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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on November 18, 2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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 1-2, 7, 10, 12, 15, 17, and 21-23 are rejected under 35 U.S.C. 103 as being unpatentable over Bathen et al. (U.S. Patent Publication No. 2019/0147300) (“Bathen”) in view of Lu et al. (GAN-Based Data Augmentation Strategy for Sensor Anomaly Detection in Industrial Robots) (“Lu”).
Regarding claim 1, Bathen teaches a computer implemented method for identifying anomalies in time-series data, the method comprising: receiving a time-series data comprising a sequence of data values, each data value associated with a time value, and wherein the time-series data represents computing network resource usage by a communication channel (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.”; [0040] “At step S202, data collection module 132 communicates with historical data device 110 and monitored device 112 to retrieve data such as the historical time series data and the monitored time series data set, accordingly.” Bathen provides receiving time series data, which may include time series data of the monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like, corresponding to receiving a time-series data comprising a sequence of data values, each data value associated with a time value, and wherein the time-series data represents computing network resource usage by a communication channel.); identifying a time window representing a range of time values (Bathen [0016] “In an example, the model may be generated for use by the system to forecast or predict a value expected of a future data point in a monitored time series data set based on the previous data points of the monitored time series data set” Bathen provides a time-series data set comprising time data values, which corresponds to identifying a time window representing a range of time values.); training a neural network to identify anomalies associated with the computing network resource usage (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.”; [0033] “Training module 134 trains pairs of neural networks simultaneously for use in detecting anomalies in the monitored time series data set. In an embodiment of the present invention, training module 134 may train a pair of individual recurrent multilayer perceptron neural networks simultaneously by implementing, for example, an adversarial training process.” Bathen provides training a neural network to identify anomalies associated with a computing network resource usage.), …the anomaly indicating a shortage of a computing network resource for a task (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.”; [0048] “In an embodiment of the present invention, an action or operation such as a security action, a maintenance action, or the like, may be generated and implemented in response to detecting the anomalies. …The maintenance action may include, for example, halting or otherwise controlling operation of a device based on detected anomalies indicative of potential or impending operational issues, or the like.” Bathen provides identifying anomalies in time series data including with respect to usage of a processor, a network transmission rate, main memory utilization, or the like, corresponding to an anomaly indicating shortage of a computing network resource for a task.); and allocating an additional network bandwidth to the communication channel (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.”; [0048] “In an embodiment of the present invention, an action or operation such as a security action, a maintenance action, or the like, may be generated and implemented in response to detecting the anomalies. …The maintenance action may include, for example, halting or otherwise controlling operation of a device based on detected anomalies indicative of potential or impending operational issues, or the like.”; [0063] “Cloud computing is a model of service delivery for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, network bandwidth, servers, processing, memory, storage, applications, virtual machines, and services) that can be rapidly provisioned and released with minimal management effort or interaction with a provider of the service.” Bathen provides allocating network bandwidth and monitoring time series data representative of resource usage and subsequently performing maintenance actions including halting or otherwise controlling operation of a device based on detected anomalies indicative of potential or impending operational issues, corresponding to allocating an additional network bandwidth to the communication channel).
Bathen fails to teach …wherein the neural network is trained for a predetermined number of iterations that is configurable or set to a default value, and the predetermined number of iterations is less than a number of iterations needed for an aggregate loss value associated with the sequence of data values to be below a loss threshold value, the training comprising: for time values of the time-series data comprising the sequence of data values: executing the neural network to predict a sequence of predicted data values corresponding to the time values; determining one or more loss values based on differences between the sequence of predicted data values and the sequence of data values; and adjusting parameters of the neural network based on the aggregate loss value of the one or more loss values; and determining, using the neural network trained for the predetermined number of iterations, a first loss value associated with the time-series data comprising the sequence of data values, the first loss value corresponding to a first time value of the one or more time values exceeding a threshold; identifying a first data value of the one or more data values corresponding to the one or more time value as an anomaly based on the first loss value.
However, Lu teaches …wherein the neural network is trained for a predetermined number of iterations that is configurable or set to a default value (Lu Algorithm 1, line 4 “for number i of training iterations do…”; Section IV MSGAN-Based Data Augmentation Algorithm “Generator represents a mapping from latent noise space to real sample space as G(z,θg),where G is a neural network with parameters θg. Correspondingly, the discriminator D(x,θd) is also defined by a neural network with parameters θd… The primary training process of MSGAN is shown in Algorithm1… The procedures iterate until convergence or reaching the predefined training iteration”; Section 5 Performance evaluation “.We train MSGAN and original GAN for 3k iterations and use Adam optimizer” Lu provides an iterative neural network training algorithm, as shown in Algorithm 1, which is configurable or set to a default value for a predefined number i, including for example, 3k iterations (i=3000).), and the predetermined number of iterations is less than a number of iterations needed for an aggregate loss value associated with the sequence of data values to be below a loss threshold value (Lu Algorithm 1, line 4 “for number i of training iterations do…”; Section IV MSGAN-Based Data Augmentation Algorithm “The primary training process of MSGAN is shown in Algorithm1… The procedures iterate until convergence or reaching the predefined training iteration” Lu provides terminating training for a predefined number of iterations, which may happen before convergence, since the algorithm continues until convergence or reaching the predefined training iteration, therefore providing the predetermined number of iterations is less than that needed for convergence, which happens when an aggregate loss value associated with the sequence of data values to be below a loss threshold value.), the training comprising: for time values of the time-series data comprising the sequence of data values (Lu Section I Introduction “Specifically, the offline phase reads historical data from the database for model training, i.e., timeseries forecasting and anomaly detection. For time series forecasting, the raw sensor data is segmented through sliding windows to train the time series model with multi-step intervals.”; Section III(4) Model Selection and Training “Currently, multi-step ahead samples are generated for time series prediction in offline training step (2)” Lu provides training for timeseries forecasting comprising training with time-series samples, such as the raw sensor data, which is segmented through sliding windows.): executing the neural network to predict a sequence of predicted data values corresponding to the time values (Lu Section III Methodology “Since the purpose of incremental learning is to enable the model to make correct predictions on new data, regular updates and threshold updates 391 can be combined to improve model prediction performance, which is also the method used in this work.”; Section IV MSGAN-Based Data Augmentation Algorithm “The primary training process of MSGAN is shown in Algorithm 1. We first randomly initialize the parameters of generator and discriminator as θg and θd, respectively (lines1-2). Then, we sample ˜z from latent spatial to generate fake samples through the generator Gθg (˜z) in line 7.” Lu provides executing the generator neural network, in accordance with training Algorithm 1 to generate fake samples, which are being interpreted as predictions, wherein the fake samples are time-series samples and therefore a sequence of predicted data values corresponding to the time values.); determining one or more loss values based on differences between the sequence of predicted data values and the sequence of data values (Lu Algorithm 1; Section IV MSGAN-Based Data Augmentation Algorithm “That is: Equation (6), where Lc G and Lp G represent the generator loss in the current iteration and previous iteration, respectively. Similarly, Lc Dand Lp D represent the discriminator loss in the current iteration and previous iteration, respectively…To perform automatically update both generator and discriminator, it is necessary to compare the loss Lc G and Lc D. However, as the losses on different scales, they cannot compare directly. To tackle this issue, we adopt the method proposed in [21],which applies relative loss with normalizing to compute the difference between the current iteration and the previous iteration. The relative loss can be defined as Equations 9-10… The primary training process of MSGAN is shown in Algorithm 1. We first randomly initialize the parameters of generator and discriminator as θg and θd, respectively (lines1-2). Then, we sample ˜z from latent spatial to generate fake samples through the generator Gθg (˜z) in line 7.” Lu provides calculating a plurality of loss values, including a generator loss, for determining a loss value for the fake samples generated by the generator, as shown in Algorithm 1.); and adjusting parameters of the neural network based on the aggregate loss value of the one or more loss values (Lu Algorithm 1, line 2 “generator parameters θg; discriminator parameters θd”; Section IV MSGAN-Based Data Augmentation Algorithm “Generator represents a mapping from latent noise space to real sample space as 398 G(z,θg), where G is a neural network with parameters θg. Correspondingly, the discriminator D(x,θd) is also defined by a neural network with parameters θd… To perform automatically update both generator and discriminator, it is necessary to compare the loss Lc G and Lc D. However, as the losses on different scales, they cannot compare directly. To tackle this issue, we adopt the method proposed in [21],which applies relative loss with normalizing to compute the difference between the current iteration and the previous iteration. The relative loss can be defined as Equations 9-10… The primary training process of MSGAN is shown in Algorithm 1. We first randomly initialize the parameters of generator and discriminator as θg and θd, respectively (lines1-2). Then, we sample ˜z from latent spatial to generate fake samples through the generator Gθg (˜z) in line 7.” Lu provides an iterative process or adjusting the parameters of the generator and discriminator, as shown in Algorithm 1, including the iteratively calculated generator and discriminator loss, corresponding to the aggregate loss.); and determining, using the neural network trained for the predetermined number of iterations, a first loss value associated with the time-series data comprising the sequence of data values, the first loss value corresponding to a first time value of the one or more time values exceeding a threshold (Lu Algorithm 1, line 9 “if rD>rG then…”; Section IV MSGAN-Based Data Augmentation Algorithm “The relative loss can be defined as rG, rD in Equations 9 and 10. If rG>rD, we perform the update of generator, or otherwise update the discriminator.” Lu provides determining loss values in Algorithm 1, which includes using the generator neural network trained for a predetermined number of iterations to determine a generator loss, as shown above, including a threshold for the relative loss values, including if rG>rD, as shown in line 9 of Algorithm 1, wherein the loss value rG exceeds threshold rD.); identifying a first data value of the one or more data values corresponding to the one or more time value as an anomaly based on the first loss value (Lu Section IV MSGAN-Based Data Augmentation Algorithm “Generator represents a mapping from latent noise space to real sample space as 398 G(z,θg), where G is a neural network with parameters θg. Correspondingly, the discriminator D(x,θd) is also defined by a neural network with parameters θd… To perform automatically update both generator and discriminator, it is necessary to compare the loss Lc G and Lc D. However, as the losses on different scales, they cannot compare directly. To tackle this issue, we adopt the method proposed in [21],which applies relative loss with normalizing to compute the difference between the current iteration and the previous iteration. The relative loss can be defined as Equations 9-10… The primary training process of MSGAN is shown in Algorithm 1. We first randomly initialize the parameters of generator and discriminator as θg and θd, respectively (lines1-2). Then, we sample ˜z from latent spatial to generate fake samples through the generator Gθg (˜z) in line 7.” Section V(2) Anomaly Detection “After the generator training is completed, new synthetic samples are generated by the generator for anomaly detection.” Lu provides identifying anomalies after training the neural network, which includes calculating a loss for the generator and discriminator, thus identify an anomaly for time series data based on the first loss value (i.e., the generator loss).).
Bathen and Lu are both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to time-series data. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen with the above teachings of Lu. Doing so would allow for an improved detection accuracy (Lu Section V Performance Metrics “The results show that with the decrease of the imbalance ratio, the detection accuracy improves gradually.”)
Regarding claim 2, Bathen in view of Lu teaches the computer implemented method of claim 1 as discussed above in the rejection of claim 1, wherein the one of the anomalies is a point time anomaly (Bathen [0015] “An anomaly may include, for example, an anomalous data point in the time series data set having a value significantly different than that which may be expected, such as with respect to values of one or more previous data points in the data set and a trend or pattern exhibited or supported by the previous data points.” Bathen provides anomalies as data points in a time series data set, corresponding to the time anomaly is a point time anomaly.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 7, Bathen in view of Lu teaches the computer implemented method of claim 1 as discussed above in the rejection of claim 1, further comprising: identifying a potential resource failure including the shortage of the computing network resource based on the anomaly (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.”; [0048] “In an embodiment of the present invention, an action or operation such as a security action, a maintenance action, or the like, may be generated and implemented in response to detecting the anomalies. …The maintenance action may include, for example, halting or otherwise controlling operation of a device based on detected anomalies indicative of potential or impending operational issues, or the like.” Bathen provides identifying anomalies in usage of a processor, a network transmission rate, or main memory utilization corresponding to identifying a potential resource failure including the shortage of the computing network resource based on the anomaly.); and sending a message reporting the potential resource failure (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.”; [0048] “In an embodiment of the present invention, an action or operation such as a security action, a maintenance action, or the like, may be generated and implemented in response to detecting the anomalies. …The maintenance action may include, for example, halting or otherwise controlling operation of a device based on detected anomalies indicative of potential or impending operational issues, or the like.” Bathen provides maintenance actions for identified anomalies which may include, for example, halting or otherwise controlling operation of a device based on detected anomalies indicative of potential or impending operational issues, or the like, corresponding to sending a message reporting the potential resource failure.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 10, Bathen in view of Lu teaches the computer implemented method of claim 1 as discussed above in the rejection of claim 1, wherein the neural network is a multi-layered perceptron configured to receive a scalar input and output a scalar value (Bathen [0033] “In an embodiment of the present invention, training module 134 may train a pair of individual recurrent multilayer perceptron neural networks simultaneously by implementing, for example, an adversarial training process”; [0043] “In the embodiment, where each of the neural networks are defined by multilayer perceptrons, training module 134 may utilize backpropagation and dropout algorithms in training the neural networks.” Bathen provides multilayer perceptrons, corresponding to the neural network is a multi-layered perceptron configured to receive a scalar input and output a scalar value.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 12, it is the non-transitory computer readable storage medium embodiment of claim 1 with similar limitations to claim 1 and is rejected using the same reasoning found above in the rejection of claim 1. Further, Bathen teaches a non-transitory computer readable storage medium storing instructions (Bathen [0055] “A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.” Bathen provides a non-transitory computer readable storage medium.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 15, the rejection of claim 12 is incorporated herein. Further, the limitations in this claim are taught by Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 7.
Regarding claim 17, it is the computer system comprising one or more computer processors; and non-transitory computer readable storage medium storing instructions embodiment of claim 1 with similar limitations to claim 1 and is rejected using the same reasoning above in the rejection of claim 1. Further, Bathen teaches one or more computer processors (Bathen [0059] “These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine” Bathen provides one or more computer processors.); and non-transitory computer readable storage medium storing instruction (Bathen [0055] “A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.” Bathen provides a non-transitory computer readable storage medium.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 1.
Regarding claim 21, Bathen in view of Lu teaches the non-transitory computer readable storage medium of claim 15 as discussed above in the rejection of claim 15, wherein the computing resource is one of: a processing resource, a memory resource, or a storage resource (Bathen [0029] “In an embodiment of the present invention, the monitored time series data set may include, for example, multidimensional operational or system parameter time series data of monitored device 112, such as with respect to usage of a processor, a network transmission rate, main memory utilization, or the like.” Bathen teaches processor and memory resources.).
It would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 15.
Regarding claim 22, the rejection of claim 17 is incorporated herein. Further, the limitations in this claim are taught by Bathen in view of Lu for the same reasons disclosed above in the rejection of claim 7.
Regarding claim 23, Bathen in view of Lu teaches the computer implemented method of claim 1, wherein the predetermined number of iterations is independent of any convergence value associated with the aggregate loss (Lu Algorithm 1, lines 3-4; Section IV MSGAN-Based Data Augmentation Algorithm “The primary training process of MSGAN is shown in Algorithm1… The procedures iterate until convergence or reaching the predefined training iteration” Lu provides two independent ending criteria for Algorithm 1, as shown in lines 3 and 4, including a predetermined number of iterations in line 4 and a convergence value in line 3.).
Bathen and Lu are both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to time-series data. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen with the above teachings of Lu. Doing so would allow for an improved detection accuracy (Lu Section V Performance Metrics “The results show that with the decrease of the imbalance ratio, the detection accuracy improves gradually.”)
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Bathen et al. (U.S. Patent Publication No. 2019/0147300) (“Bathen”) in view of Lu et al. (GAN-Based Data Augmentation Strategy for Sensor Anomaly Detection in Industrial Robots) (“Lu”) in further view of Seo et al. (U.S. Patent Publication No. 2020/0380409) (“Seo”).
Regarding claim 3, Bathen in view of Lu teaches the computer implemented method of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the training further comprises at a start of training, assigning random values to the parameters of the neural network.
However, Seo teaches wherein the training further comprises: at a start of training, assigning random values to the parameters of the neural network (Seo [0067] “In operation 702, the policy recommendation module 212 generates a plurality of data augmentation policies to be applied to raw space augmentation or feature space augmentation. In this case, the policy recommendation module 212 may receive the data augmentation policies from a user or randomly generate the data augmentation policies.”; [0069] “In operation 706, the policy recommendation module 212 performs model learning by applying the selected policy to training data.” Seo provides data augmentation policies for training neural networks including randomly generating the data augmentation policies, corresponding to at a start of training, assigning random values to the parameters of the neural network.).
Bathen, Lu, and Seo all both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to time-series data. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu with the above teachings of Seo. Doing so would allow for unnecessary computation to be reduced and efficiency of an optimal policy determination process to be improved (Seo [0073] “As such, when the separate early stopping condition is used, unnecessary computation may be reduced and efficiency of an optimal policy determination process may be improved.”).
