CTNF 18/240,183 CTNF 98836 Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Claim Rejections - 35 USC § 103 07-20-aia AIA 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 1–5, 7, 10, 12–13, 18, and 20 rejected under 35 U.S.C. 102(a) (1) over Hong et al., “Curved Geometric Networks for Visual Anomaly Recognition,” 2022, arXiv:2208.01188v1 (hereafter Hong ) in view of Sha et al., US Pre-Grant Publication No. 2023/0348197 (hereafter Sha ). Regarding claim 1 and analogous claims 18 and 20: Hong teaches: (bold only) “ An anomaly detection system for detecting an anomaly of an operation of a machine based on a signal indicative of the operation of the machine performing a task, comprising: at least one processor; and memory having instructions stored thereon that, when executed by the at least one processor, cause the anomaly detection system to ”: Hong , section 1, paragraph 1, “In this paper, we aim to leverage the curved geometry for learning embeddings, which in return allows us to analyze and identify anomalies [ An anomaly detection system ] or out-of-distribution (OOD) objects from normal or in-distribution (ID) input data. Non-flat geometry has gained an increasing amount of interest in various machine learning approaches [ comprising: at least one processor; and memory having instructions stored thereon ], due to its intriguing properties in encoding the hidden structural information of the data [1]–[6]. For example, hyperbolic spaces, featured with a constant negative curvature, are shown to be rich in encoding the underlying hierarchical structure in the data.” “collect hyperbolic embeddings of the signal indicative of the operation of the machine, wherein the hyperbolic embeddings lie in a hyperbolic space”: Hong , section IV, paragraph 2, “The pipeline of the baseline is illustrated in Fig. 3 (a) for the training phase and 4 (a) for the inference phase. Specifically, a feature extractor first maps the input to a feature embedding eE, lying in a Euclidean space”; Hong , section IV, paragraph 3, “A curvature-aware geometric model indicates a model where its classifier operates in a curved space MG and we term its classifier geometric classifier. Two curvature-aware geometric models are presented in this section: Geometric-in-One (GiO) and Geometric-in-Two (GiT), as shown in Fig. 3 (b) and (c). Compared to the baseline model, the curvature-aware geometric model has two geometric layers, namely, geometric transformation and geometric MLP. The geometric transformation is to transform the Euclidean embedding eE computed by a feature extractor to the geometric embedding eG [ collect hyperbolic embeddings of the signal indicative of the operation of the machine, wherein the hyperbolic embeddings lie in a hyperbolic space ]. See Eq. (2) for the spherical geometry MS and Eq. (4) for the hyperbolic geometry [ hyperbolic embeddings ] MH.” “perform the detection of the anomaly of the operation of the machine based on the hyperbolic embeddings to determine an anomaly score; and render the anomaly score ”: Hong , section IV.B, paragraph 1, “In the GiT model, in parallel with a branch of the Euclidean classifier, the other branch learns the feature embedding in the curved space, and a following geometric MLP is used as a class predictor. The embedding in the curved space MG is achieved by transforming eE to eG via a GT function, as shown in Fig. 3 (c). In such a pipeline, the geometry-aware score zEG is defined as the discrepancy between distributions of eE and eG, measured via the Kullback-Leibler (KL) divergence”; Hong , section IV.B, paragraph 2, “Our experiments show that zEG is able to provide reliable discrimination information for anomaly identification. We find that, similar to our GiO models, the anomaly score of GiT models should be AS = 1 – tanh(zEG) (See Fig. 4 (c)). Tab. I lists the networks and the geometry score in the GiT model [ perform the detection of the anomaly of the operation of the machine based on the hyperbolic embeddings to determine an anomaly score; and render the anomaly score ].” Hong does not explicitly teach (bold only) “An anomaly detection system for detecting an anomaly of an operation of a machine based on a signal indicative of the operation of the machine performing a task , comprising: at least one processor; and memory having instructions stored thereon that, when executed by the at least one processor, cause the anomaly detection system to.” Sha teaches (bold only) “An anomaly detection system for detecting an anomaly of an operation of a machine based on a signal indicative of the operation of the machine performing a task , comprising: at least one processor; and memory having instructions stored thereon that, when executed by the at least one processor, cause the anomaly detection system to”: Sha , paragraph 0004, “To solve the above-mentioned technical problem, the technical solution adopted by the invention is a sound based roller fault detecting method by using double-projection neighborhoods preserving embedding comprising the following steps.” Sha and Hong are analogous arts as they are both related to anomaly detection