DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This action is responsive to communications: RCE filed on 4/03/2026.
Claims 1-14 are pending. Claims 1 is independent.
The previous rejection of claims 1-14 under 35 USC § 103 have been withdrawn in view of the amendment.
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) 1-4, 8, 10-14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chong et al. (Abnormal Event Detection in Videos using Spatiotemporal Autoencoder) in view of Kwak et al. (Unsupervised Abnormal Sensor Signal Detection With Channelwise Reconstruction Errors) and Ahn et al. (Deep Generative Models-Based Anomaly Detection for Spacecraft Control Systems) and Wu et al. (US2021/0256290) .
In regards to claim 1, Chong et al. a machine learning apparatus comprising a processing circuit configured to
train a first learning parameter of an extraction layer configured to extract, from input data, feature data of the input data, based on a plurality of training data (Chong et al. pg5 section3.2.1 para1, autoencoder trains encoder model to extract features), and
train a second learning parameter of a reconstruction layer configured to generate reconstructed data of the input data, based on a plurality of training feature data obtained by applying the trained extraction layer to the plurality of training data (Chong et al. pg5 section3.2.1 para1, autoencoder trains decoder model to minimize reconstruction error), and
the representative vectors as many as the dimension count are defined by a weighted sum of the plurality of training data (Chong et al. pg7 section3.2.4 para2, weight matrixes trained from input vectors).
Chong et al. does not explicitly disclose wherein: the second learning parameter comprises a number of horizontal rows equal to a dimension count of the input data, and a number of vertical columns equal to a dimension count of the feature data, such that the second learning parameter is arranged as representative vectors in a number equal to the dimension count of the feature data.
However Wu et al. substantially discloses wherein: the second learning parameter comprises a number of horizontal rows equal to a dimension count of the input data, and a number of vertical columns equal to a dimension count of the feature data, such that the second learning parameter is arranged as representative vectors in a number equal to the dimension count of the feature data (Wu et al. para[0144], extracts the feature vector calculated for each channel, thereby learning the parameters for each layer of the CNN).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the feature extraction method of Wu et al. in order to efficiently optimize the parameters (Wu et al. para[0015]).
Chong et al. does not explicitly disclose the extraction layer and the reconstruction layer are included in a neural network configured to receive the input data and output a determination result of presence or absence of abnormality of the input data.
However Ahn et al. discloses The extraction layer and the reconstruction layer are included in a neural network configured to receive the input data and output a determination result of presence or absence of abnormality of the input data (Ahn et al. pg8 section 4 para2, a series of processes for anomaly detection using multidimensional time-series data include data preprocessing (including purification, integration, cleanup, and transformation), feature extraction using deep networks, and learning models represented by algorithms considering robustness and performance optimization).
It would have been obvious to one of ordinary skill in the art before the filing data of the invention to have combined the abnormality detection method of Chong et al. with the deep generative models of Ahn et al. in order to determine if a system has deviated from the proper operating range due to a failure (Ahn et al. pg1 section 1 para1).
In regards to claim 8, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 1, wherein the processing circuit trains the second learning parameter by minimizing an error between the training feature data and training reconstructed data obtained by applying the training feature data to the reconstruction layer (Chong et al. pg5 section 3.2.1 para1, model is trained to minimize reconstruction error).
In regards to claim 10, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 1, wherein the neural network including the extraction layer and the reconstruction layer includes a determination layer configured to output a determination result of presence/absence of abnormality of the input data based on comparison between a threshold and an error between the reconstructed data and the input data (Chang et al. pg9 section3.4.1, threshold determines how sensitive system is to identifying frame as anomalous based on a reconstruction error).
In regards to claim 11, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 1, wherein the representative vectors as many as the dimension count are defined by the weighted sum of the plurality of training data, and the weight has a value based on the plurality of training feature data (Chong et al. pg7 section3.2.4 para2, bias vectors have dimension based on trainable weight matrices).
In regards to claim 12, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 1, wherein the dimension count is decided in accordance with a storage capacity that is assigned to a memory of an apparatus in which the neural network including the extraction layer and the reconstruction layer is implemented, and is needed for the neural network (Chong et al. pg5 section 3.1 para2, reduces and normalizes dimensionality of data).
In regards to claim 13, Chong et al. as modified by Kwak et al. and Ahn et al. discloses a system comprising:
The machine learning apparatus according to claim 1; and
an abnormality detection apparatus comprising:
a processing circuit configured to:
extract feature data from the diagnostic data using the extraction layer trained by the machine learning apparatus (Chong et al. pg5 section3.2.1 para1, autoencoder trains encoder model to extract features),
generate reconstructed data from the feature data using the reconstruction layer trained by the machine learning apparatus, the reconstructed data being generated based on a weighted sum of the feature data (Chong et al. pg5 section3.2.1 para1, autoencoder trains decoder model to minimize reconstruction error), and
determine presence/absence of abnormality of the diagnostic data based on the diagnostic data and the reconstructed data (Chong et al. pg9 section3.4.1, determines if data is normal or anomalous based on anomaly threshold parameter).
Chong et al. does not explicitly disclose representative vectors as many as a dimension count of the feature data.
However Wu et al. substantially discloses representative vectors in the number equal to the dimension count of the feature data (Wu et al. para[0144], extracts the feature vector calculated for each channel, thereby learning the parameters for each layer of the CNN).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the feature extraction method of Wu et al. in order to efficiently optimize the parameters (Wu et al. para[0015]).
Chong et al. does not explicitly disclose obtain diagnostic data from a manufacturing machine for an abnormality determination target or an inspection device of the manufacturing machine.
However Ahn et al. discloses obtain diagnostic data from a manufacturing machine for an abnormality determination target or an inspection device of the manufacturing machine (Ahn et al. pg2 section 1 para3, Time-series data transmitted from each subsystem, part, and sensor are needed to diagnose the spacecrafts system’s mechanical condition).
It would have been obvious to one of ordinary skill in the art before the filing data of the invention to have combined the abnormality detection method of Chong et al. with the deep generative models of Ahn et al. in order to determine if a system has deviated from the proper operating range due to a failure (Ahn et al. pg1 section 1 para1).
Claim 14 recites substantially similar limitations to claim 13. Thus claim 14 is rejected along the same rationale as claim 13.
Claim(s) 2-4 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chong et al. and Wu et al. and Ahn et al. as applied to claim 1 above, and further in view of Kwak et al. ("Unsupervised Abnormal Sensor Signal Detection With Channelwise Reconstruction Errors").
In regards to claim 2, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 1,
Chong et al. does not explicitly disclose wherein the processing circuit
calculates a false detection rate concerning abnormality detection based on the training feature data obtained by applying the trained extraction layer to the training data and training reconstructed data obtained by applying the trained reconstruction layer to the training feature data, and
displays the false detection rate on a display device.
However Kwak et al. discloses wherein the processing circuit
calculates a false detection rate concerning abnormality detection based on the training feature data obtained by applying the trained extraction layer to the training data and training reconstructed data obtained by applying the trained reconstruction layer to the training feature data (Kwak et al. pg40001 section IV.B para para7, calculates false positive rate based on trained hyperparameters), and
displays the false detection rate on a display device (Kwak et al. table 2 pg40002 section IV.D para4, displays FPR results).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the anomaly detection method of Kwak et al. in order to prevent loss of valuable channel information through reconstruction errors (Kwak et al. pg39995 abstract).
In regards to claim 3, Chong et al. as modified Wu et al., Ahn et al. and Kwak discloses the apparatus according to claim 2, wherein the processing circuit
calculates a probability distribution of an error between the training feature data and the training reconstructed data (Chong et al. pg9 section 3.3, calculates the reconstruction error),
calculates, as the false detection rate, a probability that the error is not less than a threshold in the probability distribution (Chong et al. pg9 section 3.4.1, calculates at detection threshold where false positive rate is equal to false negative rate), and
displays a graph of the false detection rate for the threshold (Chong et al. fig. 4 pg13 section 4.3.2. para1-2, displays graph of regularity score).
