Notice of Pre-AIA or AIA Status 1. 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 2. This action is in response to the original filing on 09/22/ 2023. Claims 1-17 are pending and have been considered below. Information Disclosure Statement 3. The information disclosure statement (IDS (s) ) submitted on 09/22/2023 is /are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner. Claim Rejections - 35 USC § 112 4 . 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. Claims 2- 5 , 7- 10 , and 12-1 7 are 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 regards as the invention. Claims 2, 7, and 12 recite wherein the permuting is according to: Claims 3, 8, and 13 recite wherein the permutation utilizes a matrix according to: . Claims 4 and 9 recite wherein each layer of the matrix is according to . Claim 14 recites wherein the permutation is performed on an activation for a layer with Z l =[z l,1 T , . . . , z l,n T ], wherein the permutation at each layer is performed according to It is noted that there are several variables in the equations are not defined. Claims 5 , 10, 15-17 incorporate the deficiencies of claim s 4, 9, 14 , through dependency, and are also rejected. Claim Rejections – 35 USC § 103 5 . 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 . 6 . Claim 1 is rejected under 35 U.S.C. 103 as being unpatentable over Mozaffari et al. (U.S. Patent Application Pub. No. US 20240064016 A1) in view of Sharma et al. (U.S. Patent Application Pub. No. US 20210357800 A1) in view of Zhan et al. (U.S. Patent Application Pub. No. US 20230325722 A1) and further in view of Chu et al. (U.S. Patent Application Pub. No. US 20210374617 A1). Claim 1: Mozaffari teaches a method of training neural networks with federated learning (i.e. Parameter permutation is performed for federated learning to train a machine learning model; para. [0003]) , the method comprising: within a plurality of clients, transferring respective weights of machine learning models from each of the plurality of clients to a respective neighboring client (i.e. The clients encrypt their updates, send them to the server, then the server can calculate their sum (using the ∘ operation) and sends back the encrypted results to the clients. Now, the clients can decrypt the global model locally and update their models; para. [0076]) within the plurality of clients without transferring locally-stored data of the plurality of clients (i.e. in a federated learning technique N clients collaborate to train a global machine learning model without directly sharing their data. In round t, the federation server (also referred to as the “aggregation server” or the “server”) samples n out of N total clients and sends them the most recent global model θ t . Each client re-trains θ t on its private data using a machine learning technique, such as stochastic gradient descent (SGD), and sends back the model parameter updates (x i for i th client) to the server; para. [0025]) , each client computing local model updates using private local data, those updates being transferred out from the clients, and the aggregated result being sent back to clients ; permuting the machine learning models according to the respective weights to obtain permuted weights (i.e. each parameter permutation for federated learning client chooses its shuffling pattern uniformly at random for each round of federated learning, which is private to that client; para. [0022 , 0051 ]) , parameter permutation, intra-model shuffling, and client-side production of shuffled/permuted parameters ; aggregating, the permuted weights to obtain aggregated permuted weights (i.e. The server can aggregate the PIR queries and their corresponding shuffled parameters for multiple clients to get the encrypted aggregated model. The aggregated model is decrypted independently at each client; para. [0022,0076]) , generation of an aggregated model from those shuffled parameters ; at each of the plurality of clients, training a local machine learning model with the locally-stored data that is stored locally at that respective client (i.e. Local updates to the parameters of the machine model may be computed by the client system according to a machine learning technique using local training data. Each client re-trains θ t on its private data using a machine learning technique; para. [0003, 0025]) ; and updating respective weights for each of the plurality of local machine learning model (i.e. Local updates to the parameters of the machine model may be computed by the client system according to a machine learning technique using local training data. Each client re-trains θ t on its private data using a machine learning technique; para. [0003, 0025]) . Mozaffari does not explicitly teach aggregating, at the plurality of clients , training a plurality of local machine learning models , wherein the training at each client includes determining a respective cross entropy loss for each of the plurality of local machine learning models and a loss computed based on a distance between the local machine learning models to the aggregated permuted weights; and updating respective weights for each of the plurality of local machine learning models based on the determined cross entropy loss and the loss. However, Sharma teaches within a plurality of clients, transferring respective weights of machine learning models from each of the plurality of clients to a respective neighboring client within the plurality of clients without transferring locally-stored data of the plurality of clients (i.e. transmitting a copy of the updated machine learning model from the first node to a first neighboring node in the network, receiving, at the first node, a modified machine learning model from the first neighboring node, the modified machine learning model having parameters set based on local data of the first neighboring node, modifying, at the first node, the updated machine learning model based on the modified machine learning model, and applying the updated machine learning model to perform operations at the first node; para. [0001, 0002, 0030]) ; aggregating, at the plurality of clients (i.e. the method 300 may include aggregating the one or more neighbor models with the receiving node's own local model, to further update the local model; para. [0030, 0034-0036]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Mozaffari to include the feature of Sharma . One would have been motivated to make this modification because it improves