NON-FINAL REJECTION, THIRD DETAILED ACTION
Status of Prosecution
The present application 17/977,736, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
The application was filed in the Office on Oct. 31, 2022 and claims priority to Chinese patent application CN202110084293.6 with a filing date of January 21, 2021.
The Office mailed a first detailed action, non-final rejection on Sept. 30, 2025.
Applicant’s representative initiated an interview on Oct. 25, 2025 and subsequently filed amendments with remarks and arguments on Dec. 30, 2025.
The Office mailed a second detailed action, final rejection on Jan. 23, 2026.
Applicant filed an after final response asking for reconsideration with amended claims on March 20, 2026 and the Office mailed an advisory action on April 10, 2026.
Applicant filed a request for continued examination on April 16, 2026.
Claims 1-20 are pending and are all rejected in this rejection. Claims 1, 12 and 20 are independent claims.
Status of Claims
Claim 10 is rejected as being indefinite under 35 USC § 112(b).
Claims 1, 2, 6, 11-13, 17 and 20 are rejected under 35 USC § 103 as being unpatentable over Angel et al. (“Angel”), United States Patent Application Publication 2021/0174243 A1, published on June 10, 2021 in view of non-patent literature Yang et al. (“Yang”), “The Tradeoff Between Privacy and Accuracy in Anomaly Detection Using Federated XGBoost, published on Oct. 14, 2019 and in further view of Ji et al. (“Ji”), Chinese Patent Application Publication CN111784772A, published on Oct. 16, 2020.
Claims 3, 5 and 14-16 are rejected under 35 USC § 103 as being unpatentable over Angel in view of Yang in view of Ji and in further view of non-patent literature J. Yang et al. (“J. Yang”), “Knowledge-based systems,” published December 31, 2010.
Claims 4 and 15 are rejected under 35 USC § 103 as being unpatentable over Angel in view of Yang in view of and in further view of Ulm et al. (“Ulm”), “Functional Federated Learning in Erlang (ffl-erl)” published in 2019.
Claims 7-9 and 18-19 are rejected under 35 USC § 103 as being unpatentable over Angel in view of Yang in view of Ji and in further view of non-patent literature Fu et al. (“Fu”), “Attack-Resistant Federated Learning with Residual-based Reweighting,” published on January 8, 2021.
Request for Continued Examination
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 March 20, 2026 have been entered.
Response to Remarks and Arguments
Examiner thanks Applicant for the amended claims and the remarks and arguments..
First, regarding the 112(b) rejection of claim 10, it is traversed as the amendments and arguments are persuasive.
As to the prior art rejections, Examiner has adjusted the rejections accordingly as noted below and adjusted the grounds of rejection as well. Examiner has introduced Ji et al. (“Ji”), Chinese Patent Application Publication CN111784772A, published on Oct. 16, 2020 to address the amended claim feature.
The claims stand rejected.
Allowable Subject Matter
Claim 10 is 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
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.
A.
Claims 1, 2, 6, 11-13, 17 and 20 are rejected under 35 USC § 103 as being unpatentable over Angel et al. (“Angel”), United States Patent Application Publication 2021/0174243 A1, published on June 10, 2021 in view of non-patent literature Yang et al. (“Yang”), “The Tradeoff Between Privacy and Accuracy in Anomaly Detection Using Federated XGBoost, published on Oct. 14, 2019 and in further view of Ji et al. (“Ji”), Chinese Patent Application Publication CN111784772A, published on Oct. 16, 2020.
As to Claim 1, Angel teaches: A federated model training method, which is executed by an electronic device and comprises:
acquiring a first sample set associated with a first device in a service system comprising one or more real samples associated with the first device, and a second sample set associated with a second device in the service system, wherein the service system comprises at least the first device and the second device (Angel: Fig. 1, par. 0032, participants [108] (i.e. devices) have data sets associated with each one (i.e. respective sample sets) comprising the system [100] )i.e. a service system; the data sets are “real data”));
determining a sample set intersection based on the first sample set and the second sample set (Angel: par. 0002, vertical federated learning considers overlapping sample spaces on vertically partitioned datasets);
determining a first key set associated with the first device and a second key set associated with the second device (Angel: par. 0033, Fig. 1, a third party authority [102] manages a cryptosystem, generating and delivering the public keys to participants [108]);
obtaining a training sample associated with the service system based on the sample set intersection, the first key set, and the second key set (Angel: par. 0033, a collective dataset (i.e. a training sample) is created from the collective data set (sample set intersection) per the successful use of the encryption ecosystem); and
training a federated model corresponding to the service system based on the training sample (Angel: par. 0033, the model is trained per the collective training dataset).
