DETAILED ACTION
This action is responsive to the Application 18/167139 filed on 02/10/2023. Claims 1-20 are pending in the case. Claims 1, 9, and 15 are independent claims.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55 and Foreign Priority of 08/13/2020 is acknowledged.
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 .
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.
Claim Rejections - 35 USC § 112(b)
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. Dependent claims inherit the deficiencies of the independent claims.
Claims 1, 3, 6, 15, 17 and 20 recites the limitation the current round of training. There is insufficient antecedent basis for this limitation as the contingent limitation in the claims includes the BRI that includes the obtaining step is not performed therefor leaving the later claims that references to a current round of training found in the obtaining step without an antecedent basis. Examiner will interpret the references of the current round of training with a current round of training.
Claims 1, 4, 5, 15 and 18-19 recites the limitation the first training data. There is insufficient antecedent basis for this limitation as the contingent limitation in the claims includes the BRI that includes the obtaining step is not performed therefor leaving the later claims that references to a first training data found in the obtaining step without an antecedent basis. Examiner will interpret the references of the first training data with a first training data.
Claims 2-3 and 16-17 recites the limitation the local value of the parameter. There is insufficient antecedent basis for this limitation as the contingent limitation in the claims includes the BRI that includes the obtaining step is not performed therefor leaving the later claims that references to a local value of the parameter found in the obtaining step without an antecedent basis. Examiner will interpret the references of the local value of the parameter with a local value of the parameter.
Claims 1, 3, 9-11 and 14 recites the limitation the training result. There is insufficient antecedent basis for this limitation as the contingent limitation in the claims includes the BRI that includes the obtaining step is not performed therefor leaving the later claims that references to a training result found in the obtaining step without an antecedent basis. Examiner will interpret the references of the training result with a training result.
Claims 1, 3, 9-11, 14, 15 and 17 recites the limitation the training result. There is insufficient antecedent basis for this limitation as the contingent limitation in the claims includes the BRI that includes the obtaining step is not performed therefor leaving the later claims that references to a training result found in the obtaining step without an antecedent basis. Examiner will interpret the references of the training result with a training result.
Claim 14 recites the limitation second condition. There is insufficient antecedent basis for this limitation as the contingent limitation in the claims includes the BRI that includes the obtaining step is not performed therefor leaving the later claims that references to a second condition found in the obtaining step without an antecedent basis. Examiner will interpret the references of the second condition with a second condition.
Claim 8 recites the limitation wherein the performance of the machine learning. There is insufficient antecedent basis for this limitation. Examiner will interpret the references of the performance of the machine learning with a performance of the machine learning.
Claims 11-14 recites the limitation second actual value. There is insufficient antecedent basis for this limitation as there is no first actual data.
Claim 11 recites the limitation second training data. There is insufficient antecedent basis for this limitation as there is no first training data.
Claims 11-13 recites the limitation second target value. There is insufficient antecedent basis for this limitation as there is no first target value.
Claims 9 and 14 recites the limitation third value. There is insufficient antecedent basis for this limitation as there is no second value.
Claim Rejections - 35 USC § 112(d)
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claims 11-13 are rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends. The contingent limitation in claim 9 includes the BRI that includes the obtaining step is not performed therefor claims 11-13 do not further limit the claims as 11-13 provides more information on the obtaining step, which is not required with the BRI embodiment. Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1-6, 8, 15-20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Liu et al.(“ Deep Anomaly Detection for Time-series Data in Industrial IoT: A Communication-Efficient On-device Federated Learning Approach” henceforth known as Liu).
