DETAILED ACTION
The present application is being examined under the pre-AIA first to invent provisions.
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 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 1-20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Balakrishnan et al. (US 20230177349 A1).
Considering claim 1, Balakrishnan teaches a communication method, comprising:
obtaining, by a central node (120, 1208, Fig.1, 12), N indication parameters (capability data, model updates or other parameters) that are in a one-to-one correspondence with N distributed nodes (K clients, Fig.1, [0031] data sources 160 (e.g., autonomous vehicles 161, user equipment 162, business and industrial equipment 163, video capture devices 164, drones 165, smart cities and building devices 166, sensors and IoT devices 167, etc.), [0123]), wherein a first indication parameter in the N indication parameters indicates a communication capability and a computing capability of a first distributed node, and the first distributed node is any distributed node in the N distributed nodes (Fig.1, [0031], [0123] clients providing capability data, model updates or other parameters to a central server, but such capability data, model updates, or parameters may be provided by the clients to different central servers);
determining, by the central node (120, 1208, Fig,1, 12, [0191] server can then determine to use the maximum value of the T.sub.ref values), M network parameters based on the N indication parameters (Fig.15, [0123], [0154] central server sends global model weights w.sub.t to the selected K clients, [0212]), wherein the M network parameters are in a one-to-one correspondence with M distributed nodes in the N distributed nodes (K clients, Fig.15, [0145], [0191], and a first network parameter in the M network parameters indicates at least one of a communication configuration or neural network architecture configuration of the first distributed node (Fig.1, 12, [0121], [0212], [0214]); and
sending, by the central node, the M network parameters (Fig.15, [0145] central server sends global model weights w.sub.t to the selected K clients), wherein N and M are positive integers, and M is less than or equal to N (Fig.15, [0145], [0211], [0224]).
Considering claim 4, Balakrishnan teaches a communication method, comprising:
sending (client/compute node), by a first distributed node ([0123],) a first indication parameter, wherein the first indication parameter indicates a communication capability and a computing capability of the first distributed node ([0123] providing capability data, model updates or other parameters to a central server, but such capability data, model updates, or parameters may be provided by the clients to different central servers;
receiving, by the first distributed node, a first network parameter (Fig.15, [0154], [0192], server may send the T.sub.ref value to the clients), wherein the first network parameter indicates at least one of a communication configuration or a neural network architecture configuration of the first distributed node ([0115] configurations, capabilities, and operate under different conditions, [0192], [0244] central server may send a broadcast METADATA request to the clients, and the clients respond with METADATA response indicating one or more of the following information:…); and
determining, by the first distributed node ([0139] determined by the client at 1304), at least one of the communication configuration or the neural network architecture configuration based on the first network parameter (Fig.1, 22,[0121] parameters, [0244]-[0245]).
Considering claim 13, Balakrishnan teaches a communication apparatus, comprising:
a transceiver (866, Fig.8), configured to obtain N indication parameters that are in a one-to-one correspondence with N distributed nodes (Fig.1, 12, 14-15, [0031], [0123] clients providing capability data, model updates or other parameters to a central server, but such capability data, model updates, or parameters may be provided by the clients to different central servers), , wherein a first indication parameter in the N indication parameters indicates a communication capability and a computing capability of a first distributed node, and the first distributed node is any distributed node in the N distributed nodes (clients, Fig.1, 12, 14-15, [0031], [0123], [0154]); and
one or more processors (862, Fig.8), configured to determine M network parameters based on the N indication parameters (Fig.15, [0123], [0154] central server sends global model weights w.sub.t to the selected K clients, [0191],[0212]), wherein the M network parameters are in a one-to-one correspondence with M distributed nodes in the N distributed nodes (Fig.15, [0145], [0154], [0191)], and
a first network parameter in the M network parameters indicates at least one of a communication configuration or a neural network architecture configuration of the first distributed node (Fig.14-15, [0145] central server sends global model weights w.sub.t to the selected K clients), wherein the transceiver is further configured to send the M network parameters, and wherein N and M are positive integers, and M is less than or equal to N (Fig.15, [0145], [0211], [0224]).
Considering claims 2, 14, Balakrishnan teaches wherein the N indication parameters are further used to determine training parameters corresponding to the M distributed nodes (Fig.14-15, [0219]), the training parameters comprise a first training parameter (Fig.14-15, [0219] federated learning may be utilized where devices are chosen in each training round to participate in the federated update. Specifically, K clients can be scheduled in each training epoch/round to send gradient updates (e.g., as described above) along with other hyper-parameters (e.g., the learning rate, number of local iterations, redundancy, transmit power and bandwidth for the clients), [0228]), and the first training parameter indicates a training order in which the first distributed node trains a neural network ([0219], [0193] neural network).
Considering claims 3, 15, Balakrishnan teaches sending, by the central node, the training parameters (Fig.14-15, [0154]).
Considering claim 5, Balakrishnan teaches wherein the method further comprises:
receiving, by the first distributed node (client), a first training parameter, wherein the first training parameter indicates a training order in which the first distributed node trains a neural network (Fig.14-15, [0123], [0154]).
Considering claims 6, 16, Balakrishnan teaches wherein the first indication parameter (clients providing capability data, model updates or other parameters) further indicates a service characteristic of the first distributed node ([0123], [0142, [0211] determines compute hyper-parameters (learning rate) along with the communication parameters (radio resources to be allocated to the clients) to satisfy certain performance metrics).
Considering claims 7, 17, Balakrishnan teaches wherein the first indication parameter (clients providing capability data, model updates or other parameters) comprises first communication capability information, first computing capability information, and first service characteristic information, wherein the first communication capability information indicates the communication capability of the first distributed node, the first computing capability information indicates the computing capability of the first distributed node, and the first service characteristic information indicates the service characteristic of the first distributed node ([0123], [0142, [0211] determines compute hyper-parameters (learning rate) along with the communication parameters (radio resources to be allocated to the clients).
Considering claims 8, 18, Balakrishnan teaches wherein the first communication capability information (clients providing capability data, model updates or other parameters) comprises at least one of the following: channel characteristic information or communication resource information of a channel between the first distributed node and the central node ([0123], [0142, [0211] determines compute hyper-parameters (learning rate) along with the communication parameters (radio resources to be allocated to the clients); or channel characteristic information or communication resource information of a channel between the first distributed node and a second distributed node, wherein the second distributed node is a distributed node having a connection relationship with the first distributed node.
Considering claims 9, 19, Balakrishnan teaches wherein the channel characteristic information (link quality estimation, data quality) along with the communication parameters (radio resources to be allocated to the clients) comprises at least one of the following: channel state information, a signal-to-noise ratio, link quality, or a location of the first distributed node ([0211], [0214]).
Considering claims 10, 20, Balakrishnan teaches wherein the first network parameter (link quality estimation, data quality) comprises a first communication network parameter and a first neural network parameter, the first communication network parameter indicates the communication configuration, and the first neural network parameter indicates the neural network architecture configuration (Fig.14-15, [0033], [0214], [0245]).
Considering claim 11, Balakrishnan teaches wherein the first communication network parameter (link quality estimation, data quality) comprises at least one of the following: a scheduling parameter or a communication mode ([0214], [0219]).
Considering claim 12, Balakrishnan teaches wherein the first neural network parameter comprises at least one of the following: a quantity of layers of the neural network architecture, an operation corresponding to each of the layers of the neural network architecture, or a weight value of a neural network corresponding to the neural network architecture (Fig.14-15, [0154], [0214], [0245]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KHAI MINH NGUYEN whose telephone number is (571)272-7923. The examiner can normally be reached 6-3.
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, Charles Appiah can be reached at 571-272-7904. 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.
/KHAI M NGUYEN/Primary Examiner, Art Unit 2641
Prosecution Insights