US 2020/0027022 A1Published application

Distributed machine learning in an information centric network

AssigneeIntel Corporation

InventorsSatish Chandra Jha, S M Iftekharul Alam, Ned M. Smith

Life of this patent

Estimating the term…

Loading patent…

Abstract

Systems and techniques for distributed machine learning (DML) in an information centric network (ICN) are described herein. Finite message exchanges, such as those used in many DML exercises, may be efficiently implemented by treating certain data packets as interest packets to reduce overall network overhead when performing the finite message exchange. Further, network efficiency in DML may be improved achieved by using local coordinating nodes to manage devices participating in a distributed machine learning exercise. Additionally, modifying a round of DML training to accommodate available participant devices, such as by using a group quality of service metric to select the devices, or extending the round execution parameters to include additional devices, may have an impact on DML performance.

Claims24

Dependent claims

An information centric network (ICN) node for distributed machine learning protocol in an ICN, the ICN node comprising:

a first network interface to connect to an upstream segment of the ICN;
a second network interface to connect to a downstream segment of the ICN;
and
processing circuitry to:
receive, at the first network interface, a participant solicitation interest packet, the solicitation interest packet including a name identifying the solicitation interest packet as a distributed machine learning (DML) participant solicitation and a set of parameters to participate in a round of DML;
create a first pending interest table (PIT) entry for the participant solicitation interest packet, the PIT entry being maintained until the expiration of a timer, the duration of which is specified in the solicitation interest packet or derived from parameters of the solicitation interest packet;
receive, at the second network interface, a first data packet in response to the solicitation interest packet from a downstream node, the response indicating that the downstream node will participate in the round of DML;
create a second PIT entry for the first data packet in response to receiving the data packet, the second PIT entry corresponding to the first PIT entry;
transmit, on the first network interface, the first data packet in accordance with the first PIT entry;
receive, at the first network interface, a second data packet in response to the first data packet, wherein the second data packet includes an identification of a DML model for the downstream node;
and
transmit, on the second network interface, the second data packet in accordance with the second PIT entry.

9 more elements

At least one machine-readable medium including instructions for distributed machine learning protocol in an information centric network (ICN), the instructions, when executed by processing circuitry, cause the processing circuitry to perform operations comprising:

receiving a participant solicitation interest packet, the solicitation interest packet including a name identifying the solicitation interest packet as a distributed machine learning (DML) participant solicitation and a set of parameters to participate in a round of DML;
creating a first pending interest table (PIT) entry for the participant solicitation interest packet, the PIT entry being maintained until the expiration of a timer, the duration of which is specified in the solicitation interest packet or derived from parameters of the solicitation interest packet;
receiving a first data packet in response to the solicitation interest packet from a downstream node, the response indicating that the downstream node will participate in the round of DML;
creating a second PIT entry for the first data packet in response to receiving the data packet, the second PIT entry corresponding to the first PIT entry;
transmitting the first data packet in accordance with the first PIT entry;
receiving a second data packet in response to the first data packet, wherein the second data packet includes an identification of a DML model for the downstream node;
and
transmitting the second data packet in accordance with the second PIT entry.

5 more elements

Dependencies are inferred from the claim text.