Genetic Algorithm Hyperparameter Optimization for Learning Models

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing learning methods struggle to improve model accuracy due to dynamic changes in learning data based on hyperparameter values, leading to suboptimal learning modes and feature selections.

Innovation Solution

The proposed solution involves an information providing device that generates a generation index using a genetic algorithm to optimize hyperparameters and feature selections, thereby improving model accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If learning data is dynamically changed according to hyperparameter values, then the learning process becomes more flexible, but the model accuracy deteriorates due to inappropriate hyperparameter selections

Engineering Contradiction:

Improvelearning mode flexibilityVSAvoidmodel accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent implements a feedback mechanism where the evaluation unit assesses model accuracy for each hyperparameter combination, and this evaluation information feeds back to the generation unit to guide the creation of improved hyperparameter sets through genetic algorithms, resolving the contradiction between flexibility and accuracy

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent systematically changes hyperparameter values across multiple combinations and evaluates each combination's effect on model accuracy, allowing the system to identify optimal parameter settings that maintain flexibility while improving accuracy

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If multiple elements (learning data, features, learning mode) are varied to improve model accuracy, then the potential for accuracy improvement increases, but the difficulty of selecting the optimum element increases

Engineering Contradiction:

Improvemodel accuracyVSAvoidselection complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs self-service by automatically evaluating multiple hyperparameter combinations and selecting the optimal set without requiring manual intervention, thereby reducing selection complexity while maintaining the ability to improve accuracy through multiple variable elements

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent systematically varies multiple parameters (hyperparameters, learning data, features) and evaluates their combined effect on model accuracy, allowing the system to navigate the complex selection space automatically and identify optimal configurations

Inventive Principle:

Principle #35Parameter changes

3Ease of operation

If hyperparameter values are not appropriately selected, then the learning process can proceed flexibly, but the model accuracy deteriorates

Engineering Contradiction:

Improvelearning process easeVSAvoidmodel accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The evaluation unit provides feedback on model accuracy for each hyperparameter combination, enabling the system to automatically identify and select appropriate hyperparameter values without manual tuning, thus maintaining ease of operation while improving accuracy

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12321836B2Learning device, learning method, learning program, evaluation device, evaluation method, and evaluation program

Publication Date: 2025.06.03 ACTAPIO INC

AI summary

A learning device according to the present application includes a generation unit that generates, from a plurality of values indicating features of a predetermined target and indicating different types of a plurality of features, a value corresponding to a set of the types of the features, and a learning unit that causes a model to learn a feature of the predetermined target using a value generated by the generation unit.