PNG
media_image1.png
485
699
media_image1.png
Greyscale
Angel may not explicitly teach: determining one or more virtual samples associated with the first device based on the first sample set;
determining a sample set intersection based on a combination of the first sample set, the one or more virtual samples associated with the first device and the second sample set.
Yang teaches in general concepts related to a proposed horizontal federated XGBoost algorithm (Yang: Abstract). Specifically, Yang teaches that virtual data samples are used in a federated learning scheme (Yang: Sec. 4.2, the Federated XGBoost considers the cluster of samples to create a new data sample and are then transmitted). The virtual sample may be result of aggregating the original data (i.e. combining the “real data”) to create the virtual sample (Yang: Sec. 1).
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Angel disclosures and teachings by incorporating the use of virtual data samples in the first data set as taught and suggested by Yang. Such a person would have been motivated to do so with a reasonable expectation of success to allow for the better control of the tradeoff between learning performance and the privacy of the system (Yang: Sec. 4.2).
Angel, Yang and Ji may not explicitly teach: training a federated model corresponding to the service system based on a modification of one or more gradient multipliers that correspond to the tone or more virtual samples included in the training sample.
Ji teaches in general concepts related to domain randomization-based attitude estimation model training (Ji: Abstract). Specifically, Ji discusses synthetic (i.e. virtual) samples are processed separately from real samples of the attitude (pose) of individuals. Loss functions are calculated for each model parameter, which are associated with the different samples and backpropagation to update the model as a whole (Ji: p. 7). Examiner notes that this would be an iterative process and the gradient multipliers would be modified iteratively as well.
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Angel-Yang-Ji disclosures and teachings by allowing for the training of the federated model as a whole based on the changing gradient multiplier associated with the synthetic samples from the backpropagation algorithm as taught by Ji. Such a person would have been motivated to do so with a reasonable expectation of success to allow for the overall training of the federated model with respect to the virtual samples in a segregated fashion.
As to Claim 2, Angel, Yang and Ji teach the elements of claim 1.
Angel further teaches: wherein acquiring the first sample set and the second sample set comprises:
determining, based on a service type of the first device, a first service type sample set corresponding to the first device;
determining, based on a service type of the second device, a second service type sample set corresponding to the second device (Angel: par. 0033, the collective dataset may be partitioned vertically, and thus based on the device, or an associated participant, a partitioned sample set is determined); and
aligning the first service type sample set to be associated with the first device and aligning the second service type sample set to align with the second device (Examiner notes that the partitioning process would therefore align the sample sets).
As to Claim 6, Angel, Yang and Ji teach the elements of claim 1.
Angel further teaches: wherein obtaining the training sample comprises:
performing, based on the first key set and the second key set, an exchange operation between a first public key of the first device and a second public key of the second device to obtain an initial parameter of the federated model (Angel: par. 0035, the TPA [102] can facilitate the exchanging of keys between the devices for obtaining information (i.e. initial parameter);
determining a number of samples that match the service system; and obtaining the training sample based on the sample set intersection, the number of samples, and the initial parameter (Angel: par. 0041, the number of input parties (i.e. number of samples that match the service system, and others are used to ultimately create the training sample).
As to Claim 11, Angel, Yang and Ji teach the elements of claim 1.
Angel further teaches: wherein the method further comprises:
transmitting at least one of the one or more virtual samples, the sample set intersection, the first key set, the second key set, and a federated model parameter to a server (Angel: par. 0125, each of the devices and processes may reside on servers, such as the TPA); and
acquiring, by at least one of the first device or the second device, at least one of the one or more virtual samples, the sample set intersection, the first key set, the second key set, and the federated model parameter from the server while performing service processing (Angel: pars. 0125-27, the different sets and parameters may all be performed while the information is being transmitted in the networking environment).
As to Claim 12, it is rejected for similar reasons as claim 1. Angel further teaches a processor and memory and computer readable storage media (Angel: pars. 0113, 30).
As to Claim 13, it is rejected for similar reasons as claim 2.
As to Claim 17, it is rejected for similar reasons as claim 6.
As to Claim 20, it is rejected for similar reasons as claim 1 and 12.
B.
Claims 3, 5 and 14-16 are rejected under 35 USC § 103 as being unpatentable over Angel et al. (“Angel”), United States Patent Application Publication 2021/0174243 A1, published on June 10, 2021 in view of non-patent literature Yang et al. (“Yang”), “The Tradeoff Between Privacy and Accuracy in Anomaly Detection Using Federated XGBoost, published on Oct. 14, 2019 in view of Ji et al. (“Ji”), Chinese Patent Application Publication CN111784772A, published on Oct. 16, 2020 and in further view of non-patent literature J. Yang et al. (“J. Yang”), “Knowledge-based systems,” published December 31, 2010.
As to Claim 3, Angel, Yang and Ji teach the elements of claim 1.
Angel, Yang and Ji may not explicitly teaches: wherein determining the one or more virtual samples comprises:
determining a repective value parameter and a respective distribution parameter of a respective sample ID in the first sample set; and
generating a virtual sample based on the respective value parameter and the respectibe distribution parameter of the respective sample ID in the first sample set.
J. Yang teaches in general concepts related to virtual sample generation based on Gaussian distributions (J. Yang: Abstract). Specifically Yang teaches that a Gaussian distribution and value parameters are determined for generating a virtual sample around an original sample (J. Yang: Sec. 3.3, the mean of a Gaussian distribution and the standard error for the distribution is determined based on values of the original sample).
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Angel-Yang-Ji disclosures and teachings by generating the virtual sample with the consideration of distribution and value parameters with the original sample as associated with a particular sample ID as taught and suggested by J. Yang . Such a person would have been motivated to do so with a reasonable expectation of success to allow for the consideration of a population with a large number of factors will follow such a distribution allowing for better sampling.
As to Claim 5, Angel, Yang, Ji and J. Yang teach the elements of claim 3.
Angel, Yang, Ji and J. Yang as combined further teaches: wherein determining the sample set intersection comprises: combining the one or more virtual samples with the first sample set to obtain the first sample set including the virtual sample (Examiner notes that the use of a virtual sample as disclosed in Yang is for the use of providing additional data points for the sample and thus is combined);
traversing the first sample set including the one or more virtual samples to obtain an ID set of the one or more virtual samples; and
traversing the first sample set including the one or more virtual samples and the second sample set to obtain the sample set intersection of the first sample set including the one or more virtual samples and the second sample set (Examiner asserts that the disclosed vertical federal learning partitioning as disclosed would traverse the sample sets to obtain the ID set and the sample set intersection as already claimed in the parent claims).
As to Claim 14, it is rejected for similar reasons as claim 3.
As to Claim 16, it is rejected for similar reasons as claim 5.
C.
Claims 4 and 15 are rejected under 35 USC § 103 as being unpatentable over Angel et al. (“Angel”), United States Patent Application Publication 2021/0174243 A1, published on June 10, 2021 in view of non-patent literature Yang et al. (“Yang”), “The Tradeoff Between Privacy and Accuracy in Anomaly Detection Using Federated XGBoost,” published on Oct. 14, 2019 in view of Ji et al. (“Ji”), Chinese Patent Application Publication CN111784772A, published on Oct. 16, 2020 and in further view of Ulm et al. (“Ulm”), “Functional Federated Learning in Erlang (ffl-erl)” published in 2019.
As to Claim 4, Angel, Yang and Ji teach the elements of claim 1.
Angel, Yang and Ji may not explicitly teach: wherein determining the one or more virtual samples comprises: determining, based on a first device type of the first device and a second device type of the second device, a process identifier of a target application process;
determining a data intersection set of the first sample set and the second sample set;
obtaining a first virtual sample set corresponding to the first device and a second virtual sample set corresponding to the second device based on invoking the target application process, wherein the first virtual sample set and the second virtual sample set are output by the target application process; and
obtaining the one or more virtual samples based on the data intersection set, the first virtual sample set, and the second virtual sample set, wherein the one or more virtual samples are output by the target application process.
Ulm teaches in general concepts related to framework for using the functional programming language Erlang for implementing federated learning (Ulm: Abstract). Specifically, Ulm teaches that process identifiers may be called in a decentralized manner for collecting different samples from different devices and sources (Ulm: pp. 166-167, the server process shown in Code Listing 1.1 a process identifier for a current process is used to identify the current process as values are received). This information could then be used for the training of the model in federated learning fashion.
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Angel-Yang-Ji disclosures and teachings by performing the virtual sample determination by implementing it via the use of target process identifiers as taught and suggested by Ulm. Such a person would have been motivated to do so with a reasonable expectation of success to allow for the use of a decentralized framework as disclosed by Ulm.
As to Claim 15, it is rejected for similar reasons as claim 4.
D.
Claims 7-9 and 18-19 are rejected under 35 USC § 103 as being unpatentable over Angel in view of Yang in view of Ji and in further view of non-patent literature Fu et al. (“Fu”), “Attack-Resistant Federated Learning with Residual-based Reweighting,” published on January 8, 2021.
As to Claim 7, Angel, Yang and Ji teach the elements of claim 1.
Angel, Yang and Ji may not explicitly teach: wherein training the federated model comprises: substituting the training sample into a loss function corresponding to the federated model;
determining a model updating parameter of the federated model based on the loss function satisfying a convergence condition; and
determining, based on the model updating parameter of the federated model, a federated model parameter of the federated model.
Fu teaches in general concepts related to residual-base reweighting to defend against attacks in federated learning spaces (Fu: Abstract). Specifically, Fu teaches that a loss function may have as a parameter training data points sampled from a distribution (Fu: p. 7, Theoretical Analysis). Minimization of the loss function is considered for the federal model (Fu: p. 7, Theoretical analysis, parameter space seeks to minimize the population loss).
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Angel-Yang-Ji disclosures and teachings by utilizing a loss function as taught by Fu. Such a person would have been motivated to do so with a reasonable expectation of success to allow for the optimizing of the federated model.
As to Claim 8, Angel, Yang, Ji and Fu teach the elements of claim 7.
Fu further teaches: wherein the method further comprises:
adjusting, by the first device, a residual corresponding to the one or more virtual samples corresponding to the model updating parameter (Fu: p. 2, Our Algorithm, a vertical distance to a robust regression line (residual) is used for adjusting a parameter confidence for each local model), or a degree of impact of the one or more virtual samples on the federated model parameter of the federated model, based on the federated model corresponding to the service system being trained based on the training sample that matches the service system.
As to Claim 9, Angel, Yang and Ji teach the elements of claim 8.
Angel, Yang and Fu as combined further teaches: wherein the method further comprises:
triggering, based on the federated model corresponding to the service system being trained based on the training sample, a target application process to: adjust the residual corresponding to the one or more virtual samples (Fu: p. 2, Our Algorithm, the parameters are reweighted per the residuals), or the degree of impact of the one or more virtual samples on the federated model parameter of the federated model.
As to Claim 18, it is rejected for similar reasons as claim 7.
As to Claim 19, it is rejected for similar reasons as claim 8.
Conclusion
Prior art made of the record:
Gao et al., US PG Pub 2022/0254472 (Aug. 4, 2022) (describing vertical federated learning with training on different subsets);
Wu et al., Chinese Patent Application Publication CN110837653A (Feb. 25, 2020) (describing label prediction model).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JAMES T TSAI whose telephone number is (571)270-3916. The examiner can normally be reached M-F 8-5 Eastern.
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, 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 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./JAMES T TSAI/
/JAMES T TSAI/ Primary Examiner, Art Unit 2147 wa
Prosecution Insights