Claims 4, 11, 13, 16, 18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Bathen et al. (U.S. Patent Publication No. 2019/0147300) (“Bathen”) in view of Lu et al. (GAN-Based Data Augmentation Strategy for Sensor Anomaly Detection in Industrial Robots) (“Lu”) in further view of Tora et al. (U.S. Patent Publication No. 2021/0397938) (“Tora”).
Regarding claim 4, Bathen in view of Lu teaches the computer implemented method of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the time window is a first time window, the one or more data values are one or more first data values, the one or more loss values are one or more first loss values, the anomaly is a first anomaly, and the range of time values is a first range of time values, the method comprising: identifying a second time window representing a second range of time values; discarding the training of the neural network by reinitializing the neural network for the second time window; training the reinitialized neural network for the predetermined number of iterations using one or more second data values of the second time window; and responsive to a second loss value corresponding to a second time value of the second time window exceeding the threshold, identifying the second data value corresponding to the second time value as a second anomaly.
However, Tora teaches wherein the time window is a first time window (Tora [0069] “Accordingly, the preprocessing unit 131 divides the data for training that is time-series data by a sliding window for predetermined periods.” Tora provides a sliding window for a plurality of predetermined periods corresponds to a first time window.), the one or more data values are one or more first data values (Tora [0069] “The generation unit 132 then generates models on the basis of each piece of data per sliding window divided by the preprocessing unit 131.” Tora provides one or more data values.), the one or more loss values are one or more first loss values (Tora [0102] “The detection device 10 detects anomalies on the basis of the magnitude of error between data at a certain clock time that is actually collected and predicted values at that clock time. For example, in a case where the magnitude of error exceeds a threshold value, the detection device 10 detects that an anomaly has occurred at that clock time.” Tora provides error values corresponding to loss values.), the anomaly is a first anomaly (Tora [0102] “In the method of constructing a prediction model, the detection device 10 detects anomalies on the basis of error between the values of original data and predicted values, instead of reconstruction error.” Tora provides anomalies.), and the range of time values is a first range of time values (Tora [0069] “Accordingly, the preprocessing unit 131 divides the data for training that is time-series data by a sliding window for predetermined periods.” Tora provides a sliding window for a plurality of predetermined periods corresponding to a range of time values.), the method comprising: identifying a second time window representing a second range of time values (Tora [0069] “Accordingly, the preprocessing unit 131 divides the data for training that is time-series data by a sliding window for predetermined periods.” Tora provides a sliding window for a plurality of predetermined periods corresponds to a second time window.); discarding the training of the neural network by reinitializing the neural network for the second time window (Tora [0062] “When training a model, there are cases in which initial values and so forth of model parameters are randomly decided. For example, when training a model using a neural network including an autoencoder, there are cases in which initial values such as weighting between nodes and so forth are randomly decided.”; [0069] “Accordingly, the preprocessing unit 131 divides the data for training that is time-series data by a sliding window for predetermined periods. The generation unit 132 then generates models on the basis of each piece of data per sliding window divided by the preprocessing unit 131. The generation unit 132 may also perform both generation of models based on training data of a fixed period (fixed-period learning) and generation of models based on training data of each of periods into which the fixed period has been divided by the sliding window (sliding learning). Also, in sliding learning, instead of using all models generated on the basis of data that has been divided for each sliding window, one can be selected and used therefrom. For example, applying a model created using data backtracking for a predetermined period from the previous day, for anomaly detection for the following one day, may be repeated.” Tora provides sliding window fixed-period learning, wherein separate models are generated for each time window and separate training is performed for each window specific model by a neural network, corresponding to discarding the training of the neural network by reinitializing the neural network for the second time window); training the reinitialized neural network for the predetermined number of iterations using one or more second data values of the second time window (Tora [0069] “Accordingly, the preprocessing unit 131 divides the data for training that is time-series data by a sliding window for predetermined periods. The generation unit 132 then generates models on the basis of each piece of data per sliding window divided by the preprocessing unit 131. The generation unit 132 may also perform both generation of models based on training data of a fixed period (fixed-period learning) and generation of models based on training data of each of periods into which the fixed period has been divided by the sliding window (sliding learning). Also, in sliding learning, instead of using all models generated on the basis of data that has been divided for each sliding window, one can be selected and used therefrom. For example, applying a model created using data backtracking for a predetermined period from the previous day, for anomaly detection for the following one day, may be repeated.” Tora provides fixed-period learning, wherein separate models are generated for each time window and training is performed for each window specific model, corresponding to training the reinitialized neural network for the predetermined number of iterations using one or more second data values of the second time window.); and responsive to a second loss value corresponding to a second time value of the second time window exceeding the threshold, identifying the second data value corresponding to the second time value as a second anomaly (Tora [0102] “The detection device 10 detects anomalies on the basis of the magnitude of error between data at a certain clock time that is actually collected and predicted values at that clock time. For example, in a case where the magnitude of error exceeds a threshold value, the detection device 10 detects that an anomaly has occurred at that clock time.” Tora provides determining anomalies responsive to error values exceeding a threshold, corresponding to responsive to a second loss value corresponding to a second time value of the second time window exceeding the threshold, identifying the second data value corresponding to the second time value as a second anomaly.).
