Federated Learning Prediction Model Accuracy via Parameter Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing federated learning techniques for predicting user interest in commodities struggle with low accuracy, failing to generate appropriate information for receivers.
Innovation Solution
A federated learning apparatus and method that trains a prediction model using evaluation values and attribute values, integrates parameter information from multiple models, and updates the model with integrated parameter information to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If federated learning is used to predict user interest in commodities, then information concealment is ensured, but prediction accuracy is low
Solution Approach 1:
The patent segments the prediction model into multiple independent client models, each trained on local user-commodity interaction data. This segmentation enables distributed training while maintaining data privacy, resolving the contradiction between information concealment and prediction accuracy by allowing each client to maintain their own model parameters without sharing raw data.
Solution Approach 2:
The patent merges the predictions from multiple client models through a server that aggregates results from all clients. This combining approach leverages the diverse data and perspectives from multiple clients to improve overall prediction accuracy while maintaining the privacy benefits of distributed learning, as each client's data remains local but contributes to the collective intelligence.
2Adaptability or versatility
If existing federated learning techniques are used, then collaborative filtering is achieved, but the generated information is not appropriate for receivers
Solution Approach 1:
The patent implements local quality by training separate prediction models at each client tailored to their specific user-base and commodity inventory. Each client's model captures local patterns and preferences specific to their audience, ensuring that the generated recommendations are highly relevant to local receivers rather than applying a one-size-fits-all approach.
Solution Approach 2:
The patent incorporates feedback mechanisms where clients receive aggregated predictions from the server and can adjust their local models accordingly. This feedback loop enables continuous improvement of recommendation relevance by learning from aggregate results and refining local models to better match receiver preferences and behaviors.
Data Source
AI summary
To generate information appropriate for a receiver of the information, a federated learning apparatus includes: a training section which trains a first prediction model that predicts an evaluation value corresponding to a combination of a user and an evaluation target on which the evaluation value is not obtained, using a first training data set including (i) evaluation values of users on evaluation targets and (ii) attribute values of the evaluation targets; a parameter information transmitting section which transmits, to a server apparatus, first parameter information indicating the first prediction model; a parameter information obtaining section which obtains, from the server apparatus, integrated parameter information obtained by integrating the first parameter information and second parameter information indicating a second prediction model trained using a second training data set; and an updating section which updates the first prediction model by replacing the first parameter information with the integrated parameter information.