robustness in distributed training. However, Zhan teaches at each of the plurality of clients, training a plurality of local machine learning models with the locally-stored data that is stored locally at that respective client (i.e. the two models are trained on the client side device, one model participates in the federated learning and aggregation process, and the other model does not directly participate in the federated learning and aggregation process; para. [0060, 0118]) , the client-side shared model and private model , wherein the training at each client includes determining a respective cross entropy loss for each of the plurality of local machine learning models and a loss computed (i.e. the first loss value may include one or more of a cross-entropy loss value and a mutual distillation loss value, and the second loss value may include one or more of a cross-entropy loss value and a mutual distillation loss value. It can be learned from the second possible implementation of the first aspect that several typical first loss values and second loss values are provided, to increase diversity of the solution. It should be noted that, in this application, the obtained first loss value includes but is not limited to the cross-entropy loss value and the mutual distillation loss value, and the second loss value includes but is not limited to the cross-entropy loss value and the mutual distillation loss value; para. [0009]) , first lost and second loss for the shared model and private model ; and updating respective weights for each of the plurality of local machine learning models based on the determined cross entropy loss and the loss (i.e. The loss values of the two models are separately obtained based on the two trained models, the loss values of the two models are fused, and the two models are updated based on the fused loss value; para. [0007, 0011, 0012, 0060]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari and Sharma to include the feature of Zhan . One would have been motivated to make this modification because it improves the quality and reliability of each client’s local model training in the federated learning by using dual model. However, Chu teaches a loss computed based on a distance (i.e. the collaboration coefficient α ij between the set of local model parameters θ i and the set of local model parameters θ j may be computed as a simple Euclidean distance between the two sets of local model parameters; para. [0073]) between the local machine learning models to the aggregated permuted weights (i.e. T he collaboration coefficients may be considered to be a numerical or mathematical representation of the similarity between different sets of local model parameters. In particular, the collaboration coefficients may represent pair-wise similarity between pairs of sets of local model parameters. For a first set of the sets of local model parameters, collaboration coefficients may be calculated to represent similarity between the first set of local model parameters and each other set of local model parameters (of the total k sets of local model parameters received at step 502) ; para. [0068 , 0074, 0075 ]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, and Zhan to include the feature of Chu . One would have been motivated to make this modification because it improves the client-model update by making the update sensitive to how similar or close a client’s model parameters are to other client parameter sets. 7. Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Mozaffari in view of Sharma, Zhan, Chu, and further in view of Navon et al. ( Equivariant Architectures for Learning in Deep Weight Spaces ; arXiv, published 31 May 2023, pages 1-27 ). Claim 2 : Mozaffari, Sharma, Zhan, and Chu teach the method of claim 1 . Mozaffari does not explicitly teach wherein the permuting is according to: However, Navon teaches wherein the permuting is according to: (i.e. Permuting the rows and columns of the weight matrices using a permutation matrix P in the following way: W1 → PTW1, W2 → W2P will, in general, result indifferent weight matrices that represent exactly the same function. More generally, any sequence of weight matrices and bias vectors can be transformed by applying permutations to their rows and columns in a similar way, while representing the same function; Section I, page 2) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, and Chu to include the feature of Navon . One would have been motivated to make this modification because it provides the known implementation of neural network permutation using a permutation matrix. 8 . Claims 3 and 4 are rejected under 35 U.S.C. 103 as being unpatentable over Mozaffari in view of Sharma, Zhan, Chu, Navon, and further in view of Tatro et al. ( Optimizing Mode Connectivity via Neuron Alignment ; NeurIPS, published 2020, pages 1-12 ). Claim 3 : Mozaffari, Sharma, Zhan, Chu, and Navon teach the method of claim 2. Mozaffari does not explicitly teach wherein the permutation utilizes a matrix according to: . However, Tatro teaches wherein the permutation utilizes a matrix according to: (i.e. same layer of different neural networks. Given input d drawn from the input data distribution D, let X(1) l,i (d) ∈ Rk represent the activations of channel i in layer l of network θ1, where k is the number of units in the channel. A channel could be a unit in a hidden state or a filter in a convolutional layer. Given two networks of the same architecture, θ1 and θ2, we define a permutation, Pl : Kl → Kl, that maps from the index set of neurons in layer l of θ1 to θ2. This provides a correspondence between the neurons, and we can associate a cost function c : Rk × Rk → R+ for the individual correspondences to minimize. This is exactly solving the assignment problem Section 3, page 3) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, Chu, and Navon to include the feature of Tatro . One would have been motivated to make this modification because it provides a best way for choosing the permutation of one model that best aligns it with another model. Claim 4 : Mozaffari, Sharma, Zhan, Chu, Navon, and Tatro teach the method of claim 3. Mozaffari does not explicitly teach wherein each layer of the matrix is according to . However, Tatro teaches wherein each layer of the matrix is according to (i.e. same layer of different neural networks. Given input d drawn from the input data distribution D, let X(1) l,i (d) ∈ Rk represent the activations of channel i in layer l of network θ1, where k is the number of units in the channel. A channel could be a unit in a hidden state or a filter in a convolutional layer. Given two networks of the same architecture, θ1 and θ2, we define a permutation, Pl : Kl → Kl, that maps from the index set of neurons in layer l of θ1 to θ2. This provides a correspondence between the neurons, and we can associate a cost function c : Rk × Rk → R+ for the individual correspondences to minimize. This is exactly solving the assignment problem Section 3, page 3) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, Chu, and Navon to include the feature of Tatro . One would have been motivated to make this modification because it provides a best way for choosing the permutation of one model that best aligns it with another model. 9 . Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Mozaffari in view of Sharma, Zhan, Chu, Navon, Tatro , and further in view of Wierzynski et al. (U.S. Patent Application Pub. No. US 20150324689 A1). Claim 5 : Mozaffari, Sharma, Zhan, Chu, Navon, and Tatro teach the method of claim 4. Mozaffari does not explicitly teach comprising selecting respective updated weights from one of the each client via a random or pseudorandom selection. However, Wierzynski teaches comprising selecting respective updated weights from one of the each client via a random or pseudorandom selection (i.e. the central server may indicate a given model layer or set of weights for user devices to focus on (e.g., update) for the current iteration. In this case, the server may indicate the set of weights being targeted for the current model update iteration. The devices may track weight updates only related to the targeted set of weights. Similar to above, the devices may further select random subsets of this targeted set of weights. The devices may send their model weight updates to the central server at the end of an iteration. The server may, in turn compute an updated model for this set of weights and send out only these updated weights for the next model update; para. [0151]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, Chu, Navon, and Tatro to include the feature of Wierzynski . One would have been motivated to make this modification because it reduces communication and computation in the federated training process by transmitting and updating only randomly selected subset of weights. 10 . Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Mozaffari in view of Sharma, Zhan, Chu, Navon, Tatro , and further in view of Ainsworth et al. ( Git Re-Basin: Merging Models modulo Permutation Symmetries ; arXiv, published 1 Mar 2023, pages 1-29 ). Claim 14 : Mozaffari, Sharma, Zhan, Chu, Navon, and Tatro teach the robotic system of claim 13. Mozaffari does not explicitly teach wherein the permutation is performed on an activation for a layer with Z l =[z l,1 T , . . . , z l,n T ], wherein the permutation at each layer is performed according to However, Ainsworth teaches wherein the permutation is performed on an activation for a layer with Z l =[z l,1 T , . . . , z l,n T ], wherein the permutation at each layer is performed according to (i.e. For activations of the ’th layer, let Z(A),Z(B) ∈ Rd× n denote the d-dim. activations for all n training data points in models A and B, respectively. Then, ; section 3.1, pages 3-4) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, Chu, Navon, and Tatro to include the feature of Ainsworth . One would have been motivated to make this modification because it reduces communication and computation in the federated training process by transmitting and updating only randomly selected subset of weights. 11 . Claims 16 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Mozaffari in view of Sharma, Zhan, Chu, Navon, Tatro , Ainsworth , Wierzynski , and further in view of Ito et al. (U.S. Patent Application Pub. No. US 20220036190 A1). Claim 16 : Mozaffari, Sharma, Zhan, Chu, Navon, Tatro, Ainsworth, and Wierzynski teach the robotic system of claim 15. Mozaffari does not explicitly teach wherein the robotic system is an autonomous driving vehicle. However, Ito teaches wherein the robotic system is an autonomous driving vehicle (i.e. an autonomous driving system; para. [0023]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, Chu, Navon, Tatro, Ainsworth, and Wierzynski to include the feature of Ito . One would have been motivated to make this modification because to apply the known federated training framework to an autonomous driving vehicle. Claim 17 : Mozaffari, Sharma, Zhan, Chu, Navon, Tatro, Ainsworth, and Wierzynski teach the robotic system of claim 15. Mozaffari does not explicitly teach wherein the robotic system is a medical system . However, Ito teaches wherein the robotic system is an autonomous driving vehicle (i.e. a medical image diagnosis system ; para. [0023]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Mozaffari, Sharma, Zhan, Chu, Navon, Tatro, Ainsworth, and Wierzynski to include the feature of Ito . One would have been motivated to make this modification because to apply the known federated training framework to a medical system . 12 . Claims 6-13 and 15 are similar in scope to Claims 1-5 and are rejected under a similar rationale. Conclusion The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. Gu et al. (Pub. No. US 20220374763 A1) , a federated learning system and method for neural network training to defend against privacy leakage and data reconstruction attacks is described. The approach herein provides enhanced protection against information leakage by leveraging a multi-layered defense strategy that comprises several aspects, namely, trustworthy aggregation, decentralized aggregation with model partitioning, and dynamic permutation . It is noted that any citation to specific pages, columns, lines, or figures in the prior art references and any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck , 699 F.2d 1331, 1332-33, 216 U.S.P.Q. 1038, 1039 (Fed. Cir. 1983) (quoting In re Lemelson , 397 F.2d 1006, 1009, 158 U.S.P.Q. 275, 277 (C.C.P.A. 1968)). Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAN TRAN whose telephone number is (303)297-4266. The examiner can normally be reached on Monday - Thursday - 8:0 0 am - 5 :00 pm MT . 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 unsuccess ful, the examiner’s supervisor, Matt Ell can be reached on 571-270-3264 . The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /TAN H TRAN/ Primary Examiner, Art Unit 2141