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 .
Detailed Action
The following action is in response to the communication(s) received on 10/22/2025.
As of the claims filed 10/22/2025:
Claims 1, 9, 17, and 23 have been amended.
Claims 1-6, 9-14, 17-19, and 21-28 are now pending.
Claims 1, 9, 17, and 23 are independent claims.
Response to Arguments
Applicant’s arguments filed 10/22/2025 have been fully considered, but are not fully persuasive.
Regarding the rejection of 35 USC § 112(b):
Applicant agrees the claim elements should be read in light of the Specification, and thus the 35 USC § 112(b) indefiniteness rejection has been withdrawn.
Regarding the 35 USC § 101 rejection, the amendments to the claims have overcome the rejection, and thus the rejection has been withdrawn.
Regarding the prior art rejections, Applicant asserts that Lin or Yang does not teach “transmitting the training loss, via digital transmission on digital resources, to a federated learning server ... and transmitting the weight updates, via analog communication, to the federated learning server.” Examiner respectfully disagrees, as “transmitting the training loss, via digital transmission on digital resources, to a federated learning server” is taught by Lin [p.9 last ¶],[p.8 Remark 2, the transmitted “local-model” contains the training loss] via Yang/Lin, and “transmitting the weight updates, via analog communication, to the federated learning server” us taught by Yang [p.4 left ¶3], via Yang/Lin.
Claims 2-6, 10-14, 18-22, and 24-28 are rejected by virtue of dependency to their respective parent claims which are rejected for reasons stated above.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Such claim limitation(s) is/are:
In claim 23: “means for computing updates…”, “means for recording a training loss…”, “means for transmitting the training loss…”, “means for receiving…”, “means for calculating…”, “and means for transmitting the updates…”.
In claim 26: “means for receiving… a configuration”, “means for scaling the updates…”
In claim 27: “means for transmitting the training loss…”
In claim 28: “means for transmitting the training loss…”
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
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 nonobviousness.
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.
Claims 1-6, 9-14, and 17-28 are rejected under 35 U.S.C. 103 as being unpatentable over Yang et al., “Federated Learning via Over-the-Air Computation” (hereinafter Yang) in view of Lin et al., “Deploying Federated Learning in Large-Scale Cellular Networks: Spatial Convergence Analysis” (hereinafter Lin).
Regarding Claim 1, Yang teaches:
A method of wireless communication by a user equipment (UE), comprising: (Yang [Abstract] This stimulates a nascent field termed as federated learning for training a machine learning model on computation, storage, energy and bandwidth limited mobile devices in a distributed manner)
computing initial updates to an artificial neural network as part of an epoch of a federated learning process, the initial updates comprising gradients or updated model parameters; (Yang [p.3 left ¶5] This model is widely used in… as deep neural networks.
[p.3 right ¶1]
PNG
media_image1.png
436
476
media_image1.png
Greyscale
) (Note: the updated global model z[t] corresponds to the updated model parameters)
recording a training loss observed while training the artificial neural network at the epoch of the federated learning process, the training loss corresponding to UE heterogeneity; (Yang [p.3 left ¶5]
PNG
media_image2.png
477
576
media_image2.png
Greyscale
[p.3 r ¶1]
PNG
media_image3.png
337
580
media_image3.png
Greyscale
(Note: l(z;xj,yj) of each function using the stochastic gradient algorithm corresponds to the training loss. The t-th round corresponds to the epoch of the federated learning process; the aggregation of all the updates of the selected devices corresponds to the UE heterogeneity.)
receiving, from the federated learning server, a configuration of a quantity of training samples, based on the training loss; (Yang [p. 3 r. ¶1]
PNG
media_image4.png
325
537
media_image4.png
Greyscale
) (Note: the updated global model z[t-1] corresponds to the configuration of a quantity of training samples, since the previous round also comprises of aggregated weighted averages of the local models. The model being trained using stochastic gradient algorithm corresponds to being based on the training loss. It should also be noted that, although Applicant asserts claim 1 incorporates parts of claim 8, the claim language is recited differently such that the interpretation allows Yang to continue teaching this limitation.)