In regards to claim 4, Chong et al. as modified Wu et al., Ahn et al. and Kwak discloses the apparatus according to claim 3, wherein the processing circuit sets a threshold used to determine presence/absence of abnormality of the input data using the neural network including the extraction layer and the reconstruction layer to a value designated via the graph by an operator (Kwak et al. pg39999 section III.B para3, user defines anomaly rate used to set the anomaly score value corresponding to the threshold to detect anomalies).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the anomaly detection method of Kwak et al. in order to prevent loss of valuable channel information through reconstruction errors (Kwak et al. pg39995 abstract).
Claim(s) 5-7, and 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chong et al. and Wu et al. and Ahn et al. as applied to claim 1 and 8 above, and further in view of Xing et al. (US2020/0104984).
In regards to claim 5, Chong et al. as modified by Wu et al. and Ahn et al. discloses the apparatus according to claim 1. Chong et al. does not explicitly disclose wherein if the training data includes only normal data, the processing circuit trains the first learning parameter such that positive correlation between an inner product of two normal data and an inner product of two feature data corresponding to the two normal data becomes high.
However Xing et al. discloses wherein if the training data includes only normal data, the processing circuit trains the first learning parameter such that positive correlation between an inner product of two normal data and an inner product of two feature data corresponding to the two normal data becomes high (Xing et al. fig. 8 a-c para[0089], correlation becomes positive for normal data).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the dimension reduction method of Xing et al. in order to reduce a large number of variables into few comprehensive indicators (Xing et al. para[0004]).
In regards to claim 6, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 1. Chong et al. does not explicitly disclose wherein if the training data includes normal data and abnormal data, the processing circuit trains the first learning parameter such that negative correlation between an inner product of the normal data and the abnormal data and an inner product of feature data corresponding to the normal data and feature data corresponding to the abnormal data becomes high.
However Xing et al. discloses wherein if the training data includes normal data and abnormal data, the processing circuit trains the first learning parameter such that negative correlation between an inner product of the normal data and the abnormal data and an inner product of feature data corresponding to the normal data and feature data corresponding to the abnormal data becomes high (Xing et al. fig. 8 a-c para[0089], correlation becomes negative for abnormal data).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the dimension reduction method of Xing et al. in order to reduce a large number of variables into few comprehensive indicators (Xing et al. para[0004]).
In regards to claim 7,Chong et al. as modified by Wu et al. and Ahn et al. discloses the apparatus according to claim 1. Chong et al. wherein the processing circuit trains the first learning parameter by contrastive learning and decorrelation based on an inner product of the training data and an inner product of feature data corresponding to the training data.
However Xing et al. discloses wherein the processing circuit trains the first learning parameter by contrastive learning and decorrelation based on an inner product of the training data and an inner product of feature data corresponding to the training data (Xing et al. fig. 7 para[0074], generates correlation matrix from abnormal and normal sample image data).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the dimension reduction method of Xing et al. in order to reduce a large number of variables into few comprehensive indicators (Xing et al. para[0004]).
In regards to claim 9, Chong et al. as modified Wu et al. and Ahn et al. discloses the apparatus according to claim 8. Chong et al. does not explicitly disclose wherein the reconstruction layer is a linear regression model.
Xing et al. discloses wherein the reconstruction layer is a linear regression model (Xing et al. para[0080], reconstruction layer is a ridge regression classification model).
It would have been obvious to one of ordinary skill in the art before the filing date of the invention to have combined the abnormality detection method of Chong et al. with the dimension reduction method of Xing et al. in order to reduce a large number of variables into few comprehensive indicators (Xing et al. para[0004]).
Response to Arguments
Applicant’s arguments with respect to claims 1-20 have been considered but are moot because the arguments do not apply the current rejection.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Zhang et al. (US2021/0089913) teaches a learning parameter having dimensions equal to hidden layers of a neural network.
Mitani et al. (US2024/0039940) teaches a generating a matrix of representative vectors.
Yoshida (US2022/0083039) teaches generating a learning parameter for multidimensional vector data.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS HASTY whose telephone number is (571)270-7775. The examiner can normally be reached Monday-Friday 8:30am-5:00pm.
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, Matt Ell can be reached at (571)270-3264. 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.
/N.H/Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights