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 .
Response to Amendment
This Office action has been issued in response to amendment filed on 04/15/2026, Claims (1-6), 15 and 17 are pending. Applicants' arguments have been carefully and respectfully considered and addressed. Accordingly, this action has been made FINAL necessitated by amendment.
Claims (1-2, 4-9, 13), (10-11, 14-17) and (18-20) are presented for examination.
Response to Arguments
Applicants' arguments have been carefully and respectfully considered and addressed. The arguments presented are moot based on amendment.
With regards to 101 rejection, the applicant’s arguments have been fully considered and they are persuasive, therefore; the rejection is withdrawn.
Applicant arguments and amendment were fully considered and are moot in view of the new ground rejection wherein Bersia. US Patent Application Publication US 20220114637 A1 (hereinafter Bersia) in view of Eriksson et al. US Patent Application Publication US 20190073594 A1 (hereinafter Eriksson) and further in view of Salle et al. US Patent Application Publication US 20220129712 A1 (hereinafter Salle) and further in view of Oono et al. US Patent Application Publication US Patent Application Publication US 20170161635 A1 (hereinafter Oono) for teaching the amended claims.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 04/07/2026 were filed prior to current Office Action. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements is being considered by the examiner.
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 of this title, 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 1-2, 4-11 and 13-20 are rejected under AIA 35 U.S.C. 103(a) as being unpatentable over Bersia. US Patent Application Publication US 20220114637 A1 (hereinafter Bersia) in view of Eriksson et al. US Patent Application Publication US 20190073594 A1 (hereinafter Eriksson) and further in view of Salle et al. US Patent Application Publication US 20220129712 A1 (hereinafter Salle) and further in view of Oono et al. US Patent Application Publication US Patent Application Publication US 20170161635 A1 (hereinafter Oono).
Regarding claim 1, Bersia teaches A method comprising: obtaining a database of reconstruction error vector samples (FIG. 2, [0086], [0199-0200] wherein Bersia describes a mapping of encoded features obtained in data samples that is stored in a database, wherein the mapping is performed by a decoder to a reconstruction of the samples as an input).
Bersia does not teach generated by an unsupervised autoencoder from input data samples.
However in analogous art of using an autoencoder for determining of event types, Eriksson teaches generated by an unsupervised autoencoder from input data samples (Abstract, [0004], [0007], [0011] wherein Eriksson describes producing a structure of dataset using unsupervised learning with a plurality of autoencoders).
It would have been obvious to a person in the ordinary skill in the art before the effective filing date of the claimed invention to combine Eriksson with Bersia by incorporating the method of generated by an unsupervised autoencoder from input data samples of Eriksson into the method of obtaining a database of reconstruction error vector samples of Bersia for the purpose of incorporating an aggregator that combines the plurality of indications based on the weighting vector to provide a weighted combination of the plurality of indications, wherein the weighted combination corresponds to a reconstruction of the input data; and an error module determining an error between the reconstruction and the input data, wherein the controller is responsive to the error to adjust the weighting vector to reduce the error (Eriksson: [0011]).
Bersia does not teach clustering the reconstruction error vector samples sampled from said database into clusters of reconstruction error vector samples; selecting candidate samples from a first cluster included in the clusters; assigning a label to each of the candidate samples; and applying the label to all reconstruction error vector.
However in analogous art of using an autoencoder for determining of event types. Salle teaches clustering the reconstruction error vector samples sampled from said database into clusters of reconstruction error vector samples (FIG. 1, Abstract, [0013], [0019-0024] wherein Salle describes, as illustrated in FIG. 1, processing data samples as input through an autoencoder that includes an encoder and decoder and clusters data into a constructed content that includes reconstruction loss or error) selecting candidate samples from a first cluster included in the clusters; assigning a label to each of the candidate samples (FIG. 1, [0017-0018], [0022], [0024], [0029], [0047], [0052] wherein Salle describes label assignment that is based on clusterer as illustrated in FIG. 1) and applying the label to all reconstruction error vector samples (FIG. 1, [0017-0024], [0029], [0047], [0052] wherein Salle incorporates the latent feature vector that can be provided to a clustering classification layer as the clusterer in FIG. 1. The clusterer determines a cluster of the clusters to which the content belongs. The clusterer determines a distance between the latent feature vector of the content and latent features vectors of one or more points (e.g., a central value, such as a latent feature vector of a centroid) of the clusters. The distance can be converted to a predicted probability that indicates how likely it is that the content belongs to the cluster).
It would have been obvious to a person in the ordinary skill in the art before the effective filing date of the claimed invention to combine Salle with Bersia by incorporating the method of clustering the reconstruction error vector samples sampled from said database into clusters of reconstruction error vector samples; selecting candidate samples from a first cluster included in the clusters; assigning a label to each of the candidate samples; and applying the label to all reconstruction error vector samples of Salle into the method of obtaining a database of reconstruction error vector samples of Bersia for the purpose of training the clustering autoencoder to cluster an input in a latent feature (Salle: [0013]).
Bersia does not teach applying the label to all reconstruction error vector samples in the first cluster in response to determining that the label assigned to a sufficient subset of the candidate samples is the same label.
However in analogous art of using an autoencoder for determining of event types, Oono teaches applying the label to all reconstruction error vector samples in the first cluster in response to determining that the label assigned to a sufficient subset of the candidate samples is the same label. (FIG. 2A, [0006-0012], [0041-0042], [0060], [0131-0135], [0182], [0199], wherein Oono describes a system that includes autoencoder that is configured to encode latent variables and decode latent representations and generate variables over values. Wherein Oono generates reconstructions of chemical compound fingerprints, wherein the system's training is constrained by the reconstruction error. The reconstruction error may comprise the negative likelihood that an encoded chemical compound representation is drawn from the random variable generated by a probabilistic decoder. The system may be trained to optimize, for example to minimize, the reconstruction error. The training is constrained by a loss function comprising the reconstruction error and a regularization error. The probabilistic autoencoder may be trained to learn to approximate an encoding distribution. The regularization error may comprise a penalty associated with the complexity of the encoding distribution. The training may comprise minimizing the loss function. The training labels comprise one or more label elements having predetermined values. The system is configured to receive a target label comprising one or more label elements and generate chemical compound fingerprints that satisfy a specified value for each of the one or more label elements. Wherein the system further comprises a predictor that is configured to predict values of selected label elements for chemical compound fingerprints, the label comprise one or more label elements selected from the group consisting of bioassay results, toxicity, cross-reactivity, pharmacokinetics, pharmacodynamics, bioavailability, and solubility. The training may comprise (1) inputting representations of chemical compounds and their associated labels, and (2) generating reconstructions of chemical compound fingerprints. The generative model may comprise a probabilistic or variational autoencoder comprising a) a probabilistic encoder for encoding fingerprint and label data as a latent variable from which a latent representation may be sampled; b) a probabilistic decoder for converting latent representations to random variables from which reconstructions of the fingerprint data may be sampled; and c) a sampling module for sampling a latent variable to generate a latent representation, or sampling a random variable to generate a fingerprint reconstruction. The system may be trained to optimize, for example to minimize, the loss function comprising a reconstruction error and a regularization error. The reconstruction error may comprise the negative likelihood that an encoded chemical compound representation is drawn from the random variable output by the probabilistic decoder. The training may comprise having the variational or probabilistic autoencoder learn to approximate an encoding distribution. Oono provides examples as described in [0125-0135] wherein Oono clusters a group of elements and compares them as latent representations and applies the labels, wherein the labels might be different or the same based on the reconstruction error and based on calculating pairwise comparison).
It would have been obvious to a person in the ordinary skill in the art before the effective filing date of the claimed invention to combine Oono with Bersia by incorporating the method of applying the label to all reconstruction error vector samples in the first cluster in response to determining that the label assigned to a sufficient subset of the candidate samples is the same label of Oono into the method of obtaining a database of reconstruction error vector samples of Bersia for the purpose of incorporating a sampling module configured to sample from the latent variables to generate latent representations or from a random variable to generate a reconstruction of a fingerprint; and wherein the training labels comprise one or more label elements having empirical or predicted values, wherein the system's training is constrained by a reconstruction error. (Oono: claim 16 text).
Regarding claim 2, Bersia as modified by Eriksson, Salle and Oono teach collecting the input data samples from multiple nodes operating in an environment and storing the data samples in a sample database, wherein the data samples are associated to the reconstruction error vector samples (FIG. 2, [0086], [0195-0200] wherein Bersia describes a mapping of encoded features obtained in data samples that is stored in a database, wherein the mapping is performed by a decoder to a reconstruction of the samples as an input).
