Channel State Feedback Neural Networks with Classifier Weight Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In wireless communication systems, rapid changes in channel conditions due to UE transitions (e.g., waking up from sleep or handovers) lead to increased latency as previously trained neural networks become irrelevant, necessitating re-training of autoencoders at the UE.

Innovation Solution

A UE requests and selects neural network weights from multiple sets based on channel conditions using a classifier neural network, updating the selected weights with measurements and transmitting an identifier to network entities for future use.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If neural networks are re-trained at the UE when channel conditions change, then the neural network remains relevant for current conditions, but latency increases due to re-training time

Engineering Contradiction:

Improverelevance of neural network to current channel conditionsVSAvoidlatency during re-training

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The network entity pre-trains multiple neural networks for different channel conditions before the UE needs them. When the UE experiences channel changes, it receives pre-trained weights corresponding to the new conditions without needing to re-train, thus eliminating re-training latency while maintaining relevance.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Instead of re-training the neural network structure, the system changes the parameter set (weights) of the neural network based on detected channel conditions. The UE selects from multiple pre-configured weight sets corresponding to different channel scenarios, allowing rapid adaptation without time-consuming re-training.

Inventive Principle:

Principle #35Parameter changes

2Loss of time

If multiple sets of neural network weights are stored and selected based on channel conditions, then latency is reduced, but device complexity increases

Engineering Contradiction:

Improvelatency during neural network adaptationVSAvoidcomplexity of managing multiple weight sets

Core Design Contradiction:

Loss of timeVSDevice complexity

Solution Approach 1:

A classifier neural network acts as an intermediary between the channel condition detector and the multiple weight sets. The classifier automatically determines which pre-trained weight set corresponds to the current channel conditions, simplifying the selection process and reducing UE complexity while maintaining low latency.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system segments the neural network adaptation problem into distinct channel condition categories, each with its own pre-trained weight set. This segmentation allows the UE to select from discrete, manageable weight sets rather than managing a continuous space of possible networks, reducing complexity.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12445872B2Online training and augmentation of neural networks for channel state feedback

Publication Date: 2025.10.14 QUALCOMM INC

AI summary

Methods, systems, and devices for wireless communications are described. A user equipment (UE) may transmit a request for sets of neural network weights, each set corresponding to a neural network, to one or more network entities. The one or more network entities may transmit multiple sets of neural network weights to the UE in response to the request. The UE may run a classifier neural network on the sets of neural network weights to select a set of neural network weights corresponding to a neural network. The UE may transmit a neural network identifier to the one or more network entities that transmitted neural network weights.