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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 10/10/2023. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claims 1 ,12 are objected to because of the following informalities:
As to Claim 1 and analogous claim 12
a central node configured to communication with the edge nodes, should read as a central node configured to communicate with the edge nodes
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.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, 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.
As to claim 1 and analogous Claim 12
The limitation “Pruned model comprises an individual in an initial generation.” makes the claim unclear if the individual in an initial generation is the generation where first model is generated or if the individual in an initial generation is the generation where the model are trained.
The limitation “performing a search-iteration process to create a next generation of individuals or, alternatively when the halting condition is met, deploying a pruned model of a best scoring individual to one or more target edge nodes.” The use of or, alternatively in the Claim makes the claim Unclear. Examiner suggests the use of ‘and’ to require one of the of the contingencies to occur.
As to Claim 6
Claim 6 recites “ selects top fitness individuals”. There is insufficient antecedent basis for this limitation in the claim. The independent Claim , claim 1 recites “best scoring individual” which makes Claim 6 recitation unclear on what top fitness individual denotes. Examiner suggests claim limitation should read as” selects best scoring fitness individuals.”
As to claim 16
Claim 16 recites “the top fitness individuals are identified“. There is insufficient antecedent basis for this limitation in the claim. The independent Claim , claim 1 recites “best scoring individual” which makes Claim 6 recitation unclear on what top fitness individual denotes. Examiner suggests claim limitation should read as “best scoring fitness individuals are identified”.
As to Claim 10
Claim 10 recites “the random prune mask”. There is insufficient antecedent basis for this limitation in the claim. Examiner suggests claim limitation should read as respective random prune mask.
Claims 2-11 are further rejected on virtue of their dependencies to claim 1.
Claims 11-20 are further rejected on virtue of their dependencies to claim 12.
Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 1-20 is rejected under 35 USC § 101 because claimed invention is directed to the abstract idea without significantly more.
As to claim 1:
Step 1 Analysis: Is the claim to a process, machine, manufacture or composition of matter? See MPEP § 2106.03.
Claim 1 is a method claim, therefore it falls under one of four categories of statutory subject matter.
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “causing one or more of the edge nodes to generate and train an initial full candidate machine learning (ML) model, and the training of the initial full candidate ML models is performed only once at the one or more edge nodes;” is an abstract idea of mathematical relationship and as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “at the central node, applying a respective random prune mask to each of the initial full candidate ML models so as to generate a respective pruned model, and each of the pruned models comprises an individual in an initial generation” Is an abstract idea of mathematical relationship and as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “computing a fitness score for each of the individuals based on a generalization loss and on a number of pruned parameters in the model” is an abstract idea of mathematical relationship as the limitation recites, calculating a quality score on pruned parameters in the model. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “and when a halting condition is not met, performing a search-iteration process to create a next generation of individuals” is the abstract idea of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “A method, comprising: in an environment comprising edge nodes and a central node configured to communication with the edge nodes, performing operations comprising” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
The limitation “when the halting condition is met, deploying a pruned model of a best scoring individual to one or more target edge nodes” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “A method, comprising: in an environment comprising edge nodes and a central node configured to communication with the edge nodes, performing operations comprising” is an additional element to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to a method
Performed by computational unit, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
The limitation “when the halting condition is met, deploying a pruned model of a best scoring individual to one or more target edge nodes” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
Therefore, in examining elements as recited by the limitations individually and as an ordered combination, as a whole the independent claim limitations do not recite what have the courts have identified as “significantly more”.
As to Claim 12
Step 1 Analysis: Is the claim to a process, machine, manufacture or composition of matter? See MPEP § 2106.03.
Claim 12 is drawn to a storage medium consisting of Hardware components. i.e (article of manufacture), therefore claim 12 falls under one of four categories of statutory subject matter (machine/products/apparatus, process/method, manufactures and compositions of matter.
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to” is additional element is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2)
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to” is the additional claim elements of one or more processors; a memory coupled to at least one of the processors; a stored instructions stored in the memory and executed by at least one of the processors, are not sufficient to amount to significantly more than the judicial exception since these additional claim elements are recited at a high level of generality (i.e. using a generic processor and generic memory.
And for all other claim elements of claim 12 they are rejected using the PEG analysis of claim 1 since they are analogous claims.