Regarding claim 4, Bersia as modified by Eriksson, Salle and Oono teach wherein the reconstruction error vector samples are generated as an absolute element-wise difference between the input data samples input into the unsupervised encoder and reconstruction samples output from the unsupervised autoencoder (FIG. 1, [0022] wherein Salle describes how the decoder can construct content based on the latent feature vector wherein a difference between the content input into the encoder and the reconstructed content is produced by the decoder that can be determined based on a cost function known as the loss. Additionally, a difference between the label and the predicted probability from the clusterer can be determined based on the loss. The cost determined by the loss can be fed back to the clustering autoencoder. The loss can be backpropagated through the clustering autoencoder, such as for training the clustering autoencoder), (FIG. 2, [0046] wherein Eriksson describes an error determination network or module that outputs an error signal, e.g, error in FIG. 2. Module 160 processes the reconstructed data produced by block 140 and the input data 100 to determine an error. Ideal processing of the input data by an autoencoder should produce reconstructed data at block 140 that is a faithful or accurate representation or reconstruction of the input data. However, as explained further below in regard to FIG. 3, an embodiment of an autoencoder in accordance with the present principles produces a reconstruction at block 140 exhibiting an error or difference with respect to the input data. Module 160 determines that difference or error and provides the error determination, e.g., signal error, to controller 120 and to the plurality of autoencoders 110 (e.g., comprising autoencoders 110.sub.1 through 110.sub.k in FIG. 2). Controller 120 and autoencoders 110 processes the error to adjust or modify the weighting vector, e.g., w.sub.k in FIG. 2, to reduce the error), ([0158-0162] wherein Oono describes how the decoder outputs a pair of a real-valued vector of means and a real-valued vector of standard deviations wherein the dimension of vectors is the same as the dimension of the feature vectors which are the fingerprints used in the training of the model), ([0125-0135] wherein Oono clusters a group of elements and compares them as latent representations and applies the labels, wherein the labels might be different or the same based on the reconstruction error and based on calculating pairwise comparison).
Regarding claim 5, Bersia as modified by Eriksson, Salle and Oono teach retrieving context samples from the input data samples for each of the candidate samples, wherein the context samples occur immediately before and/or after the corresponding candidate sample ([0005], [0011], [0015], [0030], [0043-0045], [0047-0049] wherein Salle incorporates samples based on clustering assignment wherein the samples are assigned a value that indicates low confidence or high confidence), ([0180], [0231] wherein Bersia teaches generating a high-dimensional vector based on the plurality of features associated with a vehicle includes encoding each of the plurality of features according to a one hot encoding technique. One-hot encoding converts categorical data into integer data, e.g., binary format. This technique also creates additional features based on the number of unique values in the categorical feature. For instance, every unique value in a category will be added as a feature), ([0158-0162] wherein Oono describes how the decoder outputs a pair of a real-valued vector of means and a real-valued vector of standard deviations wherein the dimension of vectors is the same as the dimension of the feature vectors which are the fingerprints used in the training of the model), ([0125-0135] wherein Oono clusters a group of elements and compares them as latent representations and applies the labels, wherein the labels might be different or the same based on the reconstruction error and based on calculating pairwise comparison).
Regarding claim 6, Bersia as modified by Eriksson, Salle and Oono teach considering the context samples associated to the candidate samples when assigning labels to the candidate samples (FIG. 1. [0017-0018], [0022], as shown in FIG. 1, Salle assigns labels to samples).
Regarding claim 7, Bersia as modified by Salle teaches deploying an unsupervised autoencoder to nodes in an environment (FIG. 1, [0002], [0013] wherein Salle describes a system that deploys an autoencoder to nodes), (Abstract, [0004], [0007], [0011] wherein Eriksson describes producing a structure of dataset using unsupervised learning with a plurality of autoencoders).
Regarding claim 8, Bersia as modified by Eriksson, Salle and Oono teach generating, by the unsupervised autoencoder operating on a node, a first reconstructed sample output from a first data sample input to the unsupervised autoencoder; generating a first reconstruction error vector sample from the data sample input and the reconstructed sample output; determining whether the first reconstruction error vector sample belongs in the first cluster; and applying the label associated with the reconstruction error samples in the first cluster to the first reconstruction error sample (FIG. 1, [0022] wherein Salle describes how the decoder can construct content based on the latent feature vector wherein a difference between the content input into the encoder and the reconstructed content is produced by the decoder that can be determined based on a cost function known as the loss. Additionally, a difference between the label and the predicted probability from the clusterer can be determined based on the loss. The cost determined by the loss can be fed back to the clustering autoencoder. The loss can be backpropagated through the clustering autoencoder, such as for training the clustering autoencoder), (FIG. 1, [0022] wherein Salle produces a reconstruction content that includes reconstruction loss or error), ([0209-0210] wherein Bersia generates a reconstruction error), (FIG. 1, [0017-0024], [0029], [0047], [0052] wherein Salle incorporates the latent feature vector that can be provided to a clustering classification layer as the clusterer in FIG. 1. The clusterer determines a cluster of the clusters to which the content belongs. The clusterer determines a distance between the latent feature vector of the content and latent features vectors of one or more points (e.g., a central value, such as a latent feature vector of a centroid) of the clusters. The distance can be converted to a predicted probability that indicates how likely it is that the content belongs to the cluster), (FIG. 2A, [0006-0012], [0041-0042], [0060], [0131-0135], [0182], [0199], wherein Oono describes a system that includes autoencoder that is configured to encode latent variables and decode latent representations and generate variables over values. Wherein Oono generates reconstructions of chemical compound fingerprints, wherein the system's training is constrained by the reconstruction error. The reconstruction error may comprise the negative likelihood that an encoded chemical compound representation is drawn from the random variable generated by a probabilistic decoder. The system may be trained to optimize, for example to minimize, the reconstruction error. The training is constrained by a loss function comprising the reconstruction error and a regularization error. The probabilistic autoencoder may be trained to learn to approximate an encoding distribution. The regularization error may comprise a penalty associated with the complexity of the encoding distribution. The training may comprise minimizing the loss function. The training labels comprise one or more label elements having predetermined values. The system is configured to receive a target label comprising one or more label elements and generate chemical compound fingerprints that satisfy a specified value for each of the one or more label elements. Wherein the system further comprises a predictor that is configured to predict values of selected label elements for chemical compound fingerprints, the label comprise one or more label elements selected from the group consisting of bioassay results, toxicity, cross-reactivity, pharmacokinetics, pharmacodynamics, bioavailability, and solubility. The training may comprise (1) inputting representations of chemical compounds and their associated labels, and (2) generating reconstructions of chemical compound fingerprints. The generative model may comprise a probabilistic or variational autoencoder comprising a) a probabilistic encoder for encoding fingerprint and label data as a latent variable from which a latent representation may be sampled; b) a probabilistic decoder for converting latent representations to random variables from which reconstructions of the fingerprint data may be sampled; and c) a sampling module for sampling a latent variable to generate a latent representation, or sampling a random variable to generate a fingerprint reconstruction. The system may be trained to optimize, for example to minimize, the loss function comprising a reconstruction error and a regularization error. The reconstruction error may comprise the negative likelihood that an encoded chemical compound representation is drawn from the random variable output by the probabilistic decoder. The training may comprise having the variational or probabilistic autoencoder learn to approximate an encoding distribution. Oono provides examples as described in [0125-0135] wherein Oono clusters a group of elements and compares them as latent representations and applies the labels, wherein the labels might be different or the same based on the reconstruction error and based on calculating pairwise comparison).
Regarding claim 9, Bersia as modified by Eriksson, Salle and Oono teach automatically determining a threshold value for the unsupervised autoencoder based on a distance of reconstruction error vector samples from a centroid of a cluster near an origin of a cluster space (FIG. 1, Claims 8 and 17 text, [0013], [0039], [0059], [0082], [0091] wherein Salle describes a probability of being in a class that can be determined based on a distance from a central value that represents the class, such as a latent feature representation of a centroid or other central point of the cluster).
Claims 10 and 18 are similar in scope to claim 1 therefore the claim is rejected under similar rationale.
Claim 11 is similar in scope to claim 2 therefore the claim is rejected under similar rationale.
Claim 13 is similar in scope to claim 4 therefore the claim is rejected under similar rationale.
Claim 14 is similar in scope to claim 5 therefore the claim is rejected under similar rationale.
Claim 15 is similar in scope to claim 6 therefore the claim is rejected under similar rationale.
Claim 16 is similar in scope to claim 8 therefore the claim is rejected under similar rationale.
Claim 17 is similar in scope to claim 9 therefore the claim is rejected under similar rationale.
Regarding claim 19, Bersia as modified by Eriksson, Salle and Oono teach clustering reconstruction error samples associated with data samples that have been processed by an autoencoder into clusters and labelling each of the reconstruction error samples in each of the clusters based on labels assigned to candidate samples from each of the clusters (FIG. 1. [0017-0018], [0022], as shown in FIG. 1, Salle assigns labels to samples), ([0011], [0015], [0030], [0043], [0047-0049] wherein Salle incorporates samples based on clustering assignment wherein the samples are assigned a value that indicates low confidence or high confidence).
Claim 20 is similar in scope to claim 9 therefore the claim is rejected under similar rationale.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HASSAN MRABI whose telephone number is (571)272-8875. The examiner can normally be reached on Monday-Friday, 7:30am-5pm. Alt, Friday, EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Viker Lamardo can be reached on 571-270-5871. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/HASSAN MRABI/Examiner, Art Unit 2144
Prosecution Insights