methods. It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have combined the application of anomaly detection to machine processes of Sha with the teachings of Hong to arrive at the present invention, in order to improve fault detection in mechanical systems, as stated in Sha , paragraph 0002, “At present, the most common fault detecting method for conveyor belt rollers adopts manual detection. However, actual requirements cannot be met by simply relying on traditional manual detection methods. Moreover, traditional manual detection has a huge risk for workers, and has a low efficiency and a low precision, so that the production requirements of modem mines cannot be met. Therefore, there is a need for an intelligent roller fault detecting method.” Regarding claim 2: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong further teaches: “ wherein the processor is further configured to process, measurements of the signal or features extracted from the measurements, with a neural network, to produce Euclidean embedding of the signal in Euclidean space”: Hong , section IV, paragraph 2, “The pipeline of the baseline is illustrated in Fig. 3 (a) for the training phase and 4 (a) for the inference phase. Specifically, a feature extractor first maps the input to a feature embedding eE, lying in a Euclidean space [ process, measurements of the signal or features extracted from the measurements, with a neural network, to produce Euclidean embedding of the signal in Euclidean space ].” “and project the Euclidean embedding into the hyperbolic space to produce the hyperbolic embeddings”: Hong , section IV.B, paragraph 1, “In the GiT model, in parallel with a branch of the Euclidean classifier, the other branch learns the feature embedding in the curved space, and a following geometric MLP is used as a class predictor. The embedding in the curved space MG is achieved by transforming eE to eG via a GT function [ project the Euclidean embedding into the hyperbolic space to produce the hyperbolic embeddings ], as shown in Fig. 3 (c).” Regarding claim 3: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong further teaches: “ wherein the processor is further configured to process the signal with an embedding neural network to produce the hyperbolic embeddings of the signal ”: Hong , section IV, paragraph 2, “The pipeline of the baseline is illustrated in Fig. 3 (a) for the training phase and 4 (a) for the inference phase. Specifically, a feature extractor first maps the input to a feature embedding eE, lying in a Euclidean space”; Hong , section IV.B, paragraph 1, “In the GiT model, in parallel with a branch of the Euclidean classifier, the other branch learns the feature embedding in the curved space, and a following geometric MLP is used as a class predictor. The embedding in the curved space MG is achieved by transforming eE to eG via a GT function [ process the signal with an embedding neural network to produce the hyperbolic embeddings of the signa ], as shown in Fig. 3 (c).” “and process the hyperbolic embeddings with a classifying neural network to produce at least a portion of the anomaly score”: Hong , section IV, paragraph 3, “As shown in Fig. 3, we learn geometric classifiers where embeddings extracted from a feature extractor are manipulated in the curved spaces [ process the hyperbolic embeddings with a classifying neural network ]. As explained in Fig. 4 (b) and (c), in contrast to the baseline model, during the inference phase, the anomaly score in our geometric model is defined as AS = 1 - tanh(z) [ to produce at least a portion of the anomaly score ]. The geometric score z is a framework-dependent value whose details will be introduced below.” Regarding claim 4: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 3 .” Hong further teaches “ wherein the classifying neural network is trained with training data generated from non-anomalous operations of the machine, wherein the non- anomalous operations of the machine are defined relative to standard operating performance data of the machine ”: Hong , section IV, paragraph 4, “Visual anomaly recognition aims to identify abnormal (or OOD) samples from normal (or ID) samples. During the training process, as illustrated in Fig. 3, only normal or ID data can be accessed [ trained with training data generated from non-anomalous operations of the machine, wherein the non- anomalous operations of the machine are defined relative to standard operating performance data of the machine ]. For the evaluation stage, as shown in Fig. 4, both normal (or ID) and anomalous (or OOD) inputs to be recognized, exist.” Regarding claim 5: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 3 .” Hong further teaches “ wherein the embedding neural network is jointly trained with the classifying neural network such that weights of the embedding and classifying neural networks are interdependent on each other ”: Hong , Fig. 3, PNG media_image1.png 351 994 media_image1.png Greyscale [ showing in models (b) and (c), a combination of feature extractor and geometric transform (GT) for producing hyperbolic embeddings, trained along with classifiers formed by multi-layer perceptron (MLP) and softmax components, hence , the embedding neural network is jointly trained with the classifying neural network such that weights of the embedding and classifying neural networks are interdependent on each other]. Regarding claim 7: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong further teaches: “ wherein the hyperbolic space is approximated by projection on one of a Poincare ball or a Poincare disk ”: Hong , section III.C, paragraph 1, “In contrast to the n-sphere S n-1 k , the hyperbolic space is a curved space with a constant negative curvature (i.e., K < 0). In this paper, we employ the Poincare ball [1], [4] to model and work with the hyperbolic space [ the hyperbolic space is approximated by projection on one of a Poincare ball ].” “and the anomaly score is a function of a distance between a hyperbolic embedding of the hyperbolic embeddings and an origin of one of the Poincare ball or the Poincare disk ”: Hong , section IV.A, paragraph 2, “For the hyperbolic space MH, the geometric score is defined as zH(eH) = d Geo (eH, 0H), where d Geo (eH; 0H) is also known as the geodesic distance (See Fig. 6) between the point eH and the origin 0H, for eH, 0H ∈ H n k ”; Hong , section IV.A, paragraph 3, “Having the geometry score zG at our disposal, the anomaly score is defined as AS = 1 - tanh(zG) [ anomaly score is a function of a distance between a hyperbolic embedding of the hyperbolic embeddings and an origin of one of the Poincare ball or the Poincare disk ].” Regarding claim 10: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong further teaches (bold only) “wherein the processor is further configured to: split the signal into a plurality of segments; generate, for each segment of the plurality of segments, a hyperbolic embedding of the hyperbolic embeddings ”: Hong , section IV, paragraph 3, “A curvature-aware geometric model indicates a model where its classifier operates in a curved space MG and we term its classifier geometric classifier. Two curvature-aware geometric models are presented in this section: Geometric-in-One (GiO) and Geometric-in-Two (GiT), as shown in Fig. 3 (b) and (c). Compared to the baseline model, the curvature-aware geometric model has two geometric layers, namely, geometric transformation and geometric MLP. The geometric transformation is to transform the Euclidean embedding eE computed by a feature extractor to the geometric embedding eG [ hyperbolic embedding ]. See Eq. (2) for the spherical geometry MS and Eq. (4) for the hyperbolic geometry [ hyperbolic embedding ] MH.” Sha further teaches (bold only) “ wherein the processor is further configured to: split the signal into a plurality of segments; generate, for each segment of the plurality of segments, a hyperbolic embedding of the hyperbolic embeddings ”: Sha , paragraphs 0005-0008, “Step 1: acquiring operation sound data of a normal roller; Step 2: pre-processing the sound data through sliding windows to obtain n sound data samples as training samples [ split the signal into a plurality of segments ]; Step 3: performing a wavelet transform energy feature extraction on the sound data samples to obtain primary feature data; Step 4: obtaining an optimal projection matrix W of the primary feature data by using a double-projection neighborhoods preserving embedding feature extraction method [ generate, for each segment of the plurality of segments, a … embedding of the … embeddings ].” Sha and Hong are combinable for the rationale given under claim 1. Regarding claim 12: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Sha further teaches “ wherein the signal includes one or a combination of an acoustic signal and a video signal ”: Sha , paragraphs 0005-0008, “Step 1: acquiring operation sound data of a normal roller [ the signal includes one or a combination of an acoustic signal and a video signal ]; Step 2: pre-processing the sound data through sliding windows to obtain n sound data samples as training samples; Step 3: performing a wavelet transform energy feature extraction on the sound data samples to obtain primary feature data; Step 4: obtaining an optimal projection matrix W of the primary feature data by using a double-projection neighborhoods preserving embedding feature extraction method.” Sha and Hong are combinable for the rationale given under claim 1. Regarding claim 13: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Sha further teaches “ wherein the signal includes measurements of vibration of the machine caused by the operation of the machine ”: Sha , paragraphs 0005-0008, “Step 1: acquiring operation sound data of a normal roller [ wherein the signal includes measurements of vibration of the machine caused by the operation of the machine ]; Step 2: pre-processing the sound data through sliding windows to obtain n sound data samples as training samples; Step 3: performing a wavelet transform energy feature extraction on the sound data samples to obtain primary feature data; Step 4: obtaining an optimal projection matrix W of the primary feature data by using a double-projection neighborhoods preserving embedding feature extraction method.” Sha and Hong are combinable for the rationale given under claim 1. Claims 6, 11, and 19 rejected under 35 U.S.C. 103 over Hong as modified by Sha in view of Marwah et al., US Patent No. 11,150,976 (hereafter Marwah ). Regarding claim 6: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 3 .” Hong further teaches: (bold only) “ wherein the anomaly score is a function of a combination of a distance between the hyperbolic embeddings and an origin of the hyperbolic space and a probability of correct classification returned by the classifying neural network”: Hong , section IV.A, paragraph 2, “For the hyperbolic space MH, the geometric score is defined as zH(eH) = d Geo (eH, 0H), where d Geo (eH; 0H) is also known as the geodesic distance (See Fig. 6) between the point eH and the origin 0H, for eH, 0H ∈ H n k ”; Hong , section IV.A, paragraph 3, “Having the geometry score zG at our disposal, the anomaly score is defined as AS = 1 - tanh(zG) [ the anomaly score is a function of … a distance between the hyperbolic embeddings and an origin of the hyperbolic space ].” (bold only) “ wherein the anomaly score is a function of a combination of a distance between the hyperbolic embeddings and an origin of the hyperbolic space and a probability of correct classification returned by the classifying neural network ”: Hong , section II.A, paragraph 2, “It is well known that the confidence from the softmax in a classifier helps to detect OOD samples from ID samples since ID samples are more likely to have a greater maximum softmax confidence compared to ODD samples [ wherein the anomaly score is a function of a … probability of correct classification returned by the classifying neural network ].” Hong as modified by Sha does not explicitly teach (bold only) “wherein the anomaly score is a function of a combination of a distance between the hyperbolic embeddings and an origin of the hyperbolic space and a probability of correct classification returned by the classifying neural network.” Marwah teaches (bold only) “wherein the anomaly score is a function of a combination of a distance between the hyperbolic embeddings and an origin of the hyperbolic space and a probability of correct classification returned by the classifying neural network”: Marwah , col. 2, lines 20–32, “For a given type of target system, there may be a large number of different anomaly detectors that can be employed. Different anomaly detectors may detect the same type of anomaly, but using different techniques or algorithms. Different anomaly detectors may detect different types of anomalies at the same target system. Using multiple anomaly detectors can ensure that if an anomaly at the target system escapes detection by one anomaly detector, another anomaly detector may still detect the anomaly. The anomaly scores output by the anomaly detectors may be aggregated or otherwise considered en masse in this respect [ a combination ], particularly in the case in which a large number of anomaly detectors are being employed.” Marwah and Hong are analogous arts as they are both related to anomaly detection. It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have employed the aggregate anomaly score concept from Marwah with the specific anomaly detection methods taught by Hong to arrive at the present invention, in order to improve anomaly detection, as stated in Marwah , col. 2, lines 22–32, “Different anomaly detectors may detect the same type of anomaly, but using different techniques or algorithms. Different anomaly detectors may detect different types of anomalies at the same target system. Using multiple anomaly detectors can ensure that if an anomaly at the target system escapes detection by one anomaly detector, another anomaly detector may still detect the anomaly. The anomaly scores output by the anomaly detectors may be aggregated or otherwise considered en masse in this respect, particularly in the case in which a large number of anomaly detectors are being employed.” Regarding claim 11 and analogous claim 19: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 10 .” Hong further teaches (bold only) “ wherein to determine the anomaly score the processor is further configured to: compute, for each hyperbolic embedding of the hyperbolic embeddings , a probability of a corresponding segment belonging to a certain attribute type ”: Hong , section IV, paragraph 2, “The pipeline of the baseline is illustrated in Fig. 3 (a) for the training phase and 4 (a) for the inference phase. Specifically, a feature extractor first maps the input to a feature embedding eE, lying in a Euclidean space”; Hong , section IV, paragraph 3, “A curvature-aware geometric model indicates a model where its classifier operates in a curved space MG and we term its classifier geometric classifier. Two curvature-aware geometric models are presented in this section: Geometric-in-One (GiO) and Geometric-in-Two (GiT), as shown in Fig. 3 (b) and (c). Compared to the baseline model, the curvature-aware geometric model has two geometric layers, namely, geometric transformation and geometric MLP. The geometric transformation is to transform the Euclidean embedding eE computed by a feature extractor to the geometric embedding eG [ hyperbolic embedding of the hyperbolic embeddings ]. See Eq. (2) for the spherical geometry MS and Eq. (4) for the hyperbolic geometry [ hyperbolic embeddings ] MH.” Hong as modified by Sha does not explicitly teach: (bold only) “wherein to determine the anomaly score the processor is further configured to: compute, for each hyperbolic embedding of the hyperbolic embeddings, a probability of a corresponding segment belonging to a certain attribute type” “aggregate the computed probabilities; and determine the anomaly score, based on the aggregated probabilities ” Marwah teaches: (bold only) “wherein to determine the anomaly score the processor is further configured to: compute, for each hyperbolic embedding of the hyperbolic embeddings, a probability of a corresponding segment belonging to a certain attribute type”: Marwah , col. 2, lines 1–16, “The anomaly detector may process such received information to generate an anomaly score indicating a likelihood that malicious infiltration has occurred. If the likelihood is greater than a threshold, for instance, then the target system may be reconfigured as a remedial action to ameliorate the detected anomaly. As another example, as to the predictive detection of hardware and/or software component failure of a target computing system, an anomaly detector may output an anomaly score indicative of whether such a component is likely to fail soon [ a probability of a corresponding segment belonging to a certain attribute type ]. The anomaly detector may receive as input logs regarding the component in question, as well as other components of the target system. The anomaly detector may process the received information to generate an anomaly score indicating a likelihood that the component is on the verge of failure.” “ aggregate the computed probabilities; and determine the anomaly score, based on the aggregated probabilities ”: Marwah , col. 2, lines 20–36, “For a given type of target system, there may be a large number of different anomaly detectors that can be employed. Different anomaly detectors may detect the same type of anomaly, but using different techniques or algorithms. Different anomaly detectors may detect different types of anomalies at the same target system. Using multiple anomaly detectors can ensure that if an anomaly at the target system escapes detection by one anomaly detector, another anomaly detector may still detect the anomaly. The anomaly scores output by the anomaly detectors may be aggregated or otherwise considered en masse in this respect, particularly in the case in which a large number of anomaly detectors are being employed. For example, the anomaly scores may be aggregated to generate an overall anomaly score [ aggregate the computed probabilities; and determine the anomaly score, based on the aggregated probabilities ]], which can then itself be compared to a threshold to detect an anomaly at a target system.” Marwah and Hong are analogous arts as they are both related to anomaly detection. It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have employed the aggregate anomaly score concept from Marwah with the specific anomaly detection methods taught by Hong to arrive at the present invention, in order to improve anomaly detection, as stated in Marwah , col. 2, lines 22–32, “Different anomaly detectors may detect the same type of anomaly, but using different techniques or algorithms. Different anomaly detectors may detect different types of anomalies at the same target system. Using multiple anomaly detectors can ensure that if an anomaly at the target system escapes detection by one anomaly detector, another anomaly detector may still detect the anomaly. The anomaly scores output by the anomaly detectors may be aggregated or otherwise considered en masse in this respect, particularly in the case in which a large number of anomaly detectors are being employed.” Claims 8–9 and 14 rejected under 35 U.S.C. 103 over Hong as modified by Sha in view of Hendler et al., US Pre-Grant Publication No. 2008/0010531 (hereafter Hendler ). Regarding claim 8: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong further teaches (bold only) “ wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector indicative of a unique attribute type associated with the machine”: Hong , section III.A, paragraph 1, “We use K to denote the curvature of a manifold. In general, a vectorized representation or an embedding can be embedded in three types of manifolds: the Euclidean space ME, the spherical space MS and the hyperbolic space MH [ wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector ], corresponding to K = 0, K > 0, and K < 0, respectively.” Hong as modified by Sha does not explicitly teach (bold only) “wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector indicative of a unique attribute type associated with the machine .” Hendler teaches (bold only) “wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector indicative of a unique attribute type associated with the machine ”: Hendler , paragraph 0013, “In some embodiments, the comparison value is determined according to Pearson's correlation equation. The fault vector can be associated with the first set of fault vectors when the comparison value exceeds a predetermined value or threshold. The predetermined value or threshold can be determined by a user. In some embodiments, the first set of fault vectors is associated with an attribute indicative of a type of fault vector that is associated with the first set of fault vectors [ a vector indicative of a unique attribute type associated with the machine ]. Associating the fault vector with the first set of fault vectors can include modifying the attribute.” Hendler and Hong are analogous arts as they are both related to anomaly detection methods It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have combined the fault vector attributes of Hendler with the teachings of Hong to arrive at the present invention, in order to better mitigate anomalous situations, as stated in Hendler , paragraph 0040, “Knowledge of the attributes associated with the set 150 allows the operator to quickly assess and implement corrective actions to prevent further faulty or defective wafers.” Regarding claim 9: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong further teaches (bold only) “ wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector indicative of a unique attribute from the set of attributes ”: Hong , section III.A, paragraph 1, “We use K to denote the curvature of a manifold. In general, a vectorized representation or an embedding can be embedded in three types of manifolds: the Euclidean space ME, the spherical space MS and the hyperbolic space MH [ wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector ], corresponding to K = 0, K > 0, and K < 0, respectively.” Sha further teaches “ wherein the signal indicative of the operation of the machine is an audio signal produced in an operating environment of the machine ” and (bold only) “wherein a source of the audio signal and the operating environment are characterized by a set of attributes ”: Sha , paragraphs 0005-0008, “Step 1: acquiring operation sound data of a normal roller [ wherein the signal indicative of the operation of the machine is an audio signal produced in an operating environment of the machine ][ source of the audio signal and the operating environment ]; Step 2: pre-processing the sound data through sliding windows to obtain n sound data samples as training samples; Step 3: performing a wavelet transform energy feature extraction on the sound data samples to obtain primary feature data; Step 4: obtaining an optimal projection matrix W of the primary feature data by using a double-projection neighborhoods preserving embedding feature extraction method.” Sha and Hong are combinable for the rationale given under claim 1. Hong as modified by Sha do not explicitly teach: (bold only) “ wherein a source of the audio signal and the operating environment are characterized by a set of attributes ” (bold only) “wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector indicative of a unique attribute from the set of attributes” Hendler teaches (bold only) “ wherein a source of the audio signal and the operating environment are characterized by a set of attributes ” and “wherein each hyperbolic embedding of the hyperbolic embeddings corresponds to a vector indicative of a unique attribute from the set of attributes”: Hendler , paragraph 0013, “In some embodiments, the comparison value is determined according to Pearson's correlation equation. The fault vector can be associated with the first set of fault vectors when the comparison value exceeds a predetermined value or threshold. The predetermined value or threshold can be determined by a user. In some embodiments, the first set of fault vectors is associated with an attribute indicative of a type of fault vector that is associated with the first set of fault vectors [ a vector indicative of a unique attribute type associated with the machine ][ wherein a source … are characterized by a set of attributes ]. Associating the fault vector with the first set of fault vectors can include modifying the attribute.” Hendler and Hong are analogous arts as they are both related to anomaly detection methods It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have combined the fault vector attributes of Hendler with the teachings of Hong to arrive at the present invention, in order to better mitigate anomalous situations, as stated in Hendler , paragraph 0040, “Knowledge of the attributes associated with the set 150 allows the operator to quickly assess and implement corrective actions to prevent further faulty or defective wafers.” Regarding claim 14: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong as modified by Sha does not explicitly teach “ wherein the signal includes measurements of one or a combination of a voltage and a current controlling the machine, and a torque produced by the machine .” Hendler teaches “ wherein the signal includes measurements of one or a combination of a voltage and a current controlling the machine, and a torque produced by the machine ”: Hendler , paragraph 0003, “A typical process tool used in current semiconductor manufacturing can be described by a set of several thousand process variables. The variables are generally related to physical parameters of the manufacturing process and/or tools used in the manufacturing process. In some cases, of these several thousand variables, several hundred variables will be dynamic ( e.g., changing in time during the manufacturing process or between manufacturing processes). The dynamic variables, for example, gas flow, gas pressure, delivered power [ torque ], current [ current ], voltage [ voltage ], and temperature change based on, for example, a specific processing recipe, the particular step in the overall sequence of processing steps, or errors and faults that occur during the manufacturing process.” Hendler and Hong are analogous arts as they are both related to anomaly detection methods It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have combined the monitoring of power, current, and voltage signals of Hendler with the teachings of Hong to arrive at the present invention, in order to improve anomaly detection, as stated in Hendler , paragraph 0007, “In this way, for real-time wafer processing, variables associated with a particular wafer processing recipe can be monitored. If a particular set of values of a set of measured variables resembles a previously-measured set of variables stored in the database, the type of fault responsible for the measured variables can be determined. In general, similar faults will be associated with similar patterns of values for the variables (e.g., the processing parameters).” Claims 15–17 rejected under 35 U.S.C. 103 over Hong as modified by Sha in view of Hirano et al., US Pre-Grant Publication No. 2021/0349997 (hereafter Hirano ). Regarding claim 15: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong as modified by Sha does not explicitly teach “ wherein the control system is configured to control the operation of the machine based on the anomaly score rendered by the anomaly detection system .” Hirano teaches “ wherein the control system is configured to control the operation of the machine based on the anomaly score rendered by the anomaly detection system ”: Hirano , paragraph 0258, “As illustrated in FIG. 23 and FIG. 24, anomaly countermeasure notifier 109 may request a vehicle of the same vehicle type as the vehicle determined to be an anomalous vehicle or a vehicle located in the anomalous area where the vehicle determined to be an anomalous vehicle is located to be subjected to any one or more of the shutdown of a network interface, the limiting of the address being accessed and the address being accessed from, the limiting of the number of network devices to be connected, the alerting to the driver, the limiting of the network connection, the limiting of the vehicle control function, the stopping of the vehicle control function from starting, an increase in the frequency at which the vehicle log is transmitted, an increase in the number of types of the vehicle log, and the notification to the driver based on the value of the anomaly score ( e.g., the value of the anomaly score of the vehicle determined to be an anomalous vehicle) or the type of the suspicious behavior (e.g., the network analysis or the system analysis) [ wherein the control system is configured to control the operation of the machine based on the anomaly score rendered by the anomaly detection system ]. The one or more requests are made to the anomalous vehicle, the vehicle of the same vehicle type as the vehicle determined to the anomalous vehicle, and vehicles other than the anomalous vehicle that are located in the anomalous area.” Hirano and Hong are analogous arts as they are both related to anomaly detection methods. It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have combined the controls reactive to anomalies of Hirano with the teachings of Hong to arrive at the present invention, in order to mitigate problems signaled by anomalies, as stated in Hirano , paragraph 0043, “Accordingly, the present inventors have diligently contemplated anomalous vehicle detection servers and so on that can capture the activities of an attacker in the stage of investigating an in-vehicle network and conceived of an anomalous vehicle detection server and so on described below. For example, the present inventors have found that the security of an in-vehicle network can be increased effectively by monitoring vehicle logs of a plurality of vehicles on a server, capturing a vehicle behavior that is different from a predetermined behavior ( e.g., a normal behavior) and occurs due to the reverse engineering performed by an attacker as a suspicious behavior, calculating an anomaly score indicating how likely it is that reverse engineering is being performed on a vehicle, detecting a vehicle having an anomaly score that is greater than a statistical value ( e.g., a mean value) of anomaly scores of vehicles of an identical vehicle type, and taking a countermeasure against an anomaly based on the value of the anomaly score and an anomaly category.” Regarding claim 16: Hong as modified by Sha and Hirano teaches “ [t]he anomaly detection system of claim 15 .” Hirano further teaches “ wherein the control system is configured to change a mode of the operation of the machine based on the anomaly score”: Hirano , paragraph 0258, “As illustrated in FIG. 23 and FIG. 24, anomaly countermeasure notifier 109 may request a vehicle of the same vehicle type as the vehicle determined to be an anomalous vehicle or a vehicle located in the anomalous area where the vehicle determined to be an anomalous vehicle is located to be subjected to any one or more of the shutdown of a network interface, the limiting of the address being accessed and the address being accessed from, the limiting of the number of network devices to be connected, the alerting to the driver, the limiting of the network connection, the limiting of the vehicle control function, the stopping of the vehicle control function from starting [ wherein the control system is configured to change a mode of the operation of the machine based on the anomaly score ], an increase in the frequency at which the vehicle log is transmitted, an increase in the number of types of the vehicle log, and the notification to the driver based on the value of the anomaly score ( e.g., the value of the anomaly score of the vehicle determined to be an anomalous vehicle) or the type of the suspicious behavior (e.g., the network analysis or the system analysis). The one or more requests are made to the anomalous vehicle, the vehicle of the same vehicle type as the vehicle determined to the anomalous vehicle, and vehicles other than the anomalous vehicle that are located in the anomalous area.” Hirano and Hong are combinable for the rationale given under claim 15. Regarding claim 17: Hong as modified by Sha teaches “ [t]he anomaly detection system of claim 1 .” Hong as modified by Sha does not explicitly teach “ wherein the control system is configured to control the operation of the machine based on the anomaly score rendered by the anomaly detection system .” Hirano teaches “ further comprising a display device configured to display the rendered anomaly score ”: Hirano , paragraph 0070, “For example, the anomalous vehicle detection server may further include an anomaly display that displays, in a list form, anomalous vehicles in a descending order of the anomaly score, and the anomalous vehicles are each the anomalous vehicle [ a display device configured to display the rendered anomaly score ].” Hirano and Hong are analogous arts as they are both related to anomaly detection methods. It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have combined the anomaly display of Hirano with the teachings of Hong to arrive at the present invention, in order to allow users to mitigate problems signaled by anomalies, as stated in Hirano , paragraph 0071, “This configuration allows an operator analyzing an anomalous vehicle by checking the content displayed on the anomaly display to analyze a more suspicious vehicle preferentially. Therefore, the operator can carry out the analysis work effectively.” Conclusion 07-96 AIA The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Zambon et al., “Anomaly and Change Detection in Graph Streams through Constant-Curvature Manifold Embeddings,” 2018, arXiv:1805.01360v1, discloses methods for anomaly detection using hyperbolic embeddings. Any inquiry concerning this communication or earlier communications from the examiner should be directed to VINCENT SPRAUL whose telephone number is (703) 756-1511. The examiner can normally be reached M-F 9:00 am - 5:00 pm. 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, MICHAEL HUNTLEY can be reached at (303) 297-4307. 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. /VAS/ Examiner, Art Unit 2129 /MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129 Application/Control Number: 18/240,183 Page 2 Art Unit: 2129 Application/Control Number: 18/240,183 Page 3 Art Unit: 2129 Application/Control Number: 18/240,183 Page 4 Art Unit: 2129 Application/Control Number: 18/240,183 Page 5 Art Unit: 2129 Application/Control Number: 18/240,183 Page 6 Art Unit: 2129 Application/Control Number: 18/240,183 Page 7 Art Unit: 2129 Application/Control Number: 18/240,183 Page 8 Art Unit: 2129 Application/Control Number: 18/240,183 Page 9 Art Unit: 2129 Application/Control Number: 18/240,183 Page 10 Art Unit: 2129 Application/Control Number: 18/240,183 Page 11 Art Unit: 2129 Application/Control Number: 18/240,183 Page 12 Art Unit: 2129 Application/Control Number: 18/240,183 Page 13 Art Unit: 2129 Application/Control Number: 18/240,183 Page 14 Art Unit: 2129 Application/Control Number: 18/240,183 Page 15 Art Unit: 2129 Application/Control Number: 18/240,183 Page 16 Art Unit: 2129 Application/Control Number: 18/240,183 Page 17 Art Unit: 2129 Application/Control Number: 18/240,183 Page 18 Art Unit: 2129 Application/Control Number: 18/240,183 Page 19 Art Unit: 2129 Application/Control Number: 18/240,183 Page 20 Art Unit: 2129 Application/Control Number: 18/240,183 Page 21 Art Unit: 2129 Application/Control Number: 18/240,183 Page 22 Art Unit: 2129 Application/Control Number: 18/240,183 Page 23 Art Unit: 2129
Prosecution Insights