calculating weight updates based on the quantity of training samples; (Yang [p. 3 r. ¶1]
PNG
media_image5.png
332
518
media_image5.png
Greyscale
) (Note: running the local update algorithm corresponds to calculating the weight updates; each resulting updated local model zk[t] corresponds to each weight update)
and transmitting the weight updates, via analog communication, to the federated learning server that is configured to aggregate the gradients from a plurality of UEs in accordance with weights determined based on the training loss, (Yang [p.3 r]
[AltContent: textbox (Plurality of user equipment)][AltContent: textbox (Federated learning server)]
PNG
media_image6.png
446
598
media_image6.png
Greyscale
(Note: the arrow from device 1 to device m to the Base Station corresponds to transmitting the (local) updates to a federated learning server)
the transmitting the weight updates comprising the analog communication via shared uplink resources that are orthogonal to the digital resources, (Yang [p.4 left ¶3] Our approach is based on the principles of over-the-air computation by leveraging the signal superposition property of a multiple-access channel. The key observation in the FedAvg algorithm is that the global model is updated through computing the weighted average of locally computed updates at each selected device, which falls in the category of computing nomographic functions of distributed data) (Note: over-the-air computation and signal superposition corresponds to analog communication; analog communication methods are different from digital communication methods and thus are orthogonal to the digital resources; each selected device correspond to each shared uplink resource)
the weight updates being based on the training loss. (Yang [p.3 r]
[AltContent: textbox (Plurality of user equipment)][AltContent: textbox (Federated learning server)]
PNG
media_image6.png
446
598
media_image6.png
Greyscale
[p. 3 r. ¶1]
PNG
media_image5.png
332
518
media_image5.png
Greyscale
) (Note: each zi[t] calculates the training loss of each device via the stochastic gradient algorithm. Thus, each zi[t] corresponds to the weight updates based on the training loss)
Yang does not teach, but Lin further teaches:
transmitting the training loss, via digital transmission on digital resources, to a federated learning server during each round of the federated learning process (Lin [p.9 last ¶] 1) Digital Transmission: For digital transmission, each coefficient of the local gradient at each device is quantized into a sufficiently large number of bits, denoted as D, such that the effect of quantization errors on learning performance is negligible. Then the quantized gradient is encoded and transmitted at the fixed rate B/M log(1 + θ) with θ being a chosen constant.
[p.8 Remark 2] Remark 2 (Extension to Local-model Uploading). The current analysis can be extended to the alternative FEEL implementation with local-model uploading by accounting for multi-round local-gradient descent [3]. First, in each round, the local model at device X is updated via w˜ (n+1) X = w (n) X −µg˜ (n) X ; then w˜X (n+1) is transmitted to the server for updating the global model: w (n+1) X = 1 A(n) P X∈C0∩Φ (n) d w˜ (n+1) X . The analysis can be modified accordingly and the modification is straightforward and does not change the findings.) (Note: the local-model transmitted to the server (through the straightforward modification) contains the weights updated by the training loss. Thus, digital transmission of the model corresponds to the transmission of the training loss)
Yang and Lin are analogous to the present invention because both are from the same field of endeavor of federated learning through sending updated models to servers. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to implement the digital transmission taught by Lin to Yang’s distributed learning method. The motivation would be that “When the network is relatively sparse…digital transmission is observed to outperform the analog scheme as the latter exposes uncoded signals to the perturbation of inter-cell interference.” (Lin [p.23 ¶1])
Regarding Claim 2, Yang, via Yang/Lin, further teaches:
The method of claim 1, in which the weight updates are scaled for aggregation based on a previous value of the training loss. (Yang [p.4 r]
PNG
media_image7.png
218
443
media_image7.png
Greyscale
[p. 3 r. ¶1] To reduce the number of communication rounds for global model aggregation, the federated averaging (FedAvg) algorithm [12] has recently been proposed. Specifically, at the t -th round:
The BS selects a subset of mobile devices St⊆{1,⋯,M} ;
The BS sends the updated global model z[t−1] to the selected devices St ;
Each selected device i∈St runs a local update algorithm (e.g., stochastic gradient algorithm) based on its local dataset Di and the global model z[t−1] , whose output is the updated local model z[t]i ;
The BS aggregates all the local updates z[t]i with i∈St , i.e., computing their weighted average as the updated global model z[t] .)
(Note: computing the weighted average corresponds to scaling for aggregation. The BS sending the global model z[t-1] corresponds to the previous value of the training loss.)
Regarding Claim 3, Yang, via Yang/Lin, further teaches:
The method of claim 1, in which the weight updates are scaled for aggregation based on a function of the training loss. (Yang [p.4 r]
PNG
media_image7.png
218
443
media_image7.png
Greyscale
[p.3 left ¶5]
PNG
media_image2.png
477
576
media_image2.png
Greyscale
[p. 3 r. ¶1]
PNG
media_image3.png
337
580
media_image3.png
Greyscale
) (Note: z is the global model; zi is the local model. Step 3 states the local update algorithm; 4 corresponds to the aggregation of the updates. The normalization of the local model (in eq. 2) corresponds to scaling based on the function of the training loss.)
Regarding Claim 4, Yang, via Yang/Lin, further teaches:
The method of claim 1, further comprising: receiving, from the federated learning server, a configuration for scaling the weight updates; and scaling the weight updates based on the training loss, prior to transmitting the weight updates to the federated learning server. (Yang [p. 3 r. ¶1]
PNG
media_image3.png
337
580
media_image3.png
Greyscale
[p.4 r]
PNG
media_image7.png
218
443
media_image7.png
Greyscale
(Note: |Di| means the size of the local dataset; the FedAvg algorithm including the normalization of the updated local model corresponds to receiving the configuration for scaling the updates. Equation 2 is performed in step 2 of the federated averaging algorithm, thus corresponding to scaling the updates prior to transmitting the updates to the federated learning server)
Regarding Claim 5, Yang, via Yang/Lin, further teaches:
The method of claim 1, further comprising transmitting the training loss to the federated learning server during each round of the federated learning process. (Yang [p.3 r ¶1] To reduce the number of communication rounds for global model aggregation, the federated averaging (FedAvg) algorithm [10] has recently been proposed. Specifically, at the t -th round:
PNG
media_image3.png
337
580
media_image3.png
Greyscale
PNG
media_image8.png
483
623
media_image8.png
Greyscale
PNG
media_image7.png
218
443
media_image7.png
Greyscale
PNG
media_image9.png
813
414
media_image9.png
Greyscale
)
(Note: the value of t includes 1 round, thus t-th round corresponds to each round of the federated learning process. Figure 2 shows that the devices are transmitting a training loss to the federated learning server at each round.)
Regarding Claim 6, Yang teaches:
The method of claim 1, further comprising transmitting the training loss to the federated learning server every N rounds of the federated learning process, where N is greater than or equal to two. (Yang [p.3 left ¶5] To reduce the number of communication rounds for global model aggregation, the federated averaging (FedAvg) algorithm [12] has recently been proposed. Specifically, at the t -th round:
PNG
media_image3.png
337
580
media_image3.png
Greyscale
.
PNG
media_image9.png
813
414
media_image9.png
Greyscale
) (Note: since the federated averaging algorithm transmits the loss every t-th round, then the algorithm also transmits the loss every second round, thus corresponding to each N rounds of the federated learning process.)
[AltContent: textbox (The apparatus for wireless communication comprises of a server, which has memory and at least one processor)]Independent Claim 9 recites An apparatus for wireless communication by a user equipment (UE), comprising: a memory; and at least one processor coupled to the memory, the at least one processor configured (Yang
PNG
media_image6.png
446
598
media_image6.png
Greyscale
) to perform precisely the methods of Claim 1. Thus, Claim 9 is rejected for reasons set forth in Claim 1.
Claims 10-14, dependent on Claim 9, also recite the apparatus configured to perform precisely the methods of Claims 2-6, respectively. Thus, Claims 10-14 are rejected for reasons set forth in Claims 2-6, respectively.
Independent Claim 17 recites A non-transitory computer-readable medium having program code recorded thereon, the program code comprising (Yang [AltContent: textbox (The base station which executes the federated learning program code comprises of a server, which has memory and at least one processor)]
PNG
media_image6.png
446
598
media_image6.png
Greyscale
) precisely the methods of Claim 1. Thus, Claim 17 is rejected for reasons set forth in Claim 1.
Claims 18, 19, 21, and 22, dependent on Claim 17, also recite the system configured to perform precisely the methods of Claims 2, 3, 5, and 6, respectively. Thus, Claims 18, 19, 21, and 22 are rejected for reasons set forth in Claims 2, 3, 5, and 6, respectively.
Independent Claim 23 recites An apparatus for wireless communication by a user equipment (UE), comprising: means for (Yang [Abstract] We thus propose a novel over-the-air computation …) to perform precisely the methods of Claim 1. Thus, Claim 23 is rejected for reasons set forth in Claim 1. (Note: over-the-air corresponds to wireless communication; computation requires a processor and memory, thus corresponding to an apparatus)
Claims 24-28, dependent on Claim 23, also recite the apparatus configured to perform precisely the methods of Claims 2-6, respectively. Thus, Claims 24-28 are rejected for reasons set forth in Claims 2-6, respectively.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 nonprovisional extension fee (37 CFR 1.17(a)) 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 JOSEP HAN whose telephone number is (703)756-1346. The examiner can normally be reached Mon-Fri 9am-5pm.
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, Kakali Chaki can be reached on (571) 272-3719. 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.
/J.H./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights