Detailed Action
This action is written in response to the application filed 10/18/2024. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Subject Matter Eligibility
In determining whether the claims are subject matter eligible, the examiner has considered and applied guidance from MPEP § 2106. The examiner finds that the independent claims are directed to the practical application of distributed machine learning, ie incrementally training a global machine learning model tasks on a local computing device using local data.
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.
Claims 1 and 3-6 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Li.
(Li et al., "Privacy-Preserving Federated Learning Framework Based on Chained Secure Multiparty Computing", IEEE INTERNET OF THINGS JOURNAL, VOL. 8, NO. 8, APRIL 15, 2021, pages 6178-6186, XP011848022. Cited by Applicant in IDS dated 10/18/24.)
Regarding claims 1 and 3, Li discloses a client computing device (and a related server computing device) including:
a processing unit; and
P. 6181 and passim: ‘client’ or ‘edge server’.
a non-volatile, computer-readable storage medium containing local dataset and instructions that, upon execution by the aforementioned processing unit, direct the client computing device to carry out a series of functions, said functions consisting of:
P. 6181, fig. 2 (reproduced below).
PNG
media_image1.png
568
588
media_image1.png
Greyscale
P. 6178, first-second col., “The main principle of FL is to protect the privacy of data owners (i.e., participants) by training machine-learning models based on data sets distributed across multiple devices. During the entire model training process, participants only exchange gradients rather than raw data or encrypted/desensitized raw data with the server.
a) obtaining global values corresponding to a group of parameters associated with a machine learning-based model;
P. 6180, “FL typically aims to minimize the empirical risk over heterogeneous data distributed across multiple devices [37]. Before the start of training every iteration, each participating user gets the global model parameters from the aggregation node (or the server). Then, the user uses the global model parameters to train a local model with the local data set and sends new local model parameters as the local update to the aggregation node.” (Emphasis added.)
P. 6181, fig. 2 (reproduced above): “Distributing the aggregating results to edge-servers”.
b) incrementally training the machine-learned model based, at least in part, on a partition of the local dataset to obtain updated machine learning-based model parameters; and
P. 6180, “FL typically aims to minimize the empirical risk over heterogeneous data distributed across multiple devices [37]. Before the start of training every iteration, each participating user gets the global model parameters from the aggregation node (or the server). Then, the user uses the global model parameters to train a local model with the local data set and sends new local model parameters as the local update to the aggregation node.” (Emphasis added.)
c) communicating the updated machine learning-based model parameters to the server computing device.
P. 6180, “FL typically aims to minimize the empirical risk over heterogeneous data distributed across multiple devices [37]. Before the start of training every iteration, each participating user gets the global model parameters from the aggregation node (or the server). Then, the user uses the global model parameters to train a local model with the local data set and sends new local model parameters as the local update to the aggregation node.” (Emphasis added.)
Regarding claim 4, Li discloses the further limitation wherein the global sequence of the client computing devices generated by a computational method comprising a random or pseudorandom mathematical or statistical algorithm.
P. 6180, algorithm 1, “St ← (random set of m users)”
P. 6182, first paragraph, “The server sends the token to the first participant (named user 1) selected randomly from the neighbor users in the current round.”
Regarding claim 5, Li discloses the following further limitations:
a) generate a set of initial parameters associated with a machine learning-based model;
P. 6180, algorithm 1 (reproduced infra): “initialize wIG // Global Model Parameters”.
b) generate a parameter Number of partitions;
P. 6180, table I, “D the number of data-partitions by Pd”.
c) generate a parameter Minimum number of data points in the partitions; and
P. 6180, table I, nd the number of data-samples in each data-partition”.
The Examiner interprets “minimum number of data points in the partitions” according to its broadest reasonable interpretation as encompassing the specification of an exact value for data points per partition, which is effectively a minimum and a maximum value.
d) generate a parameter Random sequence of client computing devices.
P. 6180, algorithm 1, “St ← (random set of m users)”
P. 6182, first paragraph, “The server sends the token to the first participant (named user 1) selected randomly from the neighbor users in the current round.”
Regarding claim 6, Li discloses a computer-implemented method for computationally efficient machine learning, comprising:
a) generating, by a server computing device, a set of initial model parameters;
P. 6180, algorithm 1 (reproduced below): “initialize wIG // Global Model Parameters”.
PNG
media_image2.png
502
602
media_image2.png
Greyscale
b) generating, by the server computing device, a set of partitioning parameters including at least the parameters Number of partitions and Minimum number of data points in the partitions;
P. 6180, algorithm 1, “FOR each user k ∈ St …”
P. 6180, “If there are D data partitions selected with nd = |Pd|,”
c) communicating, by the server computing device, the set of initial model parameters and the partitioning parameters to each of a plurality of client computing devices;
P. 6180, “FL typically aims to minimize the empirical risk over heterogeneous data distributed across multiple devices [37]. Before the start of training every iteration, each participating user gets the global model parameters from the aggregation node (or the server). Then, the user uses the global model parameters to train a local model with the local data set and sends new local model parameters as the local update to the aggregation node.” (Emphasis added.)
d) obtaining, by each client computing device, global values corresponding to the set of initial parameters from the server computing device;
Id.
e) selecting, by the server computing device, a client computing device based on a global sequence of all client computing devices;
P. 6180, algorithm 1, “FOR each user k ∈ St …”
f) communicating, by the server computing device, machine learning-based model parameters to the selected client computing device;
P. 6182, “P. 6182, algorithm 2: “initialize and broadcast wIG // Global Model Parameters”.
g) obtaining, by the selected client computing device, global values corresponding to the machine learning-based model parameters;
Id.
h) incrementally training, by the selected client computing device, the machine-learned model based, at least in part, on a partition of the local dataset to obtain updated machine learning-based model parameters, said partition being selected from the local dataset according to a predefined scheme;
P. 6180, “FL typically aims to minimize the empirical risk over heterogeneous data distributed across multiple devices [37]. Before the start of training every iteration, each participating user gets the global model parameters from the aggregation node (or the server). Then, the user uses the global model parameters to train a local model with the local data set and sends new local model parameters as the local update to the aggregation node.” (Emphasis added.)
i) communicating, by the selected client computing device, the updated machine learning-based model parameters to the server computing device;
P. 6180, “FL typically aims to minimize the empirical risk over heterogeneous data distributed across multiple devices [37]. Before the start of training every iteration, each participating user gets the global model parameters from the aggregation node (or the server). Then, the user uses the global model parameters to train a local model with the local data set and sends new local model parameters as the local update to the aggregation node.” (Emphasis added.)
j) repeating steps d)-i) until all predefined conditions are satisfied.
P. 6178, second col., “For each round of iteration …”.
P. 6180, algorithm 1, “FOR round t = 1, 2, . . .”
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103(a) which forms the basis for all obviousness rejections set forth in this Office action:
(a) A patent may not be obtained though the invention is not identically disclosed or described as set forth in section 102 of this title, if the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter pertains. Patentability shall not be negatived by the manner in which the invention was made.
The following references are relied upon in the rejections below:
Li (Li et al., "Privacy-Preserving Federated Learning Framework Based on Chained Secure Multiparty Computing", IEEE INTERNET OF THINGS JOURNAL, VOL. 8, NO. 8, APRIL 15, 2021, pages 6178-6186, XP011848022. Cited by Applicant in IDS dated 10/18/24.)
Zhu (US 2022/0114475 A1)
Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Li and Zhu.
Regarding claim 2, Li discloses the further limitations wherein generating the partition of the local dataset comprises:
a) obtaining global values corresponding to a set of parameters associated with the Number of partitions to be generated and the Minimum number of data points in the partitions; and
P. 6180, table I, “D the number of data-partitions by Pd”.
P. 6180, table I, “nd the number of data-samples in each data-partition”.
The Examiner interprets “minimum number of data points in the partitions” according to its broadest reasonable interpretation as encompassing the specification of an exact value for data points per partition, which is effectively a minimum and a maximum value.
b) … wherein one partition has at least a number smin of data points, wherein the value of smin is defined by the parameter Minimum number of data points in the partitions.
Id.
Zhu discloses the following further limitation which Li does not disclose:
b) dividing the local datasets into a plurality (n) of partitions defined by the parameter Number of partitions...
[0005] “In another example, instead of dividing the local dataset D.sub.j into minibatches of data ξ.sub.j and training the local model 104 based on all minibatches, the training may be performed using only one minibatch.”
At the time of filing, it would have been obvious to a skilled machine learning engineer to apply the technique disclosed by Zhu for dividing local datasets to the Li system because this would provide for faster training iterations, and could also provide for data with witch to test the trained model.
Additional Relevant Prior Art
The following references were identified by the Examiner as being relevant to the disclosed invention, but are not relied upon in any rejection:
Jakkam Reddi discloses a federated learning system with training set batching according to a batch size parameter. See eg [0114]. (US 2021/0073639 A1)
Conclusion
Information regarding the status of an application may be found at the USPTO Patent Center at https://patentcenter.uspto.gov. Any inquiry concerning this communication should be directed to Vincent Gonzales at (571) 270-3837. The examiner can normally be reached Monday-Friday 7 am to 4 pm MT. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Miranda Huang, can be reached at (571) 270-7092.
/Vincent Gonzales/Primary Examiner, Art Unit 2124