Bathen, Lu, and Tora are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically applied to time series data. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Lu with the above teachings of Tora. Doing so would allow for appropriate model selection and thus an improvement in detection precision (Tora [0085] “Thus, according to the present embodiment, preprocessing and selection of training data, and model selection can be appropriately performed in a case of performing anomaly detection using deep learning, and detection precision can be improved.”)
Regarding claim 11, Bathen in view of Lu teaches the computer implemented method of claim 1 as discussed above in the rejection of claim 1, but fails to teach further comprising: adjusting the threshold based on comparison of the anomaly and known anomalies; and using the adjusted threshold value for identifying the anomalies for one or more other time windows.
However, Tora teaches further comprising: adjusting the threshold based on comparison of the anomaly and known anomalies (Tora [0056] “FIG. 6 is a diagram for explaining about identifying feature values with little change. The tables at the upper portion of FIG. 6 show, when calculating standard deviation (STD) of training data of feature values, and setting threshold values, the number of feature values that come under the threshold values. For example, in a case in which a threshold value is set to 0.1, the number of feature values where STD 0.1 (number of performance values of Group1) is 132. Also, at this time, the number of feature values where STD<0.1 (number of performance values of Group2) is 48.” Tora provides setting a threshold based on comparisons of training data for anomaly detection corresponding to adjusting the threshold based on comparison of the anomaly and known anomalies); and using the adjusted threshold for identifying other anomalies for one or more other time windows (Tora [0102] “For example, in a case where the magnitude of error exceeds a threshold value, the detection device 10 detects that an anomaly has occurred at that clock time.” Tora provides using the threshold for identifying anomalies.).
Bathen, Lu and Tora are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically predictive modeling of time-series data. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bathen in view of Seo with the above teachings of Tora. Doing so would allow for appropriate model selection and thus an improvement in detection precision (Tora [0085] “Thus, according to the present embodiment, preprocessing and selection of training data, and model selection can be appropriately performed in a case of performing anomaly detection using deep learning, and detection precision can be improved.”)
Regarding claim 13, the rejection of claim 12 is incorporated herein. Further, the limitations in this claim are taught by Bathen in view of Lu in further view of Tora for the same reasons disclosed above in the rejection of claim 4.
Regarding claim 16, the rejection of claim 12 is incorporated herein. Further, the limitations in this claim are taught by Bathen in view of Lu in further view of Tora for the same reasons disclosed above in the rejection of claim 11.
Regarding claim 18, the rejection of claim 17 is incorporated herein. Further, the limitations in this claim are taught by Bathen in view of Lu in further view of Tora for the same reasons disclosed above in the rejection of claim 4.
Regarding claim 20, the rejection of claim 17 is incorporated herein. Further, the limitations in this claim are taught by Bathen in view of Lu in further view of Tora for the same reasons disclosed above in the rejection of claim 11.
Response to Arguments
Regarding the claim objections set forth in the Non-Final Rejection, dated October 01, 2025, Applicant’s amendments overcome the objections.
Regarding the rejection applied under 35 U.S.C. 103, Applicant’s arguments with respect to claims 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.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KURT NICHOLAS PRESSLY whose telephone number is (703)756-4639. The examiner can normally be reached M-F 8-4.
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, Kamran Afshar can be reached at (571) 272-7796. 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.
/KURT NICHOLAS PRESSLY/Examiner, Art Unit 2125
/KAMRAN AFSHAR/Supervisory Patent Examiner, Art Unit 2125
Prosecution Insights