Regarding claim 1:
Liu discloses receiving, by a first client, a first value of a parameter of the machine learning model from the server, wherein the first client is one of the plurality of clients(Liu, Page 4, Col. 2, Paragraph 2, “The cloud aggregator… initializes the global model and sends the global model to the all edge devices” where the model parameters(ω) of the model are considered a first value of a parameter of the machine learning model from the server(See also Liu, Page 3, Col. 2, Paragraph 1, “we consider that FL with N edge devices and a parameter aggregator, i.e., a cloud aggregator, distributes a pre-trained global model ωt”)
Liu discloses when the first value of the parameter does not meet a first condition(Claim 1 recites the following contingent limitations: “when the first value of the parameter does not meet a first condition” These limitations are contingent because they recite steps that are only required to be performed if their conditions precedent are met. Limitation “performing, by the first client, a current round of training based on a first training data, the machine learning model, and a local value of the parameter, to obtain a training result of the current round of training, wherein the first training data is data reserved on the first client” only needs to be performed if “the first value of the parameter does not meet a first condition” Therefor the BRI does not require the step of obtaining as recited, however prior art still discloses the step of obtaining), performing, by the first client, a current round of training(Liu, Page 3, Col. 2, Paragraph 3, “Both the edge devices and the cloud aggregator repeat the above process until the global model reaches convergence” where the global model that is sent to all edge devices and trained until meeting the condition of convergence is considered of a value of a parameter(model parameters(ω)) not meeting a first condition(not optimally converged)) based on a first training data, the machine learning model, and a local value of the parameter(Liu, Page 7, Col. 1, Paragraph 3, “Edge devices use the local dataset to train the local model” where the client training the local model with parameters using a local dataset of the global model is considered a first client performing a current round of training with first training data and a local value of the parameter), to obtain a training result of the current round of training(Liu, Page 6, Col. 2, Equation 11 and 12, where F(ω) is the loss function, t represents the data sample for the t-th round of training and where the model being trained to minimize a local loss of dataset Dk until convergence is a condition being met with rounds of training), wherein the first training data is data reserved on the first client(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global mode” where the local dataset is considered a first training data reserved on the first client)
Liu discloses and sending, by the first client, the training result of the current round of training and alarm information to the server, wherein the alarm information indicates that the first value of the parameter does not meet the first condition(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator and uploads the gradients to the cloud aggregator until the global model converges” where the edge device uploading gradients until global model converges correspond to a first client sending training result and alarm information that indicates parameters do not meet the first condition of being optimally converged)
Regarding claim 2:
Liu teaches claim 1 (and thus the rejection of claim 1 is incorporated).
Liu discloses wherein the local value of the parameter is equal to a value that is of the parameter and that is obtained in a previous round of training(Liu, Page 6, Col. 2, Equation 11 and 12, where F(ω) is the loss function, f(x,ω) is the loss function for the local device ω are the parameters of the model, t represents the data sample for the t-th round of training and where the model being trained to minimize a local loss of dataset Dk is a condition being met with rounds of training correspond with local value of the parameters equal to a value obtained from a previous round of training)
Regarding claim 3:
Liu teaches claim 1 (and thus the rejection of claim 1 is incorporated).
Liu discloses wherein the training result of the current round of training is a value that is of the parameter and that is obtained in the current round of training, or a difference between a value that is of the parameter and that is obtained in the current round of training and the local value of the parameter(Liu, Page 4, Col. 1, Paragraph 3, “The edge devices train a shared global model locally on their own local dataset… and upload their model updates (i.e., gradients) to the cloud aggregator” where the current claim language only requires one or the other and where training the global model parameters corresponds to training results of the current round being a value that is of the parameter and obtained in the current round of training)
Regarding claim 4:
Liu teaches claim 1 (and thus the rejection of claim 1 is incorporated).
Liu discloses determining, by the first client based on the first training data and the machine learning model, that the first value of the parameter does not meet the first condition. (Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator and uploads the gradients to the cloud aggregator until the global model converges” where edge devices training and sending gradients until the global model converges corresponds to determining that the first value of the parameter does not meet the first condition and it being decided by the edge devices training data and machine learning model)
Regarding claim 5:
Liu teaches claim 4 (and thus the rejection of claim 4 is incorporated).
Liu discloses calculating, by the first client, a first actual value of at least one performance of the machine learning model based on the first training data and the first value of the parameter(Liu, Page 6, Col. 2, Equation 11, where F(ω) is the loss function, f(x,ω) is the loss function for the local device and the calculation of the loss function corresponds to calculating a first actual value of performance of the machine learning model(i.e. the loss of the model) based on a first training data and first value of the parameter(model parameters(ω)) and determining, by the first client based on the first actual value of the at least one performance of the machine learning model and a first target value of the at least one performance of the machine learning model, that the first value of the parameter does not meet the first condition(Liu, Page 4, Col. 2, Fig. 2, “(ii) The edge device performs local model…training on the local dataset…(iii) The edge device uploads the sparse gradients…(iv) The cloud aggregator obtains a new global model by aggregating the sparse gradients uploaded by the edge device…the above steps are executed cyclically until the global model reaches optimal convergence” where the edge devices training until global convergence is considered using a clients first actual value(the loss of a local model) for determining whether the first condition(convergence) has been met based on a first target value(optimally converged global model parameters ω))
Regarding claim 6:
Liu teaches claim 5 (and thus the rejection of claim 5 is incorporated).
Liu discloses wherein a first target value of at least one performance of the machine learning model is a value that is of the at least one performance of the machine learning model and that is obtained after a previous round of training is performed, or a maximum value that is of the at least one performance of the machine learning model and that is obtained after all rounds of training are performed before the current round of training(Liu, Page 7, Col. 2, Algorithm 1, where the current claim language only requires one or the other and where a first target value(optimally converged global model parameters ω) is a value that is of performance of the machine learning model and that is obtained after a previous round of training is performed as the model changes from one iteration to the next)
Regarding claim 8:
Liu teaches claim 5 (and thus the rejection of claim 5 is incorporated).
Liu discloses wherein a performance of the learning model comprises at least one of accuracy, precision, a recall rate, or an F1 score(Liu, Page 7, Col. 2, Paragraph 2, “We adopt Root Mean Square Error (RMSE) to indicate the performance” where RMSE indicating performance corresponds to accuracy of a learning model)
Regarding claim 15:
Liu discloses non-transitory computer-readable storage medium, storing one or more instructions that, when executed by at least one processor on one of a plurality of clients, wherein a same machine learning model is deployed on the plurality of clients, cause the at least one processor(Liu, Page 7, Col. 1, Paragraph 3, “The experiment is conducted on a virtual workstation with the Ubuntu 18.04 operation system, Intel (R) Core (TM) i5-4210M CPU, 16GB RAM, 512GB SSD.”)
Liu discloses receiving, a first value of a parameter of the machine learning model from a server(Liu, Page 4, Col. 2, Paragraph 2, “The cloud aggregator… initializes the global model and sends the global model to the all edge devices” where the model parameters(ω) of the model are considered a first value of a parameter of the machine learning model from the server(See also Liu, Page 3, Col. 2, Paragraph 1, “we consider that FL with N edge devices and a parameter aggregator, i.e., a cloud aggregator, distributes a pre-trained global model ωt”)
when the first value of the parameter does not meet a first condition(Claim 1 recites the following contingent limitations: “when the first value of the parameter does not meet a first condition” These limitations are contingent because they recite steps that are only required to be performed if their conditions precedent are met. Limitation “performing, by the first client, a current round of training based on a first training data, the machine learning model, and a local value of the parameter, to obtain a training result of the current round of training, wherein the first training data is data reserved on the first client” only needs to be performed if “the first value of the parameter does not meet a first condition” Therefor the BRI does not require the step of obtaining as recited, however prior art still discloses the step of obtaining), performing, a current round of training(Liu, Page 3, Col. 2, Paragraph 3, “Both the edge devices and the cloud aggregator repeat the above process until the global model reaches convergence” where the global model that is sent to all edge devices and trained until meeting the condition of convergence is considered of a value of a parameter(model parameters(ω)) not meeting a first condition(not optimally converged)) based on a first training data, the machine learning model, and a local value of the parameter(Liu, Page 7, Col. 1, Paragraph 3, “Edge devices use the local dataset to train the local model” where the client training the local model with parameters using a local dataset of the global model is considered a first client performing a current round of training with first training data and a local value of the parameter), to obtain a training result of the current round of training(Liu, Page 6, Col. 2, Equation 11 and 12, where F(ω) is the loss function, t represents the data sample for the t-th round of training and where the model being trained to minimize a local loss of dataset Dk until convergence is a condition being met with rounds of training), wherein the first training data is data reserved on the first client(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global mode” where the local dataset is considered a first training data reserved on the first client)
Liu discloses and sending, the training result of the current round of training and alarm information to the server, wherein the alarm information indicates that the first value of the parameter does not meet the first condition(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator and uploads the gradients to the cloud aggregator until the global model converges” where the edge device uploading gradients until global model converges correspond to a first client sending training result and alarm information that indicates parameters do not meet the first condition of being optimally converged)
Regarding claim 16:
Liu teaches claim 15 (and thus the rejection of claim 15 is incorporated).
Liu discloses wherein the local value of the parameter is equal to a value that is of the parameter and that is obtained in a previous round of training(Liu, Page 6, Col. 2, Equation 11 and 12, where F(ω) is the loss function, f(x,ω) is the loss function for the local device ω are the parameters of the model, t represents the data sample for the t-th round of training and where the model being trained to minimize a local loss of dataset Dk is a condition being met with rounds of training correspond with local value of the parameters equal to a value obtained from a previous round of training)
Regarding claim 17:
Liu teaches claim 15 (and thus the rejection of claim 15 is incorporated).
Liu discloses wherein the training result of the current round of training is a value that is of the parameter and that is obtained in the current round of training, or a difference between a value that is of the parameter and that is obtained in the current round of training and the local value of the parameter(Liu, Page 4, Col. 1, Paragraph 3, “The edge devices train a shared global model locally on their own local dataset… and upload their model updates (i.e., gradients) to the cloud aggregator” where the current claim language only requires one or the other and where training the global model parameters corresponds to training results of the current round being a value that is of the parameter and obtained in the current round of training)
Regarding claim 18:
Liu teaches claim 15 (and thus the rejection of claim 15 is incorporated).
Liu discloses wherein the method further comprises: determining, based on the first training data and the machine learning model, that the first value of the parameter does not meet the first condition(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator and uploads the gradients to the cloud aggregator until the global model converges” where edge devices training and sending gradients until the global model converges corresponds to determining that the first value of the parameter does not meet the first condition and it being decided by the edge devices training data and machine learning model)
Regarding claim 19:
Liu teaches claim 18 (and thus the rejection of claim 18 is incorporated).
Liu discloses wherein the determining, based on the first training data and the machine learning model, that the first value of the parameter does not meet the first condition comprises: calculating, a first actual value of at least one performance of the machine learning model based on the first training data and the first value of the parameter(Liu, Page 6, Col. 2, Equation 11, where F(ω) is the loss function, f(x,ω) is the loss function for the local device and the calculation of the loss function corresponds to calculating a first actual value of performance of the machine learning model(i.e. the loss of the model) based on a first training data and first value of the parameter(model parameters(ω)) and determining, based on the first actual value of the at least one performance of the machine learning model and a first target value of the at least one performance of the machine learning model, that the first value of the parameter does not meet the first condition(Liu, Page 4, Col. 2, Fig. 2, “(ii) The edge device performs local model…training on the local dataset…(iii) The edge device uploads the sparse gradients…(iv) The cloud aggregator obtains a new global model by aggregating the sparse gradients uploaded by the edge device…the above steps are executed cyclically until the global model reaches optimal convergence” where the edge devices training until global convergence is considered using a clients first actual value(the loss of a local model) for determining whether the first condition(convergence) has been met based on a first target value(optimally converged global model parameters ω))
Regarding claim 20:
Liu teaches claim 18 (and thus the rejection of claim 18 is incorporated).
Liu discloses wherein a first target value of at least one performance of the machine learning model is a value that is of the at least one performance of the machine learning model and that is obtained after a previous round of training is performed, or a maximum value that is of the at least one performance of the machine learning model and that is obtained after all rounds of training are performed before the current round of training(Liu, Page 7, Col. 2, Algorithm 1, where the current claim language only requires one or the other and where a first target value(optimally converged global model parameters ω) is a value that is of performance of the machine learning model and that is obtained after a previous round of training is performed as the model changes from one iteration to the next)
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness
rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 7 and 9-14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu et al.(“Deep Anomaly Detection for Time-series Data in Industrial IoT: A Communication-Efficient On-device Federated Learning Approach” henceforth known as Liu) and further in view of Dinh et al.(“Federated Learning over Wireless Networks: Convergence Analysis and Resource Allocation” henceforth known as Dinh).
Regarding claim 7:
Liu teaches claim 5 (and thus the rejection of claim 5 is incorporated).
Liu discloses wherein the first condition is that a difference between the first target value of the performance of the machine learning model and the first actual value of the performance of the machine learning model is less than or equal to a first threshold(Din, Page 4, Col. 1, Equation 6, where F(wt) is the global loss of the model at iteration t(first actual value), F(w*) is the optimal global loss(first target value) and ϵ is a convergence tolerance(a first threshold) that specifies how close the current global model must be to the optimal model and F(wt) – F(w*) < ϵ corresponds to a first condition that is the difference between the first actual value of the performance model is less than a first threshold)
References Liu and Dinh are analogous art because they are from the same field of endeavor of using federated learning with edge/IoT deployment.
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Dinh before him or her, to modify the convergence process of Liu to include an convergence definition of Equation 6 of Dinh as Liu does not discuss or define a formal convergence condition, however states at a high level “until convergence”. The motivation to use equation 6 as a convergence condition is to balance the ideal theoretical convergence and the system design(the tradeoffs of accuracy/time/resources) from Dinh, Page 2, Col. 1, Paragraph 2. “Thus, to minimize the wall-clock training time of FL, a careful resource allocation problem for FL over wireless networks needs to consider not only the FL parameters such as accuracy level for computation-communication trade-off, but also allocating the UEs’ resources such as power and CPU cycles with respect to wireless conditions.”
Regarding claim 9:
Liu discloses separately sending, by the server, a first value of a parameter of the machine learning model to the plurality of clients(Liu, Page 4, Col. 2, Paragraph 2 “The cloud aggregator… initializes the global model and sends the global model to the all edge devices” where the model parameters(ω) of the model are considered a first value of a parameter of the machine learning model from the server(See also Liu, Page 3, Col. 2, Paragraph 1, “we consider that FL with N edge devices and a parameter aggregator, i.e., a cloud aggregator, distributes a pre-trained global model ωt)), wherein the first value of the parameter is used for a current round of training of the machine learning model(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator” where training the global model corresponds to using the first value of the parameter in a current round of training)
Liu discloses receiving, by the server, a training result of the current round of training from each of the plurality of clients(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator and uploads the gradients to the cloud aggregator until the global model converges” where the edge device uploading gradients to the cloud aggregator correspond to a server receiving a training result of the current round of training) obtaining, by the server when at least one client of the plurality of clients further reports an alarm information, a training result by screening the training results of the current round of training from the plurality of clients, wherein the alarm information indicates that the first value of the parameter does not meet a first condition(Claim 9 recites the following contingent limitations: “when at least one client of the plurality of clients further reports an alarm information.” These limitations are contingent because they recite steps that are only required to be performed if their conditions precedent are met. Limitation “obtaining… a training result by screening the training results of the current round of training from the plurality of clients, wherein the alarm information indicates that the first value of the parameter does not meet a first condition” only needs to be performed if “when at least one client of the plurality of clients further reports an alarm information” Therefor the BRI does not require the step of obtaining as recited)
.
Liu does not explicitly disclose and calculating, by the server, a third value of the parameter based on the training result obtained through screening, wherein the third value of the parameter is used for a next round of training of the machine learning model
Dinh discloses and calculating, by the server, a third value of the parameter based on the training result obtained through screening, wherein the third value of the parameter is used for a next round of training of the machine learning model(Din, Page 4, Col. 1, Algorithm 1, where step 5 shows using equation 4 to average local models and equation 5 forming a global gradient estimate to guide the next round of training where the aggregation corresponds to screening and, additionally, feedback using the server to update the global model parameters and feeding the model back to the user equipment/edge devices corresponding to a server calculating a third value of parameters that is used in the next round of training as the models are trained locally on edge devices. It is noted that the screening as recited in the step of calculating is interpreted as any screening, not necessarily requiring the details of the screening as recited in the obtaining step)
References Liu and Dinh are analogous art because they are from the same field of endeavor of using federated learning with edge/IoT deployment.
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Dinh before him or her, to modify the convergence process of Liu to include an convergence definition of Algorithm 1 step 5 of Dinh as Liu does not discuss or define a formal convergence condition, however states at a high level “until convergence”. The motivation to use equation 6 as a convergence condition is to balance the ideal theoretical convergence and the system design(the tradeoffs of accuracy/time/resources) from Dinh, Page 2, Col. 1, Paragraph 2. “Thus, to minimize the wall-clock training time of FL, a careful resource allocation problem for FL over wireless networks needs to consider not only the FL parameters such as accuracy level for computation-communication trade-off, but also allocating the UEs’ resources such as power and CPU cycles with respect to wireless conditions.”
Regarding claim 10:
Liu-Dinh teaches claim 9 (and thus the rejection of claim 9 is incorporated).
Liu discloses wherein the training result of the current round of training is a value that is of the parameter and that is obtained in the current round of training, or a difference between a value that is of the parameter and that is obtained in the current round of training and a local value of the parameter(Liu, Page 4, Col. 1, Paragraph 3, “The edge devices train a shared global model locally on their own local dataset… and upload their model updates (i.e., gradients) to the cloud aggregator” where the current claim language only requires one or the other and where training the global model corresponds to results of the current round of training being a value that is of the parameter and obtained in the current round of training)
Regarding claim 11:
Liu-Dinh teaches claim 9 (and thus the rejection of claim 9 is incorporated).
The BRI interpretation of the claims does not require wherein the obtaining, by the server when at least one client of the plurality of clients further reports an alarm information, a training result by screening the training results of the current round of training from the plurality of clients comprises: when at least one client of the plurality of clients further reports the alarm information, calculating, by the server, a second actual value of a performance of the machine learning model of each of the plurality of clients based on second training data and the training result of the current round of training from each of the plurality of clients, wherein the second actual value of the performance of the machine learning model is a value that is of the performance of the machine learning model and that is obtained after the current round of training is performed, and the second training data is data reserved on the server and obtaining, by the server by screening the training results of the current round of training from the plurality of clients based on the second actual value of the performance of the machine learning model of each of the plurality of clients and a second target value of the performance of the machine learning model, a training result that meets a second condition(based on the rejection under 112(d) of claim 11, this claim is rejection under the same reasons as set forth in the rejection of claim 9)
Regarding claim 12:
The BRI interpretation of the claims does not require wherein the second target value of the performance of the machine learning model is a maximum value that is of the performances of the machine learning models of the plurality of clients and that is obtained after all rounds of training are performed before the current round of training; or the second target value of the performance of the machine learning model is a maximum value of the second actual values of the performances of the machine learning models of the plurality of clients(based on the rejection under 112(d) of claim 12, this claim is rejection under the same reasons as set forth in the rejection of claim 9)
Regarding claim 13:
The BRI interpretation of the claims does not require wherein the second condition is that a second actual value of the performance of the machine learning model of a second client is greater than a second target value of the performance of the machine learning model or a second difference is less than a second threshold, wherein the second difference is equal to a difference between the second target value of the performance of the machine learning model and the second actual value of the performance of the machine learning model of the second client, and the second client is any one of the plurality of clients(based on the rejection under 112(d) of claim 13, this claim is rejection under the same reasons as set forth in the rejection of claim 9)
Regarding claim 14:
Liu-Dinh teaches claim 11 (and thus the rejection of claim 11 is incorporated).
Liu further discloses wherein all of training results of clients that report the alarm information meet a second condition(Liu, Page 4, Col. 2, Paragraph 3, “Each edge device uses the local dataset…to train the global model sent by the cloud aggregator and uploads the gradients to the cloud aggregator until the global model converges” where the edge device uploading gradients correspond to a first client sending training result of the current round of training and sending gradients until global model convergence corresponds to reporting alarm information that indicates parameters meeting a second condition of being optimally converged)
Liu does not explicitly disclose a maximum value of a second actual value that is of the performance of the machine learning model and that is calculated based on the training result of the client that reports the alarm information is greater than a maximum value of a second actual value that is of the performance of the machine learning model and that is calculated based on a training result of a client that does not report the alarm information and the calculating, by the server, a third value of the parameter based on the training result obtained through screening comprises: determining, by the server, the third value of the parameter based on the training result that is of the client that reports the alarm information and that is in the training result obtained through screening
Dinh discloses a maximum value of a second actual value that is of the performance of the machine learning model and that is calculated based on the training result of the client that reports the alarm information is greater than a maximum value of a second actual value that is of the performance of the machine learning model and that is calculated based on a training result of a client that does not report the alarm information(Din, Page 4, Col. 1, Paragraph 3, “For an arbitrary small constant ϵ > 0, the problem (1) achieves a global model convergence wt when its satisfies [Equation 6]” where the parameters sent that achieve global model wt at iteration t that satisfies equation 6 corresponds to not reporting the alarm information as reporting alarm information indicates parameters do not meet the first condition of being optimally converged and the model for iteration F(wt-1) has a greater value than F(wt) as the loss gets smaller) and the calculating, by the server, a third value of the parameter based on the training result obtained through screening comprises: determining, by the server, the third value of the parameter based on the training result that is of the client that reports the alarm information and that is in the training result obtained through screening(Din, Page 4, Col. 1, Algorithm 1, where step 5 shows the aggregation and feedback using the server to update the global model and feeding the model back to the user equipment/edge devices corresponds to a server calculating a third value of parameters that is used in the next round of training)
References Liu and Dinh are analogous art because they are from the same field of endeavor of using federated learning with edge/IoT deployment.
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art, having the teachings of Liu and Dinh before him or her, to modify the “repeat until convergence” process of Liu to include convergence calculation of algorithm 1 and equations 4-6 of Dinh as gradients are already being uploaded and weighted aggregation is compatible with data balance by providing the stopping criteria of equation 6 as Algorithm 1 is defined over a generic federated objective. The motivation for incorporating Algorithm 1 would be Dinh Page 3, Col. 2, Paragraph 3, “In this section, we propose a FL algorithm, named FEDL, as presented in Algorithm 1. To solve problem (1)” where problem 1 discusses minimizing the global loss.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHARLES JEFFREY JONES JR whose telephone number is (703)756-1414. The examiner can normally be reached Monday - Friday 8:00 - 5:00 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, Kakali Chaki can be reached at 571-272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/C.J.J./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122