As to claim 2
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “Computing of a fitness score” “is an abstract idea of mathematical relationship and as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “causing each of the edge nodes to perform a loss evaluation of the individual received by that edge node to perform a loss evaluation of the individual received by that edge node;” is an abstract idea of mathematical relationship and as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “transmitting each of the individuals to a respective one of the edge nodes and receiving the loss associated to the individual at the central node.” is an additional element that amounts to adding insignificant extra-solution activity of mere data gathering to the judicial exception. The claim recites generating an output in the form of instruction for an activity. See MPEP §§ 2106.04(d), 2106.05(g).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “transmitting each of the individuals to a respective one of the edge nodes and receiving the loss associated to the individual at the central node.” amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to move data to their respective nodes, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
As to claim 3
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “comprising orchestration and tracking of which edge nodes receive which individuals for loss evaluation.” is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to claim 4
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the training at each edge node is performed with real data that is local to that edge node. is an abstract idea of a of a mental process. Mental process can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to claim 5
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the training is performed without any exchange of local data between the edge nodes.” is the abstract idea of a mental process and a mathematical concept that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to claim 6
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation ‘’wherein the search-iteration process selects top fitness individuals and a random sample of individuals for inclusion in the next generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering). See MPEP §§ 2106.04(d), 2106.05(g).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation ‘’wherein the search-iteration process selects top fitness individuals and a random sample of individuals for inclusion in the next generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering). which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II). See MPEP §§ 2106.04(d), 2106.05(g).
As to claim 7
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the top fitness individuals are identified according to their respective fitness scores.” is the abstract idea of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to claim 8
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “, wherein the search-iteration process comprises generating new individuals for the next generation” generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering in a loop).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation ‘’wherein the search-iteration process selects top fitness individuals and a random sample of individuals for inclusion in the next generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering). which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II). See MPEP §§ 2106.04(d), 2106.05(g).
As to Claim 9
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “wherein the new individuals are generated based on the individuals in the initial generation” is an additional element that amounts to adding insignificant extra-solution activity of mere data input/output to the judicial exception. The claim recites generating a model from a previous generation in an iterative process. See MPEP §§ 2106.04(d), 2106.05(g).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation ‘’ wherein the new individuals are generated based on the individuals in the initial generation’ is an additional element that amounts to adding insignificant extra-solution activity of mere data input/output to the judicial exception. The claim recites generating a model from a previous generation in an iterative process, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II). See MPEP §§ 2106.04(d), 2106.05(g).
As to Claim 10
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “the new individuals are generated using new prune masks” is an abstract idea is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “that include portions of the random prune masks that were used to generate the individuals in the initial generation.” of mathematical relationship, as directed to a mathematical relationship is a relationship between variables or numbers. Claim recites using a portion of array of the initial generation to create a new model. See MPEP § 2106.04(a)(2)(I)(A).
As to Claim 11
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “wherein the halting condition is met when 'm' generations of individuals have been generated without an improvement in best fitness of a best scoring individual”. a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to Claim 13
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the training at each edge node is performed with real data that is local to that edge node. is an abstract idea of a of a mental process. Mental process can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to Claim 14
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the training is performed without any exchange of local data between the edge nodes.” is the abstract idea of a mental process and a mathematical concept that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to Claim 15
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation ‘’wherein the search-iteration process selects top fitness individuals and a random sample of individuals for inclusion in the next generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering). See MPEP §§ 2106.04(d), 2106.05(g).
As to Claim 16
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the top fitness individuals are identified according to their respective fitness scores.” is the abstract idea of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
As to Claim 17
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “, wherein the search-iteration process comprises generating new individuals for the next generation” generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering in a loop).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation ‘’wherein the search-iteration process selects top fitness individuals and a random sample of individuals for inclusion in the next generation’’ is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering). which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II). See MPEP §§ 2106.04(d), 2106.05(g).
As to Claim 18
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “wherein the new individuals are generated based on the individuals in the initial generation” is an additional element that amounts to adding insignificant extra-solution activity of mere data input/output to the judicial exception. The claim recites generating a model from a previous generation in an iterative process. See MPEP §§ 2106.04(d), 2106.05(g).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation ‘’ wherein the new individuals are generated based on the individuals in the initial generation’ is an additional element that amounts to adding insignificant extra-solution activity of mere data input/output to the judicial exception. The claim recites generating a model from a previous generation in an iterative process, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II), 2106.04(d), 2106.05(g).
As to Claim 19
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “the new individuals are generated using new prune masks” is an abstract idea is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “that include portions of the random prune masks that were used to generate the individuals in the initial generation.” of mathematical relationship, as directed to a mathematical relationship is a relationship between variables or numbers. Claim recites using a portion of array of the initial generation to create a new model. See MPEP § 2106.04(a)(2)(I)(A).
As to Claim 20
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “wherein the halting condition is met when 'm' generations of individuals have been generated without an improvement in best fitness of a best scoring individual”. a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
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.
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 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.
Claim 1-6 ,8-10 , 12-15,17-19 are rejected under 35 U.S.C. 103 as being unpatentable over Zhu,et. al. “Multi-Objective Evolutionary Federated Learning” (“Zhu”) in view of Hangyu et. Al. “Real-Time Federated Evolutionary Neural Architecture Search” (“Hangyu”) and in view of Huang et. al “Distributed Pruning Towards Tiny Neural Networks in Federated Learning” (“Huang”)
As to Claim 1
Zhu teaches “in an environment comprising edge nodes and a central node configured to communication with the edge nodes, (Zhu, “pg –1312
PNG
media_image1.png
573
769
media_image1.png
Greyscale
Examiner notes : server is interpreted as teaching a central node and that the clients are interpreted as teaching edge nodes as claimed, where server communicates with clients.
[in an environment comprising edge nodes and a central node configured to communication with the edge nodes,]” ).
“and each of the pruned models comprises an individual in an initial generation; (Zhu” pg -1313:1) Step 1: Randomly generate a parent population Pt of a size N for the first generation. 2) Step 2: Create an offspring population Qt [and each of the pruned models] of the same size as the parent population Pt [comprises an individual in an initial generation] by using crossover and mutation operators on Merge Pt and Qt into a combined population Rt, where Rt = Pt U Qt has a size of 2N.”).
Examiner notes: Under BRI, The offspring generated by crossover and mutation removes redundancy and unwanted weights is interpreted as pruned model in an initial generation.
3. computing a fitness score for each of the individuals based on a generalization loss and on a number of pruned parameters in the model; (Zhu, “[pg-1316],
PNG
media_image2.png
571
667
media_image2.png
Greyscale
Examiner notes: Individuals are interpreted as population i [for each of the individuals] and computing the fitness score maps to theta I fitness score detailed in step 18, and step 19 -20 is interpreted as based on generalization loss and on number of pruned parameters in the model. [computing a fitness score based on a generalization loss and on a number of pruned parameters in the model]”).
4. and when a halting condition is not met, performing a search-iteration process to create a next generation of individuals (Zhu, “[pg -1313], Step 1: Randomly generate a parent population Pt of a size N for the first generation. 2) Step 2: Create an offspring population Qt of the same size as the parent population Pt by using crossover and mutation operators on Pt. Merge Pt and Qt into a combined population Rt, where Rt = Pt ⋃ Qt. [create a next generation of individuals] has a size of 2N … 5) Step 5: Go to step 2 and repeat the whole procedure [performing a search iteration process ] until a stop criterion is met.[ and when a halting condition is not met] “).
Zhu does not explicitly teach
causing one or more of the edge nodes to generate and train an initial full candidate machine learning (ML) model, and the training of the initial full candidate ML models is performed only once at the one or more edge nodes
at the central node, applying a respective random prune mask to each of the initial full candidate ML models so as to generate a respective pruned model,
Or/ alternatively when the halting condition is met, deploying a pruned model of best scoring individual to one or more target nodes.
Hangyu teaches “causing one or more of the edge nodes to generate and train an initial full candidate machine learning (ML) model, and the training of the initial full candidate ML models is performed only once at the one or more edge nodes;” (Hangyu, pg ,365,”. Therefore, a distributed approach called federated learning [1] was proposed to preserve data privacy, enabling multiple local devices [causing one or more of the edge nodes] to collaboratively train a shared global model generate and train an initial full candidate machine learning (ML) model] while the training data remain to be deployed on the edge devices. [Pg (365)] Second, sampling clients without replacement makes sure that each local device needs to train only one subnetwork for once at each generation. [and the training of the initial full candidate ML models is performed only once at the one or more edge nodes]
Examiner remarks- Local devices are interpreted as corresponding to the claimed edge nodes where the model are trained.
Hangyu and Zhu are related to the same field of endeavor (Data privacy in ML models). In view of the teachings of Hangyu it would have been obvious for a person of ordinary skill in the art to apply the teachings of Hangyu to Zhu before the effective filing date of the claimed invention in order reduce optimize the performance of the model and reduce the computational cost.(Hangyu, “[pg 364, abs] we propose an evolutionary
approach to real-time federated NAS that not only optimizes the model performance but also reduces the local payload.”).
Huang teaches “at the central node, applying a respective random prune mask to each of the initial full candidate ML models so as to generate a respective pruned model, (Huang,[pg-3],
Given a large neural network with dense parameters Θ, . [initial full candidate ML models] we aim to find [so as to generate] a specialized subnetwork with sparse parameters θ [a respective pruned model] and mask m on dense parameters to achieve the optimal prediction performance for federated learning. The sparse parameters are derived by applying a mask [ applying a respective random prune mask to each ] to the dense parameters:
PNG
media_image3.png
40
195
media_image3.png
Greyscale
PNG
media_image4.png
52
570
media_image4.png
Greyscale
Examiner remarks - is interpreted as the random mask operation in central node [Central node] as claimed while applying random mask (i.e. m ∈ 0,1) to generate new pruned model .
2. Or/ alternatively when the halting condition is met, deploying a pruned model of best scoring individual to one or more target nodes. (Haung , [pg-5]
PNG
media_image5.png
228
734
media_image5.png
Greyscale
PNG
media_image6.png
180
772
media_image6.png
Greyscale
Examiner notes: Under BRI , well trained model is interpreted as best scoring individual [of best scoring individual] and R stop[when the halting condition is met] is interpreted as halting condition when executed, fetch sparse parameters in step 5 is interpreted as corresponding to claim deploying pruned model to one or more target nodes.
Huang and Zhu are related to the same field of endeavor (Data privacy in ML models). In view of the teachings of Zhu it would have been obvious for a person of ordinary skill in the art to apply the teachings of Huang to Zhu before the effective filing date of the claimed invention to apply the pruning at central nodes on the models while maintaining privacy between users, reducing memory space and computational cost. (Huang, “[pg-190] FedTiny allows devices with tight memory and computational budgets to participate in the resource-intensive pruning process by reconfiguring interactions between the server and devices.”).
As to Claim 12:
Zhu teaches A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors(zhu,”[ pg -1318] For example, one run of evolutionary optimization of CNNs with a population size of 20 for 50 generations took us more than 1 week on a computer [non-transitory storage medium having stored therein instructions that are executable] with GTX 1080Ti GPU and i7-8th 8700 CPU [by one or more hardware processors] preventing us from running the evolutionary optimization for a large number of generations.”).
And for all the other limitation of Claim 12, it is rejected under same basis as Claim 1. As the Claim are Analogous.
As to claim 2
Zhu in view of Hangyu and in view of Huang teaches the method of claim 1.
Zhu further teaches transmitting each of the individuals to a respective one of the edge nodes (Zhu, “pg -1313
PNG
media_image7.png
355
650
media_image7.png
Greyscale
Examiner remarks: Here, Theta t is interpreted as model parameter and k Clients is interpreted corresponding to the claimed edge nodes, and downloading theta t to edge clients is interpreted as transmitting each individual to respective edge nodes.
causing each of the edge nodes to perform a loss evaluation of the individual received by that edge node;( Huang, “[pg-3 ] During training, density d of sparse mask m cannot exceed target density d target. dtarget is determined by the limitation of devices’ memory resources. We formulate the problem as a constrained optimization problem:
PNG
media_image8.png
77
335
media_image8.png
Greyscale
where L(θ,m,Dk) denotes the loss function for local dataset Dk on the k-th device.
Examiner remarks: L| theta , m, dk is interpreted as the loss evaluation of transmitted edge nodes.
[ causing each of the edge nodes to perform a loss evaluation of the individual received by that edge node] “).
Huang teaches “and receiving the loss associated to the individual at the central node” (Huang “[pg –4]
PNG
media_image9.png
211
439
media_image9.png
Greyscale
[and receiving the loss associated to the individual at the central node] “).
Examiner remarks: step 21, is interpreted as receiving the loss to the central node as claimed which is Evaluated on step 19.
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to claim 3
Hangyu in view of Zhu and in view of Huang teaches the method of claim 2.
Huang further teaches additionally comprising orchestration and tracking of which edge nodes receive which individuals for loss evaluation. (Huang,” pg- 4
PNG
media_image10.png
307
704
media_image10.png
Greyscale
…
PNG
media_image11.png
277
696
media_image11.png
Greyscale
Examiner notes: pg-193 details, C as a candidate model, which distributes model to k devices, with device(k) per candidate tracking, and nested loop structure in (step 2) for k=1 to k and (step 4) for c=1 to C which is orchestration. Furthermore, step 19 calculates loss evaluation. Which is uploaded to the server on step 20 and further distributed to the individual detailed by nested loop. “).
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to claim 4 and Analogous claim 13
Hangyu in view of Zhu and in view of Huang teaches the method of claim 1.
Zhu teaches wherein the training at each edge node is performed with real data that is local to that edge node. (Zhu “[pg-1312] The procedure of federated learning is shown in Fig. 2, where each client receives the parameters θt of the global model from the central server and then trains their individual local models using their own data. [wherein the training at each edge node is performed with real data that is local to that edge node]. After local training, each local device sends their trained local parameters (i.e., θ1t) to the server to be aggregated to get an updated global model θt+1 to be used for the next iteration’s training. The subscript t denotes the time sequences or so-called communication rounds in federated learning. “).
PNG
media_image12.png
607
638
media_image12.png
Greyscale
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to Claim 5 and Analogous claim 14
Zhu in view of Hangyu and in view of Huang teaches the method of claim 1.
Zhu further teaches wherein the training is performed without any exchange of local data between the edge nodes (Zhu “[pg -1310] Instead of sending data directly to the central server, each local client downloads [between the edge nodes] the current global model from the server, updates the shared model by training its local data, [ wherein the training is performed] and then uploads the updated global model back to the server. By avoid sharing local private data, [without any exchange of local data] users’ privacy can be effectively protected in federated learning. “).
Examiner remarks: : Clients are interpreted as corresponding to the claimed edge nodes where data is being trained without being shared.
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to Claim 6 and Analogous Claim 15
Zhu in view of Hangyu and in view of Huang teaches the method of claim 1
Zhu further teaches “search-iteration process selects top fitness individuals and a random sample of individuals for inclusion in the next generation.” (Zhu [pg-1314] NSGA-II begins with the initialization of the population of size M where the binary and real-valued chromosomes
are randomly initialized, [and a random sample of individuals] which is the parent population at the first generation. Two parents are selected using the tournament selection to create two offspring by applying one-point crossover and flip mutation on the binary chromosome and the simulated binary crossover (SBX) and polynomial mutation [36] on the real-valued chromosome. This process repeats until [search-iteration process selects]M offspring are generated. We then calculate the two objectives of each individual in the offspring population. After that, the parent and offspring populations are combined and sorted according to the nondominance relationship and crowding distance. [top fitness individuals] Finally, M high-ranking individuals from the combined population are selected as the parent of the next generation. [ for inclusion in the next generation.] “).
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to claim 8 and analogous claim 17
Hangyu in view of Zhu and in view of Huang teaches the method of claim 1.
Zhu further teaches wherein the search-iteration process comprises generating new individuals for the next generation. (Zhu “[pg-1313] 1) Step 1: Randomly generate a parent population Pt of a size N for the first generation. 2) Step 2: Create an offspring population Qt of the same size as the parent population Pt by using crossover and mutation operators on Pt. Merge Pt and Qt into a combined population Rt, where Rt = Pt ⋃ Qt has a size of 2N. generating new individuals for the next generation … 5) Step 5: Go to step 2 and repeat the whole procedure [wherein the search-iteration process comprises] until a stop criterion is met.] “).
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to Claim 9 and analogous claim 18
wherein the new individuals are generated based on the individuals in the initial generation. (Hangyu, “[pg-316]
PNG
media_image13.png
183
664
media_image13.png
Greyscale
[wherein the new individuals are generated based on the individuals in the initial generation] “).
Examiner remarks- Qt is interpreted corresponding to the claimed new individual ,generated by parent Pt interpreted as individual in initial generation in step 3. Also, Rt = pt + qt in step 4 generates new individual from previous/initial generation.
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
As to claim 10 and analogous claim 19
Hangyu in view of Zhu and in view of Huang teaches the method of claim 8.
Huang further teaches “wherein the new individuals are generated using new prune masks that include portions of the random prune masks that were used to generate the individuals in the initial generation. (pg-5
PNG
media_image14.png
281
727
media_image14.png
Greyscale
….
PNG
media_image15.png
524
783
media_image15.png
Greyscale
Examiner remarks- step 24 is interpreted as computing a new mask that includes the portion of initial mask (mt)used in an initial generation and step 26 is interpreted as pruning the model to generate the new individual corresponding to claim.
Zhu, Hangyu, and Huang are combinable for the same rationale as set forth above with respect to Claim 1
Claims 7, 11, 16, and 20 are rejected rejected under 35 U.S.C. 103 as being unpatentable” (“Zhu”) in view (“Hangyu”) and in view of (“Huang”) and in further view of Tzruia et.al “Fitness Approximation through Machine Learning”(Tzruia).
As to claim 7 and analogous claim 16
Zhu in view of Hangyu and in view of Huang teaches the method of claim 6.
Zhu in view of Hangyu and in view of Huang does not teach
wherein the top fitness individuals are identified according to their respective fitness scores.
Tzruia teaches wherein the top fitness individuals are identified according to their respective fitness scores. (Tzuria “[Pg- 2] proportional inheritance, wherein the fitness score of an offspring is a weighted average of its parents, based on the similarity of the offspring to each of its parents. However, each time a fitness score is computed for an individual, we update a dataset whose features are the encoding vector of the individual and whose target value [identified according to their respective fitness scores] is the respective fitness score. An illustration of the dataset is presented in Table I.
PNG
media_image16.png
366
529
media_image16.png
Greyscale
Examiner remarks: Top fitness have been identified for an individual with their respective score.”).
Tzruia and Zhu are related to the same field of endeavor (Data privacy in ML models). In view of the teachings of Tzruia it would have been obvious for a person of ordinary skill in the art to apply the teachings of Tzruia to Zhu before the effective filing date of the claimed invention to
Identify the models according to model’s respective fitness score to determine whether or not further iteration need to be carried out in an optimization process while reducing runtime. (Tzruia, “pg -8, Our results show a significant reduction in GA runtime, with a small price in fitness for low sample rates, and no price for high sample rates.)
As to Claim 11 and analogous claim 20
Zhu in view of Hangyu and in view of Huang teaches the method of claim 1
Zhu in view of Hangyu and in view of Huang does not teach
Wherein the halting condition is met when ‘m’ generation of individuals have been generated without improvement in best fitness of a best scoring individual.
Tzruia teaches “Wherein the halting condition is met when ‘m’ generation of individuals have been generated without improvement in best fitness of a best scoring individual. (Tzruia ,pg 5. Wait for the best fitness score to stabilize [Wherein the halting condition is met] before transitioning to prediction mode. We consider the best fitness score as stable if it has not changed much (below a given threshold) [without improvement in best fitness of a best scoring individual] over the last P generations, for a given P. [‘m’ generation of individuals have been generated].
Tzruia and Zhu are related to the same field of endeavor (Data privacy in ML models). In view of the teachings of Tzruia it would have been obvious for a person of ordinary skill in the art to apply the teachings of Tzruia to Zhu before the effective filing date of the claimed invention to
Perform iteration process until the best individual model does not show any improvement in their fitness score with the goal of selecting an optimized model in a genetic algorithm. (Tzruia “pg- 1, GA is a population-based meta-heuristic optimization algorithm that operates on a population of candidate solutions, referred to as individuals, iteratively improving the quality of solutions over generations. GAs employ selection, crossover, and mutation operators to generate new individuals based on their fitness values, computed using a fitness function.”).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NIROJ KOIRALA whose telephone number is (571)270-0748. The examiner can normally be reached Monday -Friday 8am-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, MICHAEL HUNTLEY can be reached on (303) 297-4307. 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.
/N.K./Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights