DETAILED ACTION
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 .
Claim Rejections - 35 USC § 103
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.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-5 and 8-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Yang et al. US 2023/0038310 in view of Luo, Bing, et al. "Cost-effective federated learning in mobile edge networks."
Regarding claims 1 and 18, Yang teaches “a computer-implemented method for training a global model on a central server in a federated learning system having a plurality of nodes” (abstract “The client computing device is configured to receive a model comprising a set of common layers and a set of client-specific layers from the server computing device. After a training at the client computing device, the set of common layers and the set of client-specific layers are both updated”), the method comprising: splitting the global model along a width and a depth via two-dimensional uniform downscaling of the global model” ([0105] “As stated above (and as shown on the left-hand side of FIG. 3), a contribution of this embodiment is the virtual separation of the model 100 of the neural network - here it is exemplarily a CNN network—into a set of common layers 120 and a set of client-specific layers 140. The way of separating the model 100 may be performed according to the CNN's property. Here, in this embodiment, the set of common layers 120 is referred to as “Backbone”, e.g., stacked convolutional layers, and the set of client-specific layers 140 is referred to as last layers (LL), e.g., last fully connected layers. In particular, the CNN may be a common classification network using stacked convolutional layers at the beginning, followed by fully connected layers. The LL may also be referred to as “LL Classifier”, since it/they is/are the classifier that contains class specific information” this is interpreted as two-dimensional uniform downscaling since the model is split based on width and depth, or node/layer);
“creating a plurality of local models based on the splitting of the global model” ([0105] “The LL may also be referred to as “LL Classifier”, since it/they is/are the classifier that contains class specific information.” LL classifier or model); and
“training selected ones of the plurality of local models on respective selected ones of a plurality of clients […]” ([0029] “By separating the model into the set of common layers and the set of client-specific layers, the server computing device is able to contribute to the improved training of the model in a collaborative manner with one or more client computing devices. Thus, a global accuracy of the model can be assured, while a local accuracy is also improved. Further, generalization across client devices may be achieved”)
The Yang reference has been addressed above. More specifically, Luo teaches “[…] based on computational constraints of each of the plurality of clients” (Luo pg. 6
PNG
media_image1.png
342
578
media_image1.png
Greyscale
which shows training based on an expected cost)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Yang with that of Luo since “Experimental evidence verifies our derived properties and demonstrates that our proposed solution achieves near-optimal performance for different optimization metrics for various datasets and heterogeneous system and statistical settings” Luo abstract. This shows that by combining the cost-effective optimization algorithm of Luo with the system of Yang, we have better federated learning that operates more optimally.
Note that independent claim 18 recites the same substantial subject matter as independent claim 1, only differing in embodiment. The differences in embodiment, a method and computer-readable medium are obvious variations of another and therefore the claim is subject to the same rejection.
Regarding claim 2, the Yang and Luo references have been addressed above. Yang further teaches “further comprising, receiving, at the central server, local model parameters from each of the plurality of clients” (fig. 5 S503 and [0023] “the client computing device is further configured to receive an aggregated set of common layers from the server computing device, and update the model based on the aggregated set of common layers”)
Regarding claim 3, the Yang and Luo references have been addressed above. Yang further teaches “further comprising: aggregating, at the central server, the local model parameters across the plurality of clients into global model parameters” (previous citation, “the client computing device is further configured to receive an aggregated set of common layers from the server computing device, and update the model based on the aggregated set of common layers”); and
“sending the global model parameters to each of the plurality of clients to update respective local models at each of the plurality of clients” (“[0120] S503: sending, by the client computing device, the updated set of common layers to the server computing device, and [0121] S504: storing, by the client computing device, the updated set of client-specific layers.”)
Regarding claim 4, the Yang and Luo references have been addressed above. Yang further teaches “further comprising: waiting, by the central server, until the aggregated local model parameters are received from all of the plurality of clients before updating the global model parameters” (“[0123] S601: aggregating, by the server computing device, the received updated sets of common layers to obtain an aggregated set of common layers, [0124] S602: sending, by the server computing device, the aggregated set of common layers to each of the client computing devices, [0125] S603: updating, by the client computing device, the model based on the aggregated set of common layers” aggregating, and then sending the aggregated layers i.e. waiting)
Regarding claim 5, the Yang and Luo references have been addressed above. Luo further teaches “further comprising: obtaining a number of complexity levels for the federated learning system; obtaining a target computational overhead reduction ratio for each of the complexity levels” (Luo abstract “the iterative on-device learning process (e.g., local computations and global communications with the server) incurs a considerable cost in terms of learning time and energy consumption, which depends crucially on the number of selected clients and the number of local iterations in each training round” learning time and energy consumption are both complexity and overhead costs); and
“computing a computational overhead of each of the plurality of local models at each of the complexity levels” (Luo pg. 6
PNG
media_image1.png
342
578
media_image1.png
Greyscale
which entails computing the cost)
Regarding claim 8, the Yang and Luo references have been addressed above. Yang further teaches “further comprising initiating, by the central server, a federated learning round t for t = 1, 2, …, T” ([0111] “For each round of communications, Step 1, the client computing devices 210 may update the local model 100 by copying the Backbone. If it is the first round of communication, the LL (Classifier) may be copied as well.” each round i.e. multiple rounds at a given time point)
Regarding claim 9, the Yang and Luo references have been addressed above. Luo further teaches “wherein the central server ends the training after T rounds and outputs a trained global model” (Luo pg. 4 table II shows round numbers)
Regarding claim 10, the Yang and Luo references have been addressed above. Luo further teaches “further comprising identifying, by the central server, sK available clients among the plurality of clients” (previous citation, shows number of selected clients or identified clients)
Regarding claim 11, the Yang and Luo references have been addressed above. Luo further teaches “further comprising sending, by the central server, a selected one of the plurality of the local models to each of the sK available clients” (previous citation)
Regarding claim 12, the Yang and Luo references have been addressed above. Luo further teaches “further comprising: assigning a complexity level for each of the sK available clients such that a computational overhead of an assigned local model for each of the sK available clients does not exceed a budget of each of the sK available clients” (Luo abstract “the iterative on-device learning process (e.g., local computations and global communications with the server) incurs a considerable cost in terms of learning time and energy consumption, which depends crucially on the number of selected clients and the number of local iterations in each training round” and pg. 4 table II, computation time and energy of client k);
Yang further teaches “obtaining the assigned local model by applying a split algorithm to the global model with computed downscaling ratios for each complexity level; and sending the assigned local model to a client device” ([0105] “As stated above (and as shown on the left-hand side of FIG. 3), a contribution of this embodiment is the virtual separation of the model 100 of the neural network - here it is exemplarily a CNN network—into a set of common layers 120 and a set of client-specific layers 140. The way of separating the model 100 may be performed according to the CNN's property. Here, in this embodiment, the set of common layers 120 is referred to as “Backbone”, e.g., stacked convolutional layers, and the set of client-specific layers 140 is referred to as last layers (LL), e.g., last fully connected layers. In particular, the CNN may be a common classification network using stacked convolutional layers at the beginning, followed by fully connected layers. The LL may also be referred to as “LL Classifier”, since it/they is/are the classifier that contains class specific information” this is interpreted as two-dimensional uniform downscaling since the model is split based on width and depth, or node/layer)
Regarding claim 13, the Yang and Luo references have been addressed above. Luo further teaches “further comprising training each of the plurality of local models on each of the plurality of clients in parallel” (Luo right col. ¶2 “These unique features necessitate FL algorithms to perform multiple local iterations in parallel on a fraction of randomly sampled clients and then aggregate the resulting model update via the central server periodically”)
Regarding claim 14, Yang teaches “a computer-implemented method for training a global model on a central server in a federated learning system having a plurality of nodes, the method comprising” (abstract “The client computing device is configured to receive a model comprising a set of common layers and a set of client-specific layers from the server computing device. After a training at the client computing device, the set of common layers and the set of client-specific layers are both updated”):
“splitting the global model along a width and a depth via two-dimensional uniform downscaling of the global model to create a plurality of local models” ([0105] “As stated above (and as shown on the left-hand side of FIG. 3), a contribution of this embodiment is the virtual separation of the model 100 of the neural network - here it is exemplarily a CNN network—into a set of common layers 120 and a set of client-specific layers 140. The way of separating the model 100 may be performed according to the CNN's property. Here, in this embodiment, the set of common layers 120 is referred to as “Backbone”, e.g., stacked convolutional layers, and the set of client-specific layers 140 is referred to as last layers (LL), e.g., last fully connected layers. In particular, the CNN may be a common classification network using stacked convolutional layers at the beginning, followed by fully connected layers. The LL may also be referred to as “LL Classifier”, since it/they is/are the classifier that contains class specific information” this is interpreted as two-dimensional uniform downscaling since the model is split based on width and depth, or node/layer), “wherein one of the plurality of local models corresponds to each of the number of complexity levels” ([0091] “In one embodiment, the set of client-specific layers 140 may comprise last fully connected layers of the neural network. Optionally, the set of common layers 120 may comprise convolutional layers of the neural network. Optionally, the neural network may be a convolutional neural network.” i.e. the clients are less complex)
“sending an assigned one of the plurality of local models to each of a plurality of clients […]” ([0029] “By separating the model into the set of common layers and the set of client-specific layers, the server computing device is able to contribute to the improved training of the model in a collaborative manner with one or more client computing devices. Thus, a global accuracy of the model can be assured, while a local accuracy is also improved. Further, generalization across client devices may be achieved”)
“training the assigned ones of the plurality of local models on respective ones of the plurality of clients” (previous citation [0029] “By separating the model into the set of common layers and the set of client-specific layers, the server computing device is able to contribute to the improved training of the model in a collaborative manner with one or more client computing devices. Thus, a global accuracy of the model can be assured, while a local accuracy is also improved. Further, generalization across client devices may be achieved”)
The Yang reference has been addressed above. More specifically, Luo teaches “obtaining a number of complexity levels for the federated learning system; obtaining a target computational overhead reduction ratio for each of the complexity levels” (Luo abstract “the iterative on-device learning process (e.g., local computations and global communications with the server) incurs a considerable cost in terms of learning time and energy consumption, which depends crucially on the number of selected clients and the number of local iterations in each training round” learning time and energy consumption are both complexity and overhead costs)
“computing a computational overhead of each of a plurality of local models at each of the complexity levels” (Luo pg. 6
PNG
media_image1.png
342
578
media_image1.png
Greyscale
which entails computing the cost)
“[…] based on an available computational overhead budget at each of the plurality of clients, wherein the computational overhead of the assigned one is less than the available computational overhead budget at each of the plurality of clients” (Luo pg. 6
PNG
media_image1.png
342
578
media_image1.png
Greyscale
which shows training based on an expected cost or budget and pg. 4 table II which shows the various parameters
PNG
media_image2.png
710
582
media_image2.png
Greyscale
)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Yang with that of Luo since “Experimental evidence verifies our derived properties and demonstrates that our proposed solution achieves near-optimal performance for different optimization metrics for various datasets and heterogeneous system and statistical settings” Luo abstract. This shows that by combining the cost-effective optimization algorithm of Luo with the system of Yang, we have better federated learning that operates more optimally.
Dependent claims 15-17 are mapped and correspond to dependent claims 2-3 and 11 respectively.
Dependent claim 19 is mapped to dependent claims 2 and 3.
Dependent claim 20 is mapped to dependent claim 5.
Claim(s) 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Yang and Luo further in view of Zhong, Zhengyi, et al. "Flee: A hierarchical federated learning framework for distributed deep neural network over cloud, edge, and end device."
Regarding claim 6, the Yang and Luo references have been addressed above. They do not explicitly teach early exit. Zhong however teaches “further comprising determining early exits of the global model to generate a local model for each of the complexity levels” (Zhong pg. 3 ¶ above §2 “The proposed method FLEE breaks the restriction that FL is only aimed at the single-exit model, and applies the FL framework to the training of multi-exit models deployed hierarchically.”)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Yang and Luo with that of Zhong since a combination of known methods would yield predictable results. As shown in Zhong, early exit in federated learning is a known concepts and allows resources to be saved. Therefore this technique when combined with the system above would operate as expected and predictably, therefore allowing for more optimal learning.
Claim(s) 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Yang in view of Luo further in view of Ruan, Xiaofeng, et al. "EDP: An efficient decomposition and pruning scheme for convolutional neural network compression."
Regarding claim 7, the Yang and Luo references have been addressed above. They do not explicitly teach the claim limitations. Ruan however teaches “further comprising determining a uniform two-dimensional downscaling ratio through a grid search” (Ruan pg. 10 §4
PNG
media_image3.png
486
642
media_image3.png
Greyscale
)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Yang and Luo with that of Ruan since a combination of known methods would yield predictable results. As shown in Ruan, grid search is a known technique for adjusting hyperparameter settings which would include size. Therefore this would operate with the system above as expected.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Pfeiffer, Kilian, et al. "CocoFL: Communication-and computation-aware federated learning via partial NN freezing and quantization." arXiv preprint arXiv:2203.05468 (2022).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN W FIGUEROA whose telephone number is (571)272-4623. The examiner can normally be reached Monday-Friday, 10AM-6PM EST.
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, MIRANDA HUANG can be reached at (571)270-7092. 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.
KEVIN W FIGUEROA
Primary Examiner
Art Unit 2124
/Kevin W Figueroa/Primary Examiner, Art Unit 2124
Prosecution Insights