen by the car, and determine the confidence boundary of normal/unsupported conditions.).
an autoencoder configured to receive a first set of sensor data and a second set of sensor data to detect data drift for an aircraft system (Abstract “SelfOracle uses autoencoder and time-series-based anomaly detection to reconstruct the driving scenarios seen by the car, and determine the confidence boundary of normal/unsupported conditions.”)
a training controller configured to train the autoencoder based on a first portion of the first set of sensor data (5.4.4 SelfOracle’s Configurations “In the paper, authors reduced their training set to 600 images, which were resized to 120x90. During input validation, the three dimensional representation of an online input image is compared to the nominal images by measuring the average of the top-100 minimum distances from the training set.”)
set a threshold to an initial value x at a percentile threshold of a first empirical distribution of reconstruction errors of the first portion of the first set of sensor data after decoding by a decoder layer of the autoencoder; and (Example of Threshold Estimation “Let us assume that we are willing to accept a false alarm rate ϵ = 10−2. The threshold θ with a probability mass above the threshold equal to 10−2 can be easily obtained as the inverse of the cumulative gamma distribution F (x): θ = F−1(1 − ϵ). This ensures that the cumulative probability of values ≤ θ is 1 − ϵ, leaving only a probability of ϵ to the tail following θ.”)
determine a final value of the threshold based on a comparison of a second empirical distribution of reconstruction errors of a second portion of the first set of sensor data to the first empirical distribution of reconstruction errors of the first portion of the first set of sensor data, wherein a value is determined based on a percentage of the reconstruction errors of the second portion of the first set of sensor data that are greater than the value of an initial threshold from a set value, and if the value is greater than a percentile threshold of the second empirical distribution of reconstruction errors of the second portion of the first set of sensor data, then the final value of the threshold is set equal to the value of the initial value x, and if the value is less than or equal to the percentile threshold, then the final value of the threshold is equal to a value at an inverse percentile of the percentile threshold; (Example of Threshold Estimation “Let us assume that we are willing to accept a false alarm rate ϵ = 10−2. The threshold θ with a probability mass above the threshold equal to 10−2 can be easily obtained as the inverse of the cumulative gamma distribution F (x): θ = F−1(1 − ϵ). This ensures that the cumulative probability of values ≤ θ is 1 − ϵ, leaving only a probability of ϵ to the tail following θ.” Stocco does not update their threshold but does teach creating a threshold based on reconstruction error of multiple datasets.)
a testing controller configured to test the autoencoder by encoding and decoding the second set of sensor data. (5.6 Results Effectiveness (RQ1) shows their testing.)
determine that data drift is not present when the second set of sensor data is less than the final value of the threshold. (Example of Threshold Estimation “We use the estimated θ as threshold to distinguish anomalous conditions (reconstruction error ≥ θ) from normal ones (reconstruction error < θ).”).
Stocco does not teach updating the initial threshold to get a final threshold. However Clark does teach updating thresholds (Page 4 first column 3rd paragraph describes updating the threshold based on the previous iterations of data)
Stocco and Clark are considered analogous art to the claimed invention because they are in the same field of endeavor of data drift/anomaly detection. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to use the system and threshold of Stocco with the threshold updating of Clark. One of ordinary skill in the art would have been motivated to do this to minimize the error of the autoencoders (Clark abstract).
Neither Stocco nor Clark teach detecting drift for an aircraft system. However, Lore teaches using sensors monitoring the operations of an aircraft ([0032] describes an example of using machine learning model where its data is gathered from an aircraft system. Specifically in lines 4-6 “for example, consider a non-limiting example wherein an aircraft gas turbine engine includes sensors from which engine loading can be determined”).
Stocco, Clark and Lore are considered analogous art to the claimed invention because they are in the same field of endeavor of machine learning models with data drift/anomaly detection. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to have combined the drift detection system of Stocco and Clark with the idea of using sensors to monitor the operations of an aircraft taught by Lore. This would yield predictable results as Lore already shows that anomaly detection is already possible for monitoring an aircraft system.
Regarding claim 2, Stocco in view of Clark and lore teach claim 1 as outlined above. Stocco further teaches:
wherein the autoencoder includes an input layer, an encoder layer, and the decoder layer (2 Background subsection Autoencoders “The simplest form of autoencoder (SAE) is a three-layer DNN: the input layer, the hidden layer, and the output layer.”)
Regarding claim 3, Stocco in view of Clark and Lore teach claim 2 as outlined above. Stocco further teaches:
wherein the autoencoder is a three-layer autoencoder (2 Background subsection Autoencoders “The simplest form of autoencoder (SAE) is a three-layer DNN: the input layer, the hidden layer, and the output layer.”)
Regarding claim 4, Stocco in view of Clark and Lore teaches claim 1 as outlined above. Lore further teaches:
the first set of sensor data and the second set of sensor data are captured by one or more sensors that monitor operation of the aircraft system. ([0032] describes an example of using machine learning model where its data is gathered from an aircraft system. Specifically in lines 4-6 “for example, consider a non-limiting example wherein an aircraft gas turbine engine includes sensors from which engine loading can be determined”).
Regarding claim 5, Stocco in view of Clark and Lore teach claim 1 as outlined above. Clark further teaches:
wherein the percentile threshold is 90-99.7 (Page 4 first column 2nd paragraph “This threshold is set to the mean + 2 standard deviations of the outlier scores output by the model on the validation data.”)
Regarding claim 6, Stocco in view of Clark and Lore teach claim 1 as outlined above. Clark further teaches:
for a third empirical distribution of reconstruction errors of the second set of sensor data after decoding by the decoder layer, determine a first reconstruction error at the percentile threshold; compare the first reconstruction error of the second set of sensor data with the final value of the threshold; and determine that data drift is not present when the first reconstruction error is less than the final value of the threshold. (page 2 left column 2nd paragraph “In more detail, our method uses one class learning algorithms as base algorithms for outlier detection and considers the scores such methods output when presented with new instances. The method applies a statistical test to detect significant changes to the mean of these scores. Significant changes in mean scores suggest that the output of the outlier detection algorithm changed and that the threshold needs an update in order to function in the new environment. No significant change means that the system can continue proceeding as is.”)
Regarding claim 7, Stocco in view of Clark and Lore teach claim 6 as outlined above. Stocco further teaches:
wherein the testing controller is further configured to determine that data drift is present when the first reconstruction error is equal to or greater than the final value of the threshold (Example of Threshold Estimation “We use the estimated θ as threshold to distinguish anomalous conditions (reconstruction error ≥ θ) from normal ones (reconstruction error < θ).”).
Regarding claim 9, Stocco in view of Clark teaches claim 8 as outlined above. Neither Stocco or Clark teach the sensors monitor operation of an aircraft. However, Lore teaches using sensors monitoring the operations of an aircraft ([0032] describes an example of using machine learning model where its data is gathered from an aircraft system. Specifically in lines 4-6 “for example, consider a non-limiting example wherein an aircraft gas turbine engine includes sensors from which engine loading can be determined”).
Stocco, Clark and Lore are considered analogous art to the claimed invention because they are in the same field of endeavor of machine learning models with data drift/anomaly detection. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to have combined the drift detection system of Stocco and Clark with the idea of using sensors to monitor the operations of an aircraft taught by Lore. This would yield predictable results as Lore already shows that anomaly detection is already possible for monitoring an aircraft system.
Regarding claim 17, Stocco teaches the following:
A method for detecting drift… (Abstract
initializing the autoencoder, the autoencoder including an input layer, an encoder layer and a decoder layer (2 Background subsection Autoencoders “The simplest form of autoencoder (SAE) is a three-layer DNN: the input layer, the hidden layer, and the output layer.”)
training the autoencoder based on a first portion of a first set of sensor data, the first set of sensor data detected by a first plurality of sensors (5.4.4 SelfOracle’s Configurations “In the paper, authors reduced their training set to 600 images, which were resized to 120x90. During input validation, the three dimensional representation of an online input image is compared to the nominal images by measuring the average of the top-100 minimum distances from the training set.”)
setting a threshold to an initial value x at a percentile threshold of a first empirical distribution of reconstruction errors of the first portion of the first set of sensor data after decoding by the decoder layer, wherein the percentile threshold is 90 - 99.7; (Example of Threshold Estimation “Let us assume that we are willing to accept a false alarm rate ϵ = 10−2. The threshold θ with a probability mass above the threshold equal to 10−2 can be easily obtained as the inverse of the cumulative gamma distribution F (x): θ = F−1(1 − ϵ). This ensures that the cumulative probability of values ≤ θ is 1 − ϵ, leaving only a probability of ϵ to the tail following θ.”)
determining a final value of the threshold based on a result of the comparing wherein a value is determined based on a percentage of the reconstruction errors of the second portion of the first set of sensor data that are greater than the value of an initial threshold from a set value, and if the value is greater than a percentile threshold of the second empirical distribution of reconstruction errors of the second portion of the first set of sensor data, then the final value of the threshold is set equal to the value of the initial value x, and if the value is less than or equal to the percentile threshold, then the final value of the threshold is equal to a value at an inverse percentile of the percentile threshold; and (Example of Threshold Estimation “Let us assume that we are willing to accept a false alarm rate ϵ = 10−2. The threshold θ with a probability mass above the threshold equal to 10−2 can be easily obtained as the inverse of the cumulative gamma distribution F (x): θ = F−1(1 − ϵ). This ensures that the cumulative probability of values ≤ θ is 1 − ϵ, leaving only a probability of ϵ to the tail following θ.” Stocco does not update their threshold but does teach creating a threshold based on reconstruction error of multiple datasets.)
testing the autoencoder with a second set of sensor data from a second plurality of sensors (5.6 Results Effectiveness (RQ1) shows their testing.)
receiving, encoding and decoding the second set of sensor data with the autoencoder (2 Background subsection Autoencoders “An autoencoder (AE) is a DNN designed to reconstruct its own input. It consists of two sequentially connected components (an encoder, and a decoder) that are arranged symmetrically. The simplest form of autoencoder (SAE) is a three-layer DNN: the input layer, the hidden layer, and the output layer. The hidden layer encodes any given input x ∈ RD to its internal representation (code) z ∈ RZ with a function f (x) = z. Usually Z ≪ D. The output layer (decoder) decodes the encoded input with a reconstruction function д(z) = x′, where x′ is the reconstructed input x.”)
after training and the testing the autoencoder, detecting whether data drift is present in a third set of sensor data received from a third plurality of sensors, wherein when the data drift is not present, then the autoencoder is trained (Example of Threshold Estimation “We use the estimated θ as threshold to distinguish anomalous conditions (reconstruction error ≥ θ) from normal ones (reconstruction error < θ).”).
Stocco does not teach:
comparing a second empirical distribution of reconstruction errors of a second portion of the first set of sensor data to the first empirical distribution of reconstruction errors of the first portion of the first set of sensor data;
determining a final value of the threshold based on a result of the comparing
for a third empirical distribution of reconstruction errors of the second set of sensor data after decoding by the decoder layer, determining a first reconstruction error at the percentile threshold; comparing the first reconstruction error of the second set of sensor data with the final value of the threshold; determining that data drift is not present in the second set of sensor data when the first reconstruction error of the second set of sensor data is less than the final value of the threshold;
calculating a deviation output for one or more sensors of the second plurality of sensors
However, Clark does teach these
comparing a second empirical distribution of reconstruction errors of a second portion of the first set of sensor data to the first empirical distribution of reconstruction errors of the first portion of the first set of sensor data; determining a final value of the threshold based on a result of the comparing (Page 4 first column 3rd paragraph describes updating the threshold based on the previous iterations of data)
determining a final value of the threshold based on a result of the comparing (Page 4 first column 3rd paragraph describes updating the threshold based on the previous iterations of data)
for a third empirical distribution of reconstruction errors of the second set of sensor data after decoding by the decoder layer, determining a first reconstruction error at the percentile threshold; comparing the first reconstruction error of the second set of sensor data with the final value of the threshold; determining that data drift is not present in the second set of sensor data when the first reconstruction error of the second set of sensor data is less than the final value of the threshold; (page 2 left column 2nd paragraph “In more detail, our method uses one class learning algorithms as base algorithms for outlier detection and considers the scores such methods output when presented with new instances. The method applies a statistical test to detect significant changes to the mean of these scores. Significant changes in mean scores suggest that the output of the outlier detection algorithm changed and that the threshold needs an update in order to function in the new environment. No significant change means that the system can continue proceeding as is.”)
calculating a deviation output for one or more sensors of the second plurality of sensors (Page 4 2nd paragraph “This threshold is set to the mean + 2 standard deviations of the outlier scores output by the model on the validation data.”)
Stocco and Clark are considered analogous art to the claimed invention because they are in the same field of endeavor of data drift/anomaly detection. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to use the system and threshold of Stocco with the threshold updating of Clark. One of ordinary skill in the art would have been motivated to do this to minimize the error of the autoencoders (Clark abstract).
Stocco also does not teach the data being from sensors that monitor an operation of an aircraft system. However, Lore teaches using sensors monitoring the operations of an aircraft ([0032] describes an example of using machine learning model where its data is gathered from an aircraft system. Specifically in lines 4-6 “for example, consider a non-limiting example wherein an aircraft gas turbine engine includes sensors from which engine loading can be determined”).
Stocco, Clark and Lore are considered analogous art to the claimed invention because they are in the same field of endeavor of machine learning models with data drift/anomaly detection. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to have combined the drift detection system of Stocco and Clark with the idea of using sensors to monitor the operations of an aircraft taught by Lore. This would yield predictable results as Lore already shows that anomaly detection is already possible for monitoring an aircraft system.
Regarding claim 18, Stocco in view of Clark and Lore teaches claim 17 as outlined above. Stocco further teaches:
the first set of sensor data was randomized prior to receipt by the input layer. (5.4 Procedure “Our own implementation relaxed the restrictions above by considering a training set consisting of 3,000 randomly sampled images”)
Regarding claim 19, Stocco, in view of Clark and Lore teaches the elements of claim 17 as outlined above. Lore further teaches:
the aircraft system is an air compressor([0032] describes an example of using machine learning model where its data is gathered from an aircraft system. Specifically in lines 4-6 “for example, consider a non-limiting example wherein an aircraft gas turbine engine includes sensors from which engine loading can be determined”).
Regarding claim 20, Stocco in view of Clark and Lore teaches the elements of claim 17. Stocco further teaches:
receiving, encoding and decoding the third set of sensor data with the autoencoder (2 Background subsection Autoencoders “An autoencoder (AE) is a DNN designed to reconstruct its own input. It consists of two sequentially connected components (an encoder, and a decoder) that are arranged symmetrically. The simplest form of autoencoder (SAE) is a three-layer DNN: the input layer, the hidden layer, and the output layer. The hidden layer encodes any given input x ∈ RD to its internal representation (code) z ∈ RZ with a function f (x) = z. Usually Z ≪ D. The output layer (decoder) decodes the encoded input with a reconstruction function д(z) = x′, where x′ is the reconstructed input x.”)
Stocco does not teach:
for a fourth empirical distribution of reconstruction errors of the third set of sensor data after decoding by the decoder layer, determining a second reconstruction error at the percentile threshold; comparing the second reconstruction error of the third set of sensor data with the final value of the threshold; determining that data drift is not present in the third set of sensor data when the second reconstruction error of the third set of sensor data is less than the final value of the threshold; determining that data drift is present when the reconstruction error of the third set of sensor data is equal to or greater than the final value of the threshold;
calculating a deviation output for each sensor in the third plurality of sensors when data drift is present.
However, Clark does:
for a fourth empirical distribution of reconstruction errors of the third set of sensor data after decoding by the decoder layer, determining a second reconstruction error at the percentile threshold; comparing the second reconstruction error of the third set of sensor data with the final value of the threshold; determining that data drift is not present in the third set of sensor data when the second reconstruction error of the third set of sensor data is less than the final value of the threshold; determining that data drift is present when the reconstruction error of the third set of sensor data is equal to or greater than the final value of the threshold; (page 2 left column 2nd paragraph “In more detail, our method uses one class learning algorithms as base algorithms for outlier detection and considers the scores such methods output when presented with new instances. The method applies a statistical test to detect significant changes to the mean of these scores. Significant changes in mean scores suggest that the output of the outlier detection algorithm changed and that the threshold needs an update in order to function in the new environment. No significant change means that the system can continue proceeding as is.”)
calculating a deviation output for each sensor in the third plurality of sensors when data drift is present. (Page 4 2nd paragraph “This threshold is set to the mean + 2 standard deviations of the outlier scores output by the model on the validation data.”)
Regarding claim 21, Stocco in view of Clark and Lore teaches claim 9 as outlined above. Lore further teaches:
the aircraft system is an air compressor ([0032] describes an example of using machine learning model where its data is gathered from an aircraft system. Specifically in lines 4-6 “for example, consider a non-limiting example wherein an aircraft gas turbine engine includes sensors from which engine loading can be determined”).
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Stocco in view of Clark and Zhou (Pub. No.: US 20220116281 A1, with priority to CN 201910898740.4 filed 09/23/2019)
Regarding claim 15, Stocco in view of Clark teaches claim 14 as outlined above. Neither of them teach:
wherein the encoder layer has a third quantity of neurons, further the third quantity is 60-75% of the first quantity
However, Zhou does:
wherein the encoder layer has a third quantity of neurons, further the third quantity is 60-75% of the first quantity (Zhou [0025] “the hidden layer [equivalent to encoder layer] is less than that of the input layer and that of the output layer”. This is saying it the quantity of neurons in the hidden layer could be a percentage of the quantity of neurons in the input and output layer).
Stocco, Clark and Zhou are consider analogous art to the claimed invention because they are in the same field of endeavor of machine learning models, specifically autoencoders. It would have been obvious to one of ordinary skill in the art before the effective filing data of the claimed invention to have combined the drift detection system of Stocco and Clark with the autoencoder of Zhou. One of ordinary skill in the art would want to have the quantity of neurons in the encoder layer to be less for better performance.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL PATRICK GRUSZKA whose telephone number is (571)272-5259. The examiner can normally be reached M-F 9:00 AM - 6:00 PM ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Li Zhen can be reached at (571) 272-3768. 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.
/DANIEL GRUSZKA/Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights