Federated Learning Contributor Usefulness Evaluation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Federated learning systems face challenges in determining when to proceed with the learning process due to varying contributor availability and the differing usefulness of data samples, leading to inefficiencies and potential loss of important contributions.

Innovation Solution

A method to identify currently available contributors, determine their usefulness based on metrics such as contribution impact and dataset diversity, and decide whether to perform the federated learning process or wait for more contributors, considering predicted availability and thresholds.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the server waits for contributors to reconnect to ensure their data samples are used, then the accuracy of the global model is improved, but the time required for the federated learning process increases

Engineering Contradiction:

Improveaccuracy of global modelVSAvoidtime required for federated learning process

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by tracking contributor connection status and data sample availability before initiating the federated learning process. The server proactively identifies when contributors are available and their data samples are ready, rather than waiting passively for reconnection events to occur during the learning process.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms where contributors notify the server of their connection status and data sample availability. This feedback loop allows the server to make informed decisions about when to proceed with federated learning, balancing the need for accurate models with efficient use of time.

Inventive Principle:

Principle #23Feedback

2Productivity

If the server proceeds with federated learning using currently connected contributors, then the productivity of the system is improved, but the accuracy of the global model deteriorates due to loss of important contributions

Engineering Contradiction:

Improveefficiency of federated learning processVSAvoidaccuracy of global model

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system dynamically adjusts the decision to proceed with federated learning based on real-time conditions. The server evaluates the connection status of contributors, the availability of data samples, and the potential impact on model accuracy, making flexible decisions about when to initiate or postpone learning rounds.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes parameters such as the threshold for proceeding with federated learning based on the current state of contributor availability and data sample quality. When important contributors are unavailable, the system adjusts its parameters to wait for their reconnection, whereas when contributors are available with sufficient data, it proceeds promptly.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If the server waits for contributors with high usefulness data samples, then the quality of the global model is improved, but the time required for data collection increases

Engineering Contradiction:

Improvequality of global modelVSAvoidtime required for data collection

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system replaces mechanical waiting with an intelligent evaluation mechanism. Instead of simply waiting for contributors to reconnect, the server uses a usefulness determination process that evaluates the potential value of data samples from available contributors, substituting passive time consumption with active decision-making.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameter of waiting time based on the determined usefulness of contributors. When a contributor's data samples are determined to be highly useful, the server may extend the waiting period for their reconnection. When usefulness is low, the system reduces waiting time and proceeds with available data, dynamically adjusting the time parameter based on quality assessment.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11461593B2Federated learning of clients

Publication Date: 2022.10.04 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A method, a computer program product, and a computer system determine when to perform a federated learning process. The method includes identifying currently available contributors among contributors of a federated learning task for which the federated learning process is to be performed. The method includes determining a usefulness metric of the currently available contributors for respective datasets from each of the currently available contributors used in performing the federated learning process. The method includes, as a result of the usefulness metric of the currently available contributors being at least a usefulness threshold, generating a recommendation to perform the federated learning process with the datasets of the currently available contributors. The method includes transmitting the recommendation to a processing component configured to perform the federated learning process.