DETAILED ACTION
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 .
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 11/11/2025 has been entered.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 09/09/2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Arguments
Applicant’s arguments with respect to claims 1-3,5-6,8-10,12-13 and 15-17, and 19-20 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.
Allowable Subject Matter
Claims 4, 11, and 18 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
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.
Claims 1, 8, and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Schimert et al. (US-20170193372-A1) in view of Zhu et al. (US-20240143699-A1), Jiang et al. (“Similarity learning for cover song identification using cross-similarity matrices of multi-level deep sequences”), Chang et al. (“Audio cover song identification using convolutional neural network”), and Malhotra et al. (US-20210103812-A1).
Regarding Claim 1,
Schimert (US 20170193372 A1) teaches a method, comprising:
obtaining, with at least one processor, from at least one database (para [0039] And para [0071] Time series data captured by sensors can be stored in a database.), a plurality of known time series (para [0054]-[0057] And para [0060] “Segment generator 204 may transform time series data 210 into plurality of segments 216 for use in evaluating health 211 of component 214. In particular, time series data 210 may be transformed into plurality of segments 216 based on at least one selected state 220 for aircraft 202 that is of interest.” Segments (i.e. time series data).);
for each known time series of the plurality of known time series:
computing, with the at least one processor, a pairwise distance matrix between that known time series and each learned template of a plurality of learned templates to generate a plurality of pairwise distance matrices (para [0071] segment 218 (i.e., known time series) segment 228 (i.e., learned template). para [0122] “In particular, for each of flights 802, submatrix 810 includes a distance computed for one of P parameters with respect to State 1 806. For example, each column of submatrix 810 comprises, for each one of flights 802, a distance between a nominal segment and a segment extracted from time series data for a particular parameter during State 1 806 for the corresponding flight.” See figure 8.);
providing, with the at least one processor, the feature vector for each known time series of the plurality of known time series (para [0146] “For example, all of distance types 1101 shown may be computed for a pair of segmented time series. These distances may be put into a matrix form and input into a machine learning algorithm that generates a final score. This final score may be a prognostic indicator with respect to the health of the brake system.” An output is computed for each pair of segmented time series.).
Schimert does not explicitly disclose
stacking, with the at least one processor, the plurality of pairwise distance matrices together to generate a multi-dimensional tensor; and
processing, with the at least one processor, with a residual network, the multi-dimensional tensor by:
projecting, with a 2D convolutional layer including an input dimension that includes a plurality of dimensions, the multi-dimensional tensor to Euclidean space,
generating, with a rectified linear unit (ReLU) layer, based on the multi-dimensional tensor that is projected to Euclidean space, an intermediate representation for that known-time series,
generating, with a plurality of bottleneck building blocks, based on the intermediate representation, a merged representation for that known time-series,
generating, with a global pooling layer, based on the merged representation, a multi-dimensional vector for that known time-series, and
generating, with a linear layer, based on the multi-dimensional vector, a feature vector for that known time series; and
wherein the residual network is trained using a loss function that depends on a batch of training data in which each sample in the batch is a tuple including a query time series, a positive time series, and a negative time series.
However, Zhu (US 20240143699 A1) teaches
stacking, with the at least one processor, the plurality of pairwise distance matrices together to generate a multi-dimensional tensor (para [0039] “S14, stacking inner products of the normalized spectral embedding representations into a third-order tensor and using low-rank tensor representation learning to obtain a consistent distance matrix;” And para [0052] “Φ( ) represents the stacking of matrices into a tensor,” Inner product is a mathematical function for computing distances. para [0044] “Specifically, spectral embedding representations are firstly obtained from the similarity graphs of each view, and the inner products of multiple normalized spectral embedding representations are stacked into a third-order tensor.” Similarity graphs can be based on Euclidean distances para [0045] “Similarity graphs based on Euclidean distance can effectively capture the cluster structure of the data in this case.”);
Schimert and Zhu are analogous because they are both directed to the same field of endeavor of processing pairwise distance matrices with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert with the method of stacking distance matrices of Zhu.
Doing so would allow for obtaining a consistent distance matrix from information sources with multidimensional views to achieve better clustering performance (Zhu para [0003]).
Jiang (“SIMILARITY LEARNING FOR COVER SONG IDENTIFICATION USING CROSS-SIMILARITY MATRICES OF MULTI-LEVEL DEEP SEQUENCES”) teaches
processing, with the at least one processor, with a residual network, the multi-dimensional tensor (pg. 27, section 2.1; “The second network is a convolutional network Sγ parameterized by γ which is used to estimate the similarity between two tracks by feeding the cross-similarity matrices computed from two multi-level deep sequences of the two tracks.”) by;
generating, with a plurality of bottleneck building blocks, based on the intermediate representation, a merged representation for that known time-series (pg. 28, section 2.4; “Then, the cross-similarity matrices {Ck} are feed into four convolution networks which share weights to each other. The structure of the convolutional network is shown in Fig. 5. Then the outputs the the four convolutional network are concatenated into a vector and the vector are fed to a full connection layer activated by sigmoid function.”),
generating, with a global pooling layer, based on the merged representation, a multi-dimensional vector for that known time-series (pg. 27, fig. 2; “× 1.If the input size of Global Maxpooling1d layer is B × C × W × H, the output size will be B × C × 1 × H. X 1 represents the input.”), and
generating, with a linear layer, based on the multi-dimensional vector, a feature vector for that known time series (pg. 28, section 2.4; “Then, the cross-similarity matrices {Ck} are feed into four convolution networks which share weights to each other. The structure of the convolutional network is shown in Fig. 5. Then the outputs the the four convolutional network are concatenated into a vector and the vector are fed to a full connection layer activated by sigmoid function.”); and
Schimert and Jiang are analogous because they are both directed to the same field of endeavor of processing pairwise distance matrices with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert with the convolutional neural network of Jiang.
Doing so would allow for extracting deep sequences, at a particular level of abstraction, for the time series data to estimate the similarity between the two time series data (Jiang pg. 27).
Chang (“Audio Cover Song Identification using Convolutional Neural Network”) teaches
projecting, with a 2D convolutional layer including an input dimension that includes a plurality of dimensions (pg. 2, section 1; “We use a cross-similarity matrix generated from a pair of songs as an input feature.” Cross-similarity matrix (i.e., multi-dimensional tensor).), the multi-dimensional tensor to Euclidean space (pg. 2, section 2; “For δ, we calculate the Euclidean distance after applying the key alignment algorithm proposed in Serra et al. [2008b]… More specifically, a block of convolutional layers can sequentially perform sub-sampling and cross-correlation (or convolution) for distinguishing meaningful patterns from images in many different scales.”),
generating, with a rectified linear unit (ReLU) layer, based on the multi-dimensional tensor that is projected to Euclidean space, an intermediate representation for that known-time series (pg. 3, section 2; “Table 1: Specification of convolutional neural network: Inside the brackets are unit convolutional blocks, and outside the brackets is the number of stacked blocks. Conv denotes a same convolution layer with stride = 1, and its inside parentheses is (channel×width×height).”),
Schimert and Chang are analogous because they are both directed to the same field of endeavor of processing pairwise distance matrices with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert with the distance calculation of Chang.
Doing so would allow for improving the scalability of time series data discovery by proposing a novel embedding technique or metric subspace learning for the distance calculation, respectively (Chang pg. 2).
Malhotra (US 20210103812 A1) teaches
wherein the residual network is trained using a loss function that depends on a batch of training data in which each sample in the batch is a tuple including a query time series, a positive time series, and a negative time series (para [0030] “Triplets consist of two matching time series and a non-matching time series such that the loss aims to separate the positive pair from the negative by a distance margin. Given a set S.sub.j of all valid triplets of time series for a training task T.sub.j of the form (x.sub.l.sup.a,x.sub.l.sup.P,x.sub.l.sup.n)∈S.sub.j consisting of an anchor time series x.sub.l.sup.a, a positive time series x.sub.l.sup.P, and a negative time series x.sub.l.sup.n; where the positive time series is another instance from same class as the anchor, while the negative is from a different class than the anchor.”).
Schimert and Malhotra are analogous because they are both directed to the same field of endeavor of processing distances between time series data with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert with the training method using time series data of Malhotra.
Doing so would allow for classifying time series data when the amount of training data is minimal (Malhotra para [0005])
Regarding Claim 8,
Claim 8 is the system corresponding to the method of claim 1. Claim 8 is substantially similar to claim 1 and is rejected on the same grounds.
Regarding Claim 15,
Claim 15 is the computer program product corresponding to the method of claim 1. Claim 15 is substantially similar to claim 1 and is rejected on the same grounds.
Claims 2-3, 9-10, and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Schimert/Zhu/Jiang/Chang/Malhotra, as applied above, and further in view of Hirata et al. (US 20250020547 A1).
Regarding Claim 2,
Schimert, Zhu, Jiang, Chang, and Malhotra teach the method of claim 1.
Schimert, Zhu, Jiang, Chang, and Malhotra do not explicitly disclose
further comprising:
storing, with the at least one processor, in the at least one database, the feature vector for each known time series of the plurality of known time series.
However, Hirata (US 20250020547 A1) teaches
storing, with the at least one processor, in the at least one database, the feature vector for each known time series of the plurality of known time series (para [0038] “The normal vector registration processing unit 273 forms an M-dimensional vector including M types of variables at a same time interval using the M types of time-series signals clipped by the time-series signal clipping processing unit 271, and registers the obtained M-dimensional vector at each time point as a normal vector in the normal vector DB 5.” Time series corresponding to normal vectors (i.e., known time series). para [0051] “The operational DB 4 stores M or more types of time-series signals indicating the operation state of the target facility, acquired from the target facility during the past normal operation. The normal vector DB 5 stores the normal vector created by the normal vector DB creation processing unit 27 in the offline normal vector DB creation system 2.”).
Schimert, Zhu, Jiang, Chang, Malhotra, and Hirata are analogous because they are both directed to the same field of endeavor of computing distances for time-series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert, Zhu, Jiang, Chang, and Malhotra with the time series feature vector extraction of Hiarata.
Doing so would allow for leveling of distribution density of the vectors making it possible to achieve further higher accuracy of abnormality detection, higher calculation speed, and cost reduction by data capacity compression (Hiarata para [0055]).
Regarding Claim 3,
Schimert, Zhu, Jiang, Chang, Malhotra, and Hirata teach the method of claim 2. Schimert further teaches further comprising:
obtaining, with the at least on processor, time series (para [0054]-[0057] And para [0060] “Segment generator 204 may transform time series data 210 into plurality of segments 216 for use in evaluating health 211 of component 214. In particular, time series data 210 may be transformed into plurality of segments 216 based on at least one selected state 220 for aircraft 202 that is of interest.” Segments (i.e. time series data).);
computing, with the at least one processor, a pairwise distance matrix between the time series and each learned template of the plurality of learned templates to generate a further plurality of pairwise distance matrices (para [0071] segment 218 (i.e., time series) segment 228 (i.e., learned template). para [0122] “In particular, for each of flights 802, submatrix 810 includes a distance computed for one of P parameters with respect to State 1 806. For example, each column of submatrix 810 comprises, for each one of flights 802, a distance between a nominal segment and a segment extracted from time series data for a particular parameter during State 1 806 for the corresponding flight.” See figure 8.);
processing, with the at least one processor, the further tensor, receives, as input, the further tensor, and provides, as output, a feature for the time series (para [0146] “For example, all of distance types 1101 shown may be computed for a pair of segmented time series. These distances may be put into a matrix form and input into a machine learning algorithm that generates a final score. This final score may be a prognostic indicator with respect to the health of the brake system.” An output is computed for each pair of segmented time series.);
Zhu (US 20240143699 A1) teaches
stacking, with the at least one processor, the further plurality of pairwise distance matrices together to generate a further multi-dimensional tensor (para [0039] “S14, stacking inner products of the normalized spectral embedding representations into a third-order tensor and using low-rank tensor representation learning to obtain a consistent distance matrix;” And para [0052] “Φ( ) represents the stacking of matrices into a tensor,” Inner product is a mathematical function for computing Euclidean distances. para [0044] “Specifically, spectral embedding representations are firstly obtained from the similarity graphs of each view, and the inner products of multiple normalized spectral embedding representations are stacked into a third-order tensor.” Similarity graphs can be based on Euclidean distances para [0045] “Similarity graphs based on Euclidean distance can effectively capture the cluster structure of the data in this case.”);
Schimert and Zhu are analogous because they are both directed to the same field of endeavor of processing pairwise distance matrices with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert with the method of stacking distance matrices of Zhu.
Doing so would allow for obtaining a consistent distance matrix from information sources with multidimensional views to achieve better clustering performance (Zhu para [0003]).
Jiang further teaches
generating, with the at least one processor, with the residual network, based on the further multi-dimensional tensor, a feature vector for the time series (pg. 28 section 2.4; “Then, the cross-similarity matrices {Ck} are feed into four convolution networks which share weights to each other. The structure of the convolutional network is shown in Fig. 5. Then the outputs the the four convolutional network are concatenated into a vector and the vector are fed to a full connection layer activated by sigmoid function.”);
Schimert and Jiang are analogous because they are both directed to the same field of endeavor of processing pairwise distance matrices with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert with the convolutional neural network of Jiang.
Doing so would allow for extracting deep sequences, at a particular level of abstraction, for the time series data to estimate the similarity between the two time series data (Jiang pg. 27).
Schimert, Zhu, Jiang, Chang, and Malhotra do not explicitly disclose an unknown time series.
However, Hirata teaches
obtaining, with the at least on processor, an unknown time series (para [0045]-[0047] The time series data corresponding to the abnormality determination target is an unknown time series.);
for each known time series of the plurality of known time series stored in the database, determining, with the at least one processor, based on the stored feature vector for that known time series and the feature vector for the unknown time series, a distance between that known time series and the unknown time series (para [0048] “The neighboring data extraction processing unit 375 calculates a distance between the abnormality determination target vector extracted by the monitoring target data extraction processing unit 373 and each normal vector registered in the normal vector DB 5, and extracts a predetermined number of normal vectors from the normal vector DB 5 as neighboring vectors in ascending order of distance.” Abnormality determination target vector (i.e., feature vector for unknown time series).); and
identifying, with the at least one processor, based on the distance between each known time series and the unknown time series, at least one known time series determined to correspond to the unknown time series (para [0048] “The neighboring data extraction processing unit 375 calculates a distance between the abnormality determination target vector extracted by the monitoring target data extraction processing unit 373 and each normal vector registered in the normal vector DB 5, and extracts a predetermined number of normal vectors from the normal vector DB 5 as neighboring vectors in ascending order of distance.” The data extraction unit extracts (i.e., identified) a predetermined number of normal vectors (i.e., known time series) based on the distance.).
Schimert, Zhu, Jiang, Chang, Hirata are analogous because they are both directed to the same field of endeavor of computing distances for time-series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert, Zhu, Jiang, and Chang with the time series feature vector extraction of Hiarata.
Doing so would allow for leveling of distribution density of the vectors making it possible to achieve further higher accuracy of abnormality detection, higher calculation speed, and cost reduction by data capacity compression (Hiarata para [0055]).
Regarding Claim 9,
Claim 9 is the system corresponding to the method of claim 2. Claim 9 is substantially similar to claim 2 and is rejected on the same grounds.
Regarding Claim 10,
Claim 10 is the system corresponding to the method of claim 3. Claim 10 is substantially similar to claim 3 and is rejected on the same grounds.
Regarding Claim 16,
Claim 16 is the computer program product corresponding to the method of claim 2. Claim 16 is substantially similar to claim 2 and is rejected on the same grounds.
Regarding Claim 17,
Claim 17 is the computer program product corresponding to the method of claim 3. Claim 17 is substantially similar to claim 3 and is rejected on the same grounds.
Claim(s) 5, 12, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Schimert/Zhu/Jiang/Chang/Malhotra, as applied above, and further in view of Chu et al. (US-20170185664-A1).
Regarding Claim 5,
Schimert, Zhu, Jiang, Chang, and Malhotra teach the method of claim 1.
Schimert, Zhu, Jiang, Chang, and Malhotra do not explicitly disclose
wherein the plurality of known time series includes a plurality of known transaction time series associated with a plurality of merchants, and wherein each known time series is associated with metadata associated with a merchant associated with that known time series.
However, Chu (US 20170185664 A1) teaches
wherein the plurality of known time series includes a plurality of known transaction time series associated with a plurality of merchants (para [0044] “For example, time series data 39 may include sales data collected hourly for a retail store that has hours of being open for business of 9:00 AM to 9:00 PM on Monday through Saturday, and is closed on Sundays.”), and wherein each known time series is associated with metadata associated with a merchant associated with that known time series (para [0044] “Time interval metadata determining system 22 may analyze this data and determine four extra timing parameters in addition to the hourly time intervals, including parameters for the beginning of the week (W.sub.start), the number of days per week (n.sub.d), the beginning of the day (D.sub.start), and the number of hours per day (n.sub.h).” The metadata can describe the hours of which the merchant is open or closed for business such as weekends or holidays as described in paragraph [0043].).
Schimert, Zhu, Jiang, Chang, Malhotra, and Chu are analogous because they are both directed to the same field of endeavor of analyzing time series data with machine learning models.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the machine learning model of Schimert, Zhu, Jiang, Chang, and Malhotra with the time series metadata of Chu.
Doing so would allow for identifying statistical variables of the time series data such as maximum values, minimum values, etc. to identify irregularities (Chu para [0051]).
Regarding Claim 12,
Claim 12 is the system corresponding to the method of claim 5. Claim 12 is substantially similar to claim 5 and is rejected on the same grounds.
Regarding Claim 19,
Claim 19 is the computer program product corresponding to the method of claim 5. Claim 19 is substantially similar to claim 5 and is rejected on the same grounds.
Claims 6 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Schimert/Zhu/Jiang/Chang/Malhotra, as applied above, and further in view of Lines et al. (“Time series classification with ensembles of elastic distance measures”) and Hu et al. (US-20230343319-A1).
Regarding Claim 6,
Schimert, Zhu, Jiang, Chang, and Malhotra teach the method of claim 1.
Schimert, Zhu, Jiang, Chang, and Malhotra do not explicitly disclose
wherein the plurality of learned templates includes thirty-two learned templates, wherein the plurality of pairwise distance matrices includes thirty-two pairwise distance matrices, wherein the multi-dimensional tensor includes an input dimension of thirty-two, and wherein the feature vector for each known time series of the plurality of known time series includes a size sixty-four vector.
However, Lines (“Time series classification with ensembles of elastic distance measures”) teaches
wherein the plurality of learned templates includes thirty-two learned templates, wherein the plurality of pairwise distance matrices includes thirty-two pairwise distance matrices (pg. 568, section 2.2; “Suppose we want to measure the distance between two series, a = {a1, a2,..., am} and b = {b1, b2,..., bm}. Let M(a, b) be the m × m pointwise distance matrix between a and b, where Mi,j = (ai − bj)2.” And pg. 580, section 5.1 “The graph presented in Fig. 5 demonstrates the difference in time taken between the measures when calculating the distance between two series from each of the 75 datasets.” 75 distance matrices can be computed with includes 32 learned templates and 32 pairwise distance matrices.),
Schimert, Zhu, Jiang, Chang, and Lines are analogous because they are both directed to the same field of endeavor of computing distances between time series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the time series model of Schimert, Zhu, Jiang, and Chang with the method of computing distances between time series of Lines.
Doing so would allow for improving the accuracy of the time series classifications while maintaining a competitive runtime complexity (Lines pg. 585 section 6;).
Hu (US 20230343319 A1) teaches
wherein the multi-dimensional tensor includes an input dimension of thirty-two (para [0304] “The second LSTM block 205 maps the inputs to a sequence of 32-dimensional vectors.”), and wherein the feature vector for each known time series of the plurality of known time series includes a size sixty-four vector (para [0313] “The first neural network 222 outputs a 64-dimensional vector for the decoding step j.”).
Schimert, Zhu, Jiang, Chang, Malhotra, and Hu are analogous because they are both directed to the same field of endeavor of neural networks.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the neural network of Schimert, Zhu, Jiang, Chang, and Malhotra with the method of extracting feature vectors of Hu.
Doing so would allow for drawing the feature vectors from a probability distribution which may provide increased robustness, and also provide improved naturalness (Hu para [0406]).
Regarding Claim 13,
Claim 13 is the system corresponding to the method of claim 6. Claim 13 is substantially similar to claim 6 and is rejected on the same grounds.
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Schimert/Zhu/Jiang/Chang/Malhotra, as applied above, and further in view of Lines et al. (“Time series classification with ensembles of elastic distance measures”), Hu et al. (US-20230343319-A1), and Yao et al. (US-20200257902-A1).
Regarding Claim 20,
Schimert, Zhu, Jiang, Chang, and Malhotra teach the computer program product of claim 15.
Schimert, Zhu, Jiang, Chang, and Malhotra do not explicitly disclose
wherein the plurality of learned templates includes thirty-two learned templates, wherein the plurality of pairwise distance matrices includes thirty-two pairwise distance matrices, wherein the tensor includes an input dimension of thirty-two, and wherein the feature vector for each known time series of the plurality of known time series includes a size sixty-four vector, and wherein the residual network includes a two-dimensional residual network.
However, Lines (“Time series classification with ensembles of elastic distance measures”) teaches
wherein the plurality of learned templates includes thirty-two learned templates, wherein the plurality of pairwise distance matrices includes thirty-two pairwise distance matrices (pg. 568, section 2.2; “Suppose we want to measure the distance between two series, a = {a1, a2,..., am} and b = {b1, b2,..., bm}. Let M(a, b) be the m × m pointwise distance matrix between a and b, where Mi,j = (ai − bj)2.” And pg. 580, section 5.1 “The graph presented in Fig. 5 demonstrates the difference in time taken between the measures when calculating the distance between two series from each of the 75 datasets.” 75 distance matrices can be computed with includes 32 learned templates and 32 pairwise distance matrices.),
Schimert, Zhu, Jiang, Chang, Malhotra, and Lines are analogous because they are both directed to the same field of endeavor of computing distances between time series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the time series model of Schimert, Zhu, Jiang, and Chang with the method of computing distances between time series of Lines.
Doing so would allow for improving the accuracy of the time series classifications while maintaining a competitive runtime complexity (Lines pg. 585 section 6;).
Hu (US 20230343319 A1) teaches
wherein the tensor includes an input dimension of thirty-two (para [0304] “The second LSTM block 205 maps the inputs to a sequence of 32-dimensional vectors.”), and wherein the feature vector for each known time series of the plurality of known time series includes a size sixty-four vector (para [0313] “The first neural network 222 outputs a 64-dimensional vector for the decoding step j.”).
Schimert, Zhu, Jiang, Chang, Malhotra, and Hu are analogous because they are both directed to the same field of endeavor of neural networks.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the neural network of Schimert, Zhu, Jiang, and Chang with the method of extracting feature vectors of Hu.
Doing so would allow for drawing the feature vectors from a probability distribution which may provide increased robustness, and also provide improved naturalness (Hu para [0406]).
Yao (US 20200257902 A1) teaches
wherein the residual network includes a two-dimensional residual network (para [0021] “The feature extractor 110 can be implemented by any suitable type of 2D convolutional neural network (2D-CNN), such as a residual neural network (ResNet) or a visual geometry group (VGG) neural network.”).
Schimert, Zhu, Jiang, Chang, Malhotra, and Yao are analogous because they are both directed to the same field of endeavor of neural networks.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the neural network of Schimert, Zhu, Jiang, Chang, and Malhotra with the residual network of Yao.
Doing so would allow for facilitating the training of the neural network by reducing the dimensionality of the input data and the size of the neural network (Yao para [0056]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HENRY K NGUYEN whose telephone number is (571)272-0217. The examiner can normally be reached Mon - Fri 7:00am-4:30pm.
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, Li B Zhen can be reached at 5712723768. 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.
/HENRY NGUYEN/Examiner, Art Unit 2121