End-to-end deep neural network adaptation for edge computing

Patent metadata

Field	Value
Publication number	US-12526202-B2
Application number	US-202117998323-A
Country	US
Kind code	B2
Filing date	May 21, 2021
Priority date	Jun 5, 2020
Publication date	Jan 13, 2026
Grant date	Jan 13, 2026

Abstract

Techniques and apparatuses are described for adapting an end-to-end, E2E, machine-learning, ML, configuration for processing communications transferred through an E2E communication. A network entity directs a user equipment (UE) and a base station participating in the E2E communication to implement the E2E communication by forming at least a portion of an E2E deep neural network, DNN, based on a first E2E ML configuration. The network entity determines to update the first E2E ML configuration based on a change in a participation mode of an edge compute server (ECS) in the E2E communication. The network entity identifies a second E2E ML configuration based on the change in participation mode and directs the UE or the base station to update the portion of the E2E DNN using the second E2E ML configuration.

First claim

What is claimed is:

1 . A method performed by a network entity for adapting an end-to-end (E2E) machine-learning (ML) configuration that forms an E2E deep neural network (DNN), by determining adjustments to one or more E2E ML configurations, for processing communications transferred through an E2E communication between at least two endpoints, the E2E communication using a wireless network, the method comprising:
directing a user equipment (UE) participating in the E2E communication to implement the E2E communication by forming at least a first portion of the E2E DNN based on a first E2E ML configuration;
directing a base station participating in the E2E communication to implement the E2E communication by forming at least a second portion of the E2E DNN based on the first E2E ML configuration;
determining to update the first E2E ML configuration based on a change in a participation mode of an edge compute server (ECS) in the E2E communication;
identifying a second E2E ML configuration based on the change in participation mode of the ECS in the E2E communication; and
directing at least the UE or the base station to update at least a third portion of the E2E DNN using the second E2E ML configuration for implementing the E2E communication.

2 . The method as recited in claim 1 , wherein the determining to update the first E2E ML configuration comprises:
determining to include the ECS in the E2E communication; and determining to update the first E2E ML configuration based on determining to include the ECS in the E2E communication.

3 . The method as recited in claim 2 , wherein the determining to update the first E2E ML configuration based on determining to include the ECS further comprises:
determining to update the first E2E ML configuration based on:
aggregating communications with the ECS and communications with a remote service in the E2E communication; or
excluding the remote service from the E2E communication.

4 . The method as recited in claim 3 , wherein the determining to update the first E2E ML configuration comprises determining to update the first E2E ML configuration based on the aggregating, and
the identifying the second E2E ML configuration comprises:
identifying, as at least part of the second E2E ML configuration, a downlink E2E ML configuration that forms a downlink E2E DNN directed to:
receive a first portion of application data from the ECS;
receive a second portion of the application data from the remote service; and
aggregate the first portion and the second portion to generate aggregated application data directed to the UE.

5 . The method as recited in claim 3 , wherein the determining to update the first E2E ML configuration comprises determining to update the first E2E ML configuration based on the aggregating, and
wherein the identifying the second E2E ML configuration comprises:
identifying, as at least part of the second E2E ML configuration, an uplink E2E ML configuration that forms an uplink E2E DNN directed to:
receive uplink application data from the UE;
generate, using the uplink application data, a first output directed to the ECS; and
generate, using the uplink application data, a second output directed to the remote service.

6 . The method as recited in claim 1 , wherein the determining to update the first E2E ML configuration further comprises:
receiving a request from the UE to include the ECS in the E2E communication; or determining to include the ECS in the E2E communication based on an estimated-UE location.

7 . The method as recited in claim 1 , wherein the directing at least the UE or the base station to update the at least a third portion of the E2E DNN using the second E2E ML configuration further comprises:
directing the UE to update the first portion of the E2E DNN using the second E2E ML configuration; or directing the base station to update the second portion of the E2E DNN using the second E2E ML configuration.

8 . The method in in claim 1 , wherein the identifying of the second E2E ML configuration comprises at least one of:
identifying one or more parameter changes to the E2E DNN; or identifying one or more architecture changes to the E2E DNN.

9 . A method performed by a wireless transmit/receive unit (WTRU) for adapting an end-to-end (E2E) machine-learning (ML) configuration, by determining adjustments to one or more E2E ML configurations, for processing communications transferred through an E2E communication in a wireless network, the method comprising:
forming, based on a first E2E ML configuration identified by a network entity, at least a first portion of an E2E deep neural network (DNN) that implements an E2E communication; receiving an indication to update the E2E DNN using at least a second portion of a second E2E ML configuration based on a change in a participation mode of an edge compute server (ECS) in the E2E communication; updating the E2E DNN using the at least a second portion of the second E2E ML configuration; and implementing at least a portion of the E2E communication using the updated E2E DNN.

10 . The method as recited in claim 9 , further comprising:
identifying, based on an estimated-UE location of the WTRU, the edge computing server; and requesting to include the ECS in the E2E communication.

11 . The method as recited in claim 9 , wherein the receiving of the indication to update the E2E DNN further comprises:
receiving, as the indication, directions to update one or more parameters of the DNN; or receiving, as the indication, directions to update an architecture of the DNN.

12 . The method as recited in claim 11 , wherein the receiving of the directions to update the architecture further comprises at least one of:
changing a number of processing layers used in the E2E DNN; and changing a computation mode of at least one processing layer in the E2E DNN.

13 . A network entity for adapting an end-to-end (E2E) machine-learning (ML) configuration that forms an E2E deep neural network (DNN), by determining adjustments to one or more E2E ML configurations, for processing communications transferred through an E2E communication between at least two endpoints, the E2E communication using a wireless network, the network entity comprising:
a processor; and computer-readable storage media comprising instructions, executable by the processor to implement an end-to-end machine-learning controller to:
direct a user equipment (UE) participating in an end-to-end (E2E) communication to implement the E2E communication by forming at least a first portion of an E2E deep neural network (DNN) based on a first E2E machine-learning (ML) configuration;
direct a base station participating in the E2E communication to implement the E2E communication by forming at least a second portion of the E2E DNN based on the first E2E ML configuration;
determine to update the first E2E ML configuration based on a change in a participation mode of an edge compute server (ECS) in the E2E communication;
identify a second E2E ML configuration based on the change in participation mode of the ECS in the E2E communication; and
direct at least the UE or the base station to update at least a third portion of the E2E DNN using the second E2E ML configuration for implementing the E2E communication.

14 . A user equipment for adapting an end-to-end (E2E) machine-learning (ML) configuration, by determining adjustments to one or more E2E ML configurations, for processing communications transferred through an E2E communication in a wireless network, the user equipment comprising:
a processor; and computer-readable storage

Assignees

• Google Llc

Inventors

• Wang Jibing
• Stauffer Erik Richard

Classifications

• H04W84/042

  Public Land Mobile systems, e.g. cellular systems · CPC title

• H04L67/10

  in which an application is distributed across nodes in the network (software deployment G06F8/60; multiprogramming arrangements G06F9/46) · CPC title

• H04L47/18

  End to end · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title