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 .
This is a non-final office action. Claims 1-20 were considered.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 07/30/2025 has been entered.
Response to Amendment
This action is in response to communication filed on 06/09/2025.
a. Claims 1-20 are pending in this application.
b. Claims 1-4, 12-15 and 17-19 has been amended.
Response to Arguments Regarding Claim Rejections – 35 USC § 103
Applicant's arguments, see page 7-11 of REMARKS, filed on 06/09/2025, with respect to Claim Rejections - 35 USC § 103 have been fully 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 § 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.
Claim(s) 1, 8, 10-11, 12, 18 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Jordan et al. (US 2018/0375893 A1, hereinafter Jordan) in view of Zhang et al. (US 20240169258 A1, hereinafter Zhang).
Regarding claim 1, Jordan teaches a non-transitory computer-readable medium configured to store computer logic having instructions that, when executed, cause one or more processing devices to perform the steps of
receiving a time-series dataset having a sequence of datapoints each including a set of Performance Monitoring (PM) parameters of a network ([22-24]: In step 102 of the method 100, event records are received in a timeframe. The event records may be collected from network security devices, application flows, firewalls, anti-virus systems, intrusion detection systems, intrusion prevention systems, anti-malware systems, operating systems, applications, workstations, switches, routers, Active Directory (AD), Lightweight Directory Access Protocol (LDAP), High Availability Dynamic Host Configuration Protocol (HA-DHCP), security information and event management (SIEM) systems, or the like. Each of the event records may include attributes and/or values for alerts and events and may further include a unique key.);
applying a subset of the sequence of datapoints to a Machine Learning (ML) model having a classification function (Fig. 2 and [24-25, 29]: The method 200 begins with parsing an event record in step 202. In step 204, once an event record is parsed, the event record is sent to various threat vector calculators. In step 204, once an event record is parsed, the event record is sent to various threat vector calculators. In step 240, the event records are analyzed by a machine learning system. The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation. Each time that a specified attribute-value pairing is detected, the attribute and/or value may be stored (i.e. applying learning systems to classify event record)) and an encoding/decoding function ([24-25, 29]: The method 200 begins with parsing an event record in step 202. Each of the event records may include strings of data which may be parsed in order to obtain attributes, values, attribute-value pairs, or the like. The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation (i.e. decoding the event records)), and
allowing the ML model to leverage temporal-based correlations among the datapoints of the subset to predict an event associated with the network ([29]: The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation. Each time that a specified attribute-value pairing is detected, the attribute and/or value may be stored. A threat vector may be assigned to an event record depending on the attributes and/or values stored for that event record and sent to step 106. For example, the machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system).
Jordan however does not teach wherein the temporal-based correlations are based on differences between numerical PM parameters derived from two or more datapoints from sequential timestamps defined by the time-series dataset.
Zhang teaches wherein the temporal-based correlations are based on differences between numerical PM parameters derived from two or more datapoints from sequential timestamps defined by the time-series dataset ([23, 33]: FIG. 3 illustrates a representation 300 of time-series anomaly detection. As shown, the representation 300 includes a time-series 302 of continuously observed values y∈
PNG
media_image1.png
38
29
media_image1.png
Greyscale
(y-axis) and timestamps t1, . . . , tn∈
PNG
media_image1.png
38
29
media_image1.png
Greyscale
(x-axis) indicating times at which corresponding observed values are received/observed. [56]: Δ represents a difference between a current observation and a previous observation, e.g., Δ=|ti−ti−1|) (i.e. correlations are based on difference between parameters from two datapoints in time-series dataset)).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan to incorporate the teachings of Zhang and the temporal-based correlations are based on differences between numerical PM parameters derived from two or more datapoints from sequential timestamps defined by the time-series dataset. One of ordinary skilled in the art would have been motivated to combine the teachings in order to utilize algorithms to learn from, and make predictions on, known data by analyzing the known data (Zhang, [34]).
Regarding claim 8, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 1.
Jordan teaches wherein the subset is defined by a sliding window ([07, 42]: A device includes a system that receives or retrieves a plurality of events in a sliding window), and wherein each datapoint represents the PM parameters for a time period ([42]: The time frame may be 10 seconds, 30 seconds, 1 minute, 2 minutes, 5 minutes, or any other period of time. Alternatively, the time frame may comprise a sliding window.).
Zhang teaches the subset is defined by a gradient boosted sliding window ([59-60]: The training module 202 shifts a sliding window to a current position 306 from a previous position 308. For example, the sliding window defines a subset of recent observed values included in the time-series 302 such as a most recent 10 observed values, a most recent 20 observed values, a most recent 30 observed values, and so forth.).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to further incorporate the teachings of Zhang and the subset is defined by a gradient boosted sliding window. One of ordinary skilled in the art would have been motivated to combine the teachings in order to utilize algorithms to learn from, and make predictions on, known data by analyzing the known data (Zhang, [34]).
Regarding claim 10, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 1.
Jordan teaches wherein the instructions further cause the one or more processing devices to perform the steps of allowing the ML model to leverage spatial-based metrics related to neighboring components arranged within the network ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106.), and
using the temporal-based correlations ([29]: The machine learning system may differ from the attribute tracking system in that the machine learning maintains and analyzes the value of all attributes received, rather than simply tracking a count or statistical representation. The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation. Each time that a specified attribute-value pairing is detected, the attribute and/or value may be stored. A threat vector may be assigned to an event record depending on the attributes and/or values stored for that event record and sent to step 106 (i.e. tracking the attributes and values is the temporal-based correlation)) and the spatial-based metrics to predict the event associated with the network ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106 (i.e. learning curve is a graph therefore it is a spatial-based metric)).
Regarding claim 11, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 10.
Jordan teaches wherein the ML model is configured to utilize the spatial-based metrics by using one or more of a graph-based dataset, a graph-based ML function ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106 (i.e. using learning curve is a graph based ML function)), and a Graph Neural Network (GNN).
Regarding Claims 12 and 18, they do not teach or further define over claim 1. Therefore, claim 12 and 18 are rejected for the same reason as set forth above in claim 1.
Regarding claim 20, Jordan in view of Zhang teaches the system of claim 18.
Jordan teaches wherein the instructions further enable the processing device to allow the ML model to leverage spatial-based metrics related to neighboring components arranged within the network ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106.),
utilize the temporal-based correlations ([29]: The machine learning system may differ from the attribute tracking system in that the machine learning maintains and analyzes the value of all attributes received, rather than simply tracking a count or statistical representation. The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation. Each time that a specified attribute-value pairing is detected, the attribute and/or value may be stored. A threat vector may be assigned to an event record depending on the attributes and/or values stored for that event record and sent to step 106 (i.e. tracking the attributes and values is the temporal-based correlation)) and the spatial-based metrics to predict the event associated with the network ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106 (i.e. learning curve is a graph therefore it is a spatial-based metric)),
wherein leveraging the spatial-based metrics includes utilizing one or more of a graph-based dataset, a graph-based ML function ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106 (i.e. using learning curve is a graph based ML function)), and a Graph Neural Network (GNN).
Claim 2 and 13 rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang further in view of Nguyen et al. (US 20210049700 A1, hereinafter Nguyen) further in view of Yadav et al. (US 9838317 B1, hereafter Yadav).
Regarding claim 2, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 1.
Jordan in view of Zhang however does not teach wherein the ML model jointly encodes these temporal-based correlations into a latent space using an autoencoder and classifies events directly from this latent representation using a gradient-boosted classification model, and wherein the predicted event is related to at least one of (a) a change in an Interior Gateway Protocol (IGP) configuration in the network and (b) one or more flapping links in the network.
Nguyen teaches wherein the ML model jointly encodes these temporal-based correlations into a latent space using an autoencoder (Fig. 7 and [188]: The first RNN layer is configured to receive the numerical time-series data and the averaged or filtered numerical time series data and transform one or more feature data structures within the numerical time-series data and one or more feature data structures of the averaged numerical time-series data into a latent feature representation the output of the first LSTM layer 704 is provided as an input to the second LSTM layer 706 (i.e. encoding time-series data into a latent space)) and classifies events directly from this latent representation using a gradient-boosted classification model ([207]: Output a latent representation of each of the four input variables to the second RNN layer 706, which in turn converts the latent representation of each of the four input variables into a future feature data structure having a future feature value of a price moving average output 702 (e.g., a forecast for the next 30 days or 252 days of the moving average) (i.e. classify latent representation into future feature data structure)).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to incorporate the teachings of Nguyen and the ML model jointly encodes these temporal-based correlations into a latent space using an autoencoder and classifies events directly from this latent representation using a gradient-boosted classification model. One of ordinary skilled in the art would have been motivated to combine the teachings in order for reasonable approximation of the future feature value (Nguyen, [211]).
Zhang in view of Nguyen however does not teach wherein the predicted event is related to at least one of (a) a change in an Interior Gateway Protocol (IGP) configuration in the network and (b) one or more flapping links in the network.
Yadav teaches wherein the predicted event is related to at least one of (a) a change in an Interior Gateway Protocol (IGP) configuration in the network and (b) one or more flapping links in the network ([Col 13, 55-59]: There may be a hardware failure where routing protocol daemon 40 continues to execute, such as an input clock or clock module failure. Such a failure may eventually cause router 20 to flap the impacted links, but flapping the links may cause network traffic loss (i.e. failure event relates to flapping link)).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang and Nguyen to incorporate the teachings of Yadav and the predicted event is related to one or more flapping links in the network. One of ordinary skilled in the art would have been motivated to combine the teachings in order to determine possible fault (Yadav, [Col 14, 2-3]).
Regarding Claim 13, they do not teach or further define over claim 2. Therefore, claim 13 is rejected for the same reason as set forth above in claim 2.
Claim 3, 14 and 17 rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang further in view of Kubota et al. (US 20230137995 A1, hereafter Kubota).
Regarding claim 3, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 1.
Jordan in view of Zhang however does not teach wherein the ML model implements an XGBoost model as the gradient-boosted classification model to perform the classification function and implements a Variational Auto Encoder (VAE) model to perform the encoding/decoding function.
Kubota teaches wherein the ML model implements an XGBoost model as the gradient-boosted classification model to perform the classification function ([55]: The learning unit 12 inputs the expanded data acquired by the acquisition unit 11 to the learning model 12a that performs predetermined learning to implement learning. The learning unit 12 can also apply, to the learning model 12a, any of a random forest using bagging and a decision tree, XGboost using boosting) and implements a Variational Auto Encoder (VAE) model to perform the encoding/decoding function ([56]: The learning unit 12 may also use a predetermined learning model using a neural network, which can be applied to a strong learner. A specific example of the predetermined learning model 12a may also be a VAE (Variational AutoEncoder)).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to incorporate the teachings of Kubota and the ML model implements an XGBoost model to perform the classification function and implements a Variational Auto Encoder (VAE) model to perform the encoding/decoding function. One of ordinary skilled in the art would have been motivated to combine the teachings in order to conduct filtering (Kubota, [53]).
Regarding Claim 14, they do not teach or further define over claim 3. Therefore, claim 14 are rejected for the same reason as set forth above in claim 3.
Regarding claim 17, Jordan, Zhang and Kubota teaches the method of claim 14.
Jordan further teaches wherein the steps further include allowing the ML model to leverage spatial-based metrics related to neighboring components arranged within the network ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106.), and
using the temporal-based correlations ([29]: The machine learning system may differ from the attribute tracking system in that the machine learning maintains and analyzes the value of all attributes received, rather than simply tracking a count or statistical representation. The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation. Each time that a specified attribute-value pairing is detected, the attribute and/or value may be stored. A threat vector may be assigned to an event record depending on the attributes and/or values stored for that event record and sent to step 106 (i.e. tracking the attributes and values is the temporal-based correlation)) and the spatial-based metrics to predict the event associated with the network ([29]: The machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system and the threat vector may be sent to step 106 (i.e. learning curve is a graph therefore it is a spatial-based metric)).
Claim 4 and 15 rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang to further in view of Liu et al. (US 20230409901 A1, hereinafter Liu)
Regarding claim 4, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 1.
Jordan in view of Zhang however does not teach wherein the encoding function compresses the numerical PM parameters into a latent representation, and the decoding function reconstructs the numerical PM parameters from the latent representation.
Liu teaches wherein the encoding function compresses the numerical PM parameters into a latent representation, and the decoding function reconstructs the numerical PM parameters from the latent representation ([77]: As shown in FIG. 5 , time series forecasting framework 500 receives time series dataset 502, e.g., observed time series dataset xt of a dynamical system. Using the time series forecasting framework 500, the prediction procedure disentangles the multivariate forecasting task into the following steps: (1) The encoder captures the nonlinear correlation among series and recovers the latent causal representation from the observed data; (2) Next-step prediction is generated via the auxiliary predictor in the latent space; (3) prediction results are transformed into observed space by the decoder.).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to incorporate the teachings of Liu and wherein the encoding function compresses the numerical PM parameters into a latent representation, and the decoding function reconstructs the numerical PM parameters from the latent representation. One of ordinary skilled in the art would have been motivated to combine the teachings in order forecasting using a neural network (Liu, [76]).
Regarding Claim 15, they do not teach or further define over claim 4. Therefore, claim 15 is rejected for the same reason as set forth above in claim 4.
Claim 5 and 16 rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang and Kubota further in view of Abdelaal et al. (US 20240070465 A1, hereafter Abdelaal) further in view of Burkhart et al. (US 20200320382 A1, hereinafter Burkhart).
Regarding claim 5, Jordan in view of Zhang and Kubota teaches the non-transitory computer-readable medium of claim 3.
Jordan in view of Zhang and Kubota however does not teach wherein the VAE model implements a loss function that includes a reconstruction error, a Kullback-Leibler divergence, and a binary cross entropy loss related to a loss of a classifier component associated with the XGBoost model.
Abdelaal teaches wherein the VAE model implements a loss function that includes a reconstruction error, a Kullback-Leibler divergence ([53]: A VAE is a neural network architecture that provides a probabilistic manner for describing a dataset in the latent space. It works with an encoder (sometimes referred to as the recognition model), a decoder (sometimes referred to as the generative model), and a loss function. The encoder and decoder are trained jointly such that the output minimizes reconstruction error and the KL divergence between the parametric posterior (i.e., distribution of the generated data) and the true posterior), and a classifier component associated with the XGBoost model ([74]: The techniques of certain example embodiments were analyzed with several datasets and several machine learning models, e.g., multi-layer perception (MLP), random forest, and XGBoost.).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang and Kubota to incorporate the teachings of Abdelaal and wherein the VAE model implements a loss function that includes a reconstruction error, a Kullback-Leibler divergence, and a classifier component associated with the XGBoost model. One of ordinary skilled in the art would have been motivated to combine the teachings in order for describing a dataset in the latent space (Abdelaal, [53]).
Jordan in view of Zhang, Kubota and Abdelaal however does not teach the VAE model implements a binary cross entropy loss related to a loss of a classifier component.
Burkhart teaches the VAE model implements a binary cross entropy loss related to a loss of a classifier component ([126]: At least one technique the inventors have discovered is that including at least one estimator from the variational autoencoder (VAE) estimators 222, combined with training the neural network 204 to minimize root mean square error or cross-entropy loss as discussed above, provides a recommendation performance that exceeds the performance provided by conventional techniques.).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang, Kubota and Abdelaal to incorporate the teachings of Burkhart and the VAE model implements a binary cross entropy loss related to a loss of a classifier component. One of ordinary skilled in the art would have been motivated to combine the teachings in order to exceed the performance provided by conventional techniques (Burkhart, [126]).
Regarding Claim 16, they do not teach or further define over claim 5. Therefore, claim 16 is rejected for the same reason as set forth above in claim 5.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang and Kubota further in view of Sadrieh et al. (US 11294756 B1, hereafter Sadrieh).
Regarding claim 6, Jordan in view of Zhang and Kubota teaches the non-transitory computer-readable medium of claim 3.
Jordan in view of Zhang and Kubota however does not teach wherein the VAE model uses a Long Short-Term Memory (LSTM) technique.
Sadrieh teaches wherein the VAE model uses a Long Short-Term Memory (LSTM) technique ([Col 3, 49-54]: Both the encoder 320 and decoder 350 of the VAE are implemented using a L-layer LSTM configuration).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang and Kubota to incorporate the teachings of Sadrieh and the VAE model uses a Long Short-Term Memory (LSTM) technique. One of ordinary skilled in the art would have been motivated to combine the teachings in order to provide probabilistic representation (Sadrieh, [Col 2, 58-59]).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang further in view of Ford et al. (US 2021/0351973 A1, hereafter Ford).
Regarding claim 7, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 1.
Jordan teaches wherein the temporal-based correlations are related to differences in the PM parameters between consecutive pairs of datapoints in the subset ([29]: The machine learning system may include a filter which normalizes incoming data to ensure that variation of the same value has the same representation. A threat vector may be assigned to an event record depending on the attributes and/or values stored for that event record and sent to step 106. For example, the machine learning system may assign a threat vector to an event record in step 242 if the event record is statistically outside of a learned curve established by the machine learning system (i.e. determining differences in values)).
Jordan in view of Zhang however does not teach each consecutive pair defining a first timestamped datapoint and its next adjacent time-ordered datapoint.
Ford teaches each consecutive pair defining a first timestamped datapoint and its next adjacent time-ordered datapoint (Fig. 22(2216) and [353]: At step 2212 the apparatus receives time series KPI data of KPIs at N pairs. In step 2214 a list L is generated that includes all permutation of the KPI's/alarms in pairs of two. In step 2216 a pair of KPI or alarm is selected from the list L. In step 2218 a Granger causality tests is performed. The Granger causality tests for K1 implies that K2 for all (ENB, cell) pairs. In step 2200 a determination is performed. When the pair does not pass, then in step 2222 K1 and K2 are deleted from the list. (i.e. the KPI data are time series data. Here, pair of KPI/alarm is selected form time series data for tests. Fig. 23 and [360] further shows the analysis over KPIs and alarms across an ensemble of eNBs, cell numbers, and time stamps.)).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to further incorporate the teachings of Ford and each consecutive pair defining a first timestamped datapoint and its next adjacent time-ordered datapoint. One of ordinary skilled in the art would have been motivated to combine the teachings in order to establish correlations across space (Ford, [360]).
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang further in view of Yadav et al. (US 9838317 B1, hereafter Yadav).
Regarding claim 9, Jordan in view of Zhang teaches the non-transitory computer-readable medium of claim 8.
Jordan in view of Zhang however does not teach wherein predicting the event associated with the network includes predicting Interior Gateway Protocol (IGP) configuration changes over a period of five subsequent days.
Yadav teaches wherein predicting the event associated with the network includes predicting Interior Gateway Protocol (IGP) configuration changes over a period of five subsequent days ([Col 7, 2-7]: Traffic impact prediction module 45 may be configured to periodically evaluate (e.g., once per 15 seconds, once per minute, once per hour, etc.) the operating characteristics of router 20 and, as part of this periodic evaluation, adjust the metrics for the impacted routes. [Col 8, 55-67]: RIB 38 may include information defining a topology of a network, such as network 4 of FIG. 1, learned by execution of Interior Gateway Protocol with Traffic Engineering extensions 42 (“IGP-TE 42”) by routing protocol daemon 40 (“illustrated as RP daemon 40”). IGP-TE 42 may represent OSPF-TE or IS-IS-TE.).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to incorporate the teachings of Yadav and predicting the event associated with the network includes predicting Interior Gateway Protocol (IGP) configuration changes over a period of five subsequent days. One of ordinary skilled in the art would have been motivated to combine the teachings in order to determine possible fault (Yadav, [Col 14, 2-3]).
Claim 4, 15 and 19 rejected under 35 U.S.C. 103 as being unpatentable over Jordan in view of Zhang to further in view of Sadrieh et al. (US 11294756 B1, hereafter Sadrieh) further in view of Naumov et al. (US 20220148730 A1, hereinafter Naumov).
Regarding claim 19, Jordan in view of Zhang teaches the system of claim 18.
Jordan in view of Zhang however does not teach wherein the ML model implements an XGBoost model as the gradient-boosted classification model to perform the classification function and implements a Variational Auto Encoder (VAE) model to perform the encoding/decoding function, wherein the VAE model is configured to encode the temporal-based correlations for mapping to a latent representation, and wherein the XGBoost model is configured to perform classification directly from the latent representation.
Kubota teaches wherein the ML model implements an XGBoost model as the gradient-boosted classification model to perform the classification function ([55]: The learning unit 12 inputs the expanded data acquired by the acquisition unit 11 to the learning model 12a that performs predetermined learning to implement learning. The learning unit 12 can also apply, to the learning model 12a, any of a random forest using bagging and a decision tree, XGboost using boosting) and implements a Variational Auto Encoder (VAE) model to perform the encoding/decoding function ([56]: The learning unit 12 may also use a predetermined learning model using a neural network, which can be applied to a strong learner. A specific example of the predetermined learning model 12a may also be a VAE (Variational AutoEncoder)).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang to incorporate the teachings of Kubota and the ML model implements an XGBoost model to perform the classification function and implements a Variational Auto Encoder (VAE) model to perform the encoding/decoding function. One of ordinary skilled in the art would have been motivated to combine the teachings in order to conduct filtering (Kubota, [53]).
Jordan in view of Zhang and Kubota however does not teach wherein the VAE model is configured to encode the temporal-based correlations for mapping to a latent representation, and wherein the XGBoost model is configured to perform classification directly from the latent representation.
Sadrieh teaches wherein the VAE model is configured to encode the temporal-based correlations for mapping to a latent representation ([Col 2, 58-65]: The VAE provides a probabilistic representation of latent space by using the encoder to describe a probability distribution for each latent attribute. A VAE layer can be trained based on a feature-set representing BGP behavior in the network. The VAE's reconstruction probability for each feature is input into a Random Isolation Forest (RIF) 230 layer in order to calculate the final anomaly score 240 for each observation.).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang and Kubota to incorporate the teachings of Sadrieh and wherein the VAE model is configured to encode the temporal-based correlations for mapping to a latent representation. One of ordinary skilled in the art would have been motivated to combine the teachings in order to provide probabilistic representation (Sadrieh, [Col 2, 58-59]).
Jordan in view of Zhang, Kubota and Sadrieh however does not teach wherein the XGBoost model is configured to reconstruct the temporal-based correlations from the latent representation.
Naumov teaches wherein the XGBoost model is configured to reconstruct the temporal-based correlations from the latent representation ([198]: The XGBoost algorithm may comprise an implementation of gradient boosted decision trees that is designed for speed and/or performance. The XGBoost algorithm may automatically handle missing data values).
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Jordan in view of Zhang, Kubota and Sadrieh to incorporate the teachings of Naumov and the VAE model is configured to encode the temporal-based correlations for mapping to a latent representation. One of ordinary skilled in the art would have been motivated to combine the teachings in order to improve speed and performance (Naumov, [198]).
Additional References
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
a. Vasseur et al., US 20230308359 A1: CAPTURING NETWORK DYNAMICS AND FORECASTING ACROSS MULTIPLE TIMESCALES.
b. Patodia et al., US 20230135329 A1: COMPUTER SOFTWARE ARCHITECTURE FOR EXECUTION EFFICIENCY.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SUJANA KHAKURAL whose telephone number is (571)272-3704. The examiner can normally be reached on M-F: 7:30AM - 5:30PM.
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, Kamal B Divecha can be reached on 571-272-5863. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see https://ppair-my.uspto.gov/pair/PrivatePair. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/SUJANA KHAKURAL/Examiner, Art Unit 2453
/KAMAL B DIVECHA/Supervisory Patent Examiner, Art Unit 2453
Prosecution Insights