Neural Network Model Generation via Genetic Algorithm Module Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model generation techniques are limited in flexibility, particularly in connecting modules in series, and lack the ability to generate models with more complex configurations involving multiple blocks.
Innovation Solution
An information processing method that uses a genetic algorithm to optimize the generation of models by connecting inputs from one module to another, allowing for flexible configuration and optimization of model parameters based on learning data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If models are generated with layers connected in series using conventional techniques, then model generation is straightforward, but model flexibility and configurability are limited
Solution Approach 1:
The patent applies dynamics by making the model architecture configurable and adaptable through genetic algorithms. The system dynamically optimizes model parameters, block configurations, and module connections based on learning data, transforming static series connections into flexible, optimized architectures that can adapt to different tasks and data types.
Solution Approach 2:
The patent changes parameters by using genetic algorithms to optimize various model parameters including block configurations, module connections, and hyperparameters. This allows the system to explore different architectural configurations and select the optimal parameters for each specific application, thereby improving model flexibility without excessive complexity.
2Adaptability or versatility
If models use fixed series connections between layers, then implementation is simple, but the ability to handle complex multi-block configurations is insufficient
Solution Approach 1:
The patent applies self-service by enabling the model generation system to automatically configure and optimize its own architecture using genetic algorithms. The system self-adjusts block configurations, module connections, and parameters based on learning data performance, eliminating the need for manual configuration while handling complex multi-block architectures effectively.
Solution Approach 2:
The patent implements feedback mechanisms where the genetic algorithm continuously evaluates model performance on learning data and uses this feedback to iteratively optimize the architecture. This feedback loop enables the system to handle complex multi-block configurations by learning from performance results and adjusting configurations accordingly, making complex model generation as easy as providing the learning data.
3Manufacturing precision
If conventional model generation techniques are used, then generation process is fast, but model accuracy and adaptability are suboptimal
Solution Approach 1:
The patent applies preