DETAILED ACTION
This office action is in response to the application filed on December 14, 2023.
Claims 1-11 are pending and have been examined. Claims 1-11 are rejected.
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 .
Priority
Applicants’ claim for the benefit of a prior-filed application under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, 365(c), or 386(c) is acknowledged. The present application claims foreign priority based on Chinese Patent Application No. 202110662048.9 filed June 15, 2021. The examiner notes that a certified copy (in Chinese) of the above-noted application was received on January 22, 2024.
Information Disclosure Statement
Acknowledgment is made of the information disclosure statements filed January 3, 2025 and January 10, 2025, which comply with 37 CFR 1.97. As such, the information disclosure statements have been placed in the application file and the information referred to therein has been considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. § 112(b):
(b) CONCLUSION – The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. § 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 2, which is dependent on Claim 1, is rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding Claim 2, the limitation "when sending the parameter of the first model to the server" does not clearly set the metes and bounds of the patent protection desired. There is insufficient antecedent basis for this limitation in the claim, rendering the claim indefinite because Claim 1 states “send a parameter of a first subnet in a first model to the server”. Therefore, the limitation Claim 2 is referencing is unclear.
Claim Rejections - 35 USC § 102
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless – (a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 3-7, 9, and 11 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by “Semi-Supervised Federated Learning for Activity Recognition” by Zhao et al., (non-patent literature published on November 2, 2020, hereinafter “Zhao”).
With respect to Claim 1: Zhao teaches:
“A model training system, comprising:
a server that maintains labeled data;” (Page 2, Section 1 “Introduction” recites a proposed design where a server uses a global autoencoder and supervised learning to transform a labeled dataset that it hosts (a server that maintains labeled data).)
“and a client that maintains unlabeled data,” (Page 5, Section 3.3 recites clients run unsupervised learning with their unlabeled local data (maintains unlabeled data).)
“wherein the client is configured to train a first model based on the unlabeled data, to obtain a parameter of the first model, (Page 5, Section 3.3 recites clients run unsupervised learning on a global autoencoder consisting of an encoder and decoder (train a first model) with their unlabeled local data (based on unlabeled data). The resulting local autoencoders (a parameter of the first model) are then sent back to the server where they came from.)
and send a parameter of a first subnet in the first model to the server, wherein the first model further comprises a second subnet;” (Page 5, Section 3.3 recites the outputted local autoencoders from the trained first model (a parameter of a first subnet in the first model) are then sent back to the server where they came from. Page 4, Figure 2 further clarifies that the structure of the global autoencoder consists of an encoder part (a first subnet) and a decoder part (a second subnet).)
“wherein the server is configured to train a second model based on the parameter of the first subnet reported by the client and the labeled data, (Page 5, Section 3.3 recites a server uses supervised learning to train the classifier using the encoder component of the global autoencoder (train a second model) that maps the encoded representation sent to the server from the client to activities with a labeled dataset (based on the parameter of the first subnet reported by the client and the labeled data).)
to update a parameter of the second model comprising the first subnet and a third subnet corresponding to the second subnet;” (Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where “CloudTraining” updates a classifier using the autoencoder (second model) and outputs an updated parameter of the second model. Page 2, Section 1 “Introduction” further clarifies that the server uses the encoder component (comprising the first subnet) of the global autoencoder to transform a labeled dataset into labeled representations with which a classifier (and a third subnet), that takes the place of the decoder (corresponding to the second subnet) as the task-specific output component, can be trained.)
“and the server is further configured to send an updated parameter of the first subnet and an updated parameter of the third subnet to the client;” (Page 5, Section 3.3 recites during each communication round between the client and server, the server sends the updated encoder (updated parameter of the first subnet) and the resulting classifier output (updated parameter of the third subnet) to selected clients.)
“and wherein the client is further configured to obtain a target model based on the parameter of the first subnet and the parameter of the third subnet that are from the server, wherein the target model comprises the first subnet and the third subnet.” (Page 5, Section 3.3 recites the client receives the updated global encoder (parameter of the first subnet) and updated global classifier (parameter of the third subnet) from the server and utilizes the encoder along with the classifier to perform inference. Under the broadest reasonable interpretation, it is understood that the client would thereby obtain a predictive model that performs inference (target model) comprised of the encoder and classifier (comprises the first subnet and the third subnet).)
With respect to Claim 3:
Zhao teaches:
“wherein the client is further configured to send a parameter other than the parameter of the
first subnet in the first model to the server.” (Page 5, Section 3.3 recites the outputted local autoencoders from the trained first model (a parameter of a first subnet in the first model) are then sent back to the server where they came from. Page 4, Figure 2 further clarifies that the structure of the autoencoders sent to the server consist of an encoder part (a first subnet) and a decoder part (a second subnet), akin to sending a parameter other than the parameter of the first subnet in the first model to the server.)
With respect to Claim 4:
Zhao teaches:
“wherein a quantity of clients is K, K is an integer greater than 1, (Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where during each communication round, there are “K” number of clients (a quantity of clients is K) and the server selects a set of clients (“St”) during each communication round (K is an integer greater than 1).)
and the server is further configured to perform aggregation processing on parameters of K first subnets from the K clients, to obtain a processed parameter of the first subnet;” (Page 6, Figure 6, Step 6 recites receiving locally trained autoencoders (models) from multiple clients and applying the “FedAvg” algorithm to the local autoencoder parameters in order to generate a new global autoencoder. Under the broadest reasonable interpretation, since each local autoencoder comprises an encoder corresponding to the first subnet, the aggregation of the local autoencoder parameters includes aggregation of the respective encoder parameters from the multiple clients, which produces processed encoder parameters in the global autoencoder.)
“and when training the second model of the server based on the parameter of the first subnet reported by the client and the labeled data, to update the parameter of the second model, the server is configured to train the second model of the server based on the processed parameter of the first subnet and the labeled data, to update the parameter of the second model” (Page 5, Section 3.3 recites a server uses supervised learning to train the classifier using the encoder component of the global autoencoder (train a second model) that maps the encoded representation sent to the server from the client to activities with a labeled dataset (based on the parameter of the first subnet reported by the client and the labeled data). Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where “CloudTraining” updates a classifier using the autoencoder (second model) and outputs an updated parameter of the second model.)
With respect to Claim 5:
Zhao teaches:
“wherein the third subnet of the second model is configured to output a calculation result of the second model, (Page 5, Section 3.3 recites a classifier (third subnet) outputting an activity recognition result (calculation result) of the second model (the encoder combined with the classifier).)
the second subnet of the first model is configured to output a calculation result of the first model, (Page 4, Figure 2 recites the network structure of the autoencoder containing an encoder and decoder (first model), where the encoder part compresses the input “X” into a representation “h” with fewer dimensions, while the decoder part (second subnet) generates a reconstruction “X’” from “h” (outputs a calculation result of the first model).)
and the third subnet of the second model has a different structure from the second subnet of the first model” (Page 5, Section 3.3 recites a classifier (third subnet) outputting an activity recognition result of the second model (the encoder combined with the classifier). Page 4, Figure 2 recites the network structure of the autoencoder containing an encoder and decoder (first model), where the encoder part compresses the input “X” into a representation “h” with fewer dimensions, while the decoder part (second subnet) generates a reconstruction “X’” from “h”. The third subnet and second subnet are separate neural network components of different models with different structures that produce different outputs, akin to having different structures.)
With respect to Claim 6:
Zhao teaches:
“A model training method applied to a server that maintains labeled data, comprising: training a second model based on a parameter of a first subnet reported by a client and the labeled data, (Page 5, Section 3.3 recites a server uses supervised learning to train the classifier using the encoder component of the global autoencoder (train a second model) that maps the encoded representation sent to the server from the client to activities with a labeled dataset (based on the parameter of the first subnet reported by the client and the labeled data).)
to update a parameter of the second model comprising the first subnet and a third subnet;” (Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where “CloudTraining” updates a classifier using the autoencoder (second model) and outputs an updated parameter of the second model. Page 2, Section 1 “Introduction” further clarifies that the server uses the encoder component (comprising the first subnet) of the global autoencoder to transform a labeled dataset into labeled representations with which a classifier (and a third subnet) can be trained.)
“and sending an updated parameter of the first subnet and an updated parameter of the third subnet to the client.” (Page 5, Section 3.3 recites during each communication round between the client and server, the server sends the updated encoder (updated parameter of the first subnet) and the resulting classifier output (updated parameter of the third subnet) to selected clients.)
With respect to Claim 7:
Zhao teaches:
“wherein a quantity of clients is K, K is an integer greater than 1, and the method further comprises: (Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where during each communication round, there are “K” number of clients (a quantity of clients is K) and the server selects a set of clients (“St”) during each communication round (K is an integer greater than 1).)
performing aggregation processing on parameters of K first subnets from the K clients, to obtain a processed parameter of the first subnet,” (Page 6, Figure 6, Step 6 recites receiving locally trained autoencoders (models) from multiple clients and applying the “FedAvg” algorithm to the local autoencoder parameters in order to generate a new global autoencoder. Under the broadest reasonable interpretation, since each local autoencoder comprises an encoder corresponding to the first subnet, the aggregation of the local autoencoder parameters includes aggregation of the respective encoder parameters from the multiple clients, which produces processed encoder parameters in the global autoencoder.)
“wherein the training the second model based on the parameter of the first subnet reported by the client and the labeled data, to update the parameter of the second model comprises: training the second model based on the processed parameter of the first subnet and the labeled data, to update the parameter of the second model” (Page 5, Section 3.3 recites a server uses supervised learning to train the classifier using the encoder component of the global autoencoder (train a second model) that maps the encoded representation sent to the server from the client to activities with a labeled dataset (based on the parameter of the first subnet reported by the client and the labeled data). Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where “CloudTraining” updates a classifier using the autoencoder (second model) and outputs an updated parameter of the second model.)
With respect to Claim 9:
Zhao teaches:
“A model training applied to a client that maintains unlabeled data, and the method comprising:
training a first model based on the unlabeled data, to obtain a parameter of the first model;” (Page 5, Section 3.3 recites clients run unsupervised learning on a global autoencoder consisting of an encoder and decoder (train a first model) with their unlabeled local data (based on unlabeled data). The resulting local autoencoders (a parameter of the first model) are then sent back to the server where they came from.)
“sending a parameter of a first subnet in the first model to a server, wherein the first model further comprises a second subnet;” (Page 5, Section 3.3 recites the outputted local autoencoders from the trained first model (a parameter of a first subnet in the first model) are then sent back to the server where they came from. Page 4, Figure 2 further clarifies that the structure of the global autoencoder consists of an encoder part (a first subnet) and a decoder part (a second subnet).)
“and obtaining a target model based on the parameter of the first subnet and a parameter of a third subnet from the server, wherein the target model comprises the first subnet and the third subnet corresponding to the second subnet.” (Page 5, Section 3.3 recites the client receives the updated global encoder (parameter of the first subnet) and updated global classifier (parameter of the third subnet) from the server and utilizes the encoder along with the classifier to perform inference. Under the broadest reasonable interpretation, it is understood that the client would thereby obtain a predictive model that performs inference (target model) comprised of the encoder and classifier, where the classifier takes the place of the decoder as the task-specific output component (comprises the first subnet and the third subnet corresponding to the second subnet).)
With respect to Claim 11:
Zhao teaches:
“sending a parameter other than the parameter of the first subnet in the first model to the
server” (Page 5, Section 3.3 recites the outputted local autoencoders from the trained first model (a parameter of a first subnet in the first model) are then sent back to the server where they came from. Page 4, Figure 2 further clarifies that the structure of the autoencoders sent to the server consist of an encoder part (a first subnet) and a decoder part (a second subnet), akin to sending a parameter other than the parameter of the first subnet in the first model to the server.)
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or non-obviousness.
Claim(s) 2 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over “Semi-Supervised Federated Learning for Activity Recognition” by Zhao et al., (non-patent literature published on November 2, 2020, hereinafter “Zhao”), in view of “Federated Learning with Personalization Layers” by Arivazhagan et al., (non-patent literature published on December 2, 2019, hereinafter “Arivazhagan”).
With respect to Claim 2:
Zhao does not appear to explicitly disclose:
“wherein when sending the parameter of the first model to the server, the client is configured to send only the parameter of the first subnet in the first model to the server”
However, Arivazhagan teaches:
“wherein when sending the parameter of the first model to the server, the client is configured to send only the parameter of the first subnet in the first model to the server” (Page 4, Section “Algorithm 2 FEDPER-SERVER” from Arivazhagan recites that only shared base-layer parameters (parameters of a first subnet) are sent to a server by the client. Page 2, “Figure 1” under Section 1 “Introduction” further supports that only base layers are shared by the client with the parameter server while the personalization layers are kept private by each device.)
It would have been obvious to a PHOSITA before the effective filing date of the present application to implement a claim that utilized the teachings of Zhao with the teachings of Arivazhagan, which are both in the same field of invention. A PHOSITA would have been motivated to modify the communication protocol used in the semi-supervised federated learning system from Zhao to implement the selective sharing of parameters from Arivazhagan. Arivazaghan’s “FedPer” communication strategy explains how transmitting only shared subnetwork parameters while retaining remaining parameters locally reduces communication overhead. This would predictably improve the efficiency of the parameter exchange if combined with Zhao’s system.
With respect to Claim 10:
Zhao does not appear to explicitly disclose:
“wherein the client sends only the parameter of the first subnet in the first model to the server, and does not send a parameter other than the parameter of the first subnet in the first model to the server”
However, Arivazhagan teaches:
“wherein the client sends only the parameter of the first subnet in the first model to the server, and does not send a parameter other than the parameter of the first subnet in the first model to the server” (Page 4, Section “Algorithm 2 FEDPER-SERVER” from Arivazhagan recites that only shared base-layer parameters (parameters of a first subnet) are sent to a server by the client. Page 2, “Figure 1” under Section 1 “Introduction” further supports that only base layers are shared by the client with the parameter server while the personalization layers are kept private by each device.)
It would have been obvious to a PHOSITA before the effective filing date of the present application to implement a claim that utilized the teachings of Zhao with the teachings of Arivazhagan, which are both in the same field of invention. A PHOSITA would have been motivated to modify the communication protocol used in the semi-supervised federated learning system from Zhao to implement the selective sharing of parameters from Arivazhagan. Arivazaghan’s “FedPer” communication strategy explains how transmitting only shared subnetwork parameters while retaining remaining parameters locally reduces communication overhead. This would predictably improve the efficiency of the parameter exchange if combined with Zhao’s system.
Claim(s) 8 is rejected under 35 U.S.C. 103 as being unpatentable over “Semi-Supervised Federated Learning for Activity Recognition” by Zhao et al., (non-patent literature published on November 2, 2020, hereinafter “Zhao”), in view of “Ensemble Distillation for Robust Model Fusion in Federated Learning” by Lin et al., (non-patent literature published in December 2020, hereinafter “Lin”).
With respect to Claim 8:
Zhao teaches:
“wherein the server further maintains unlabeled data,
and the training the second model based on the parameter of the first subnet reported by the client and the labeled data, to update the parameter of the second model comprises: training a third model based on the parameter of the first subnet reported by the client and the unlabeled data, to update a parameter of the third model;” (Page 5, Section 3.3 recites a server uses supervised learning to train the classifier using the encoder component of the global autoencoder (train a second model) that maps the encoded representation sent to the server from the client to activities with a labeled dataset (based on the parameter of the first subnet reported by the client and the labeled data).)
“and training the second model based on the parameter of the third model and the labeled data, to update the parameter of the second model.” (Page 5, Section 3.3 recites a server uses supervised learning to train the classifier using the encoder component of the global autoencoder (train a second model) that maps the encoded representation sent to the server from the client to activities with a labeled dataset. Page 6, Figure 6 recites the semi-supervised federated learning algorithm, where “CloudTraining” updates a classifier using the autoencoder (second model) and outputs an updated parameter of the second model.)
Zhao does not appear to explicitly disclose:
“wherein the server further maintains unlabeled data,
and the training the second model based on the parameter of the first subnet reported by the client and the labeled data, to update the parameter of the second model comprises: training a third model based on the parameter of the first subnet reported by the client and the unlabeled data, to update a parameter of the third model;”
“and training the second model based on the parameter of the third model and the labeled data, to update the parameter of the second model.”
However, Lin teaches:
“wherein the server further maintains unlabeled data, (Page 3, Section 3 “Ensemble distillation” recites unlabeled data maintained on a server.)
and the training the second model based on the parameter of the first subnet reported by the client and the labeled data, to update the parameter of the second model comprises: training a third model based on the parameter of the first subnet reported by the client and the unlabeled data, to update a parameter of the third model;” (Page 3, Section 3, “Algorithm 1” recites the “FedDF” algorithm where a server student model (third model) is trained using client model parameters and unlabeled data, where the weights of the server student model are updated (update a parameter of the third model) before being used in the overall server-side learning pipeline.)
“and training the second model based on the parameter of the third model and the labeled data, to update the parameter of the second model.” (Page 3, Section 3, “Algorithm 1” recites the “FedDF” algorithm where a server student model (third model) is trained using client model parameters and unlabeled data, where the weights of the server student model are updated (parameter of the third model) before being used in the overall server-side learning pipeline.)
It would have been obvious to a PHOSITA before the effective filing date of the present application to implement Claim 8 that utilized the teachings of Zhao with the teachings of Lin, which are both in the same field of invention. A PHOSITA would have been motivated to incorporate the “FedDF” server-side distillation technique from Lin into the federated semi-supervised learning system from Zhao because Lin discusses how FedDF improves the quality and robustness of the learned feature representations before supervised training is performed afterwards. This would in turn improve the accuracy of Zhao’s server-side classifier by leveraged additional unlabeled data available at the server from Zhao.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Vibha Bhat whose telephone number is (571)-272-7091. The examiner can normally be reached on Monday – Thursday from 8:00 AM to 5:00 PM EST and every other Friday from 8:00 AM to 4:00 PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. See MPEP § 713.01. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at https://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mariela Reyes, can be reached at telephone number (571)-270-1006. 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 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://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 (572)-272-1000.
/Vibha Bhat/Examiner
Art Unit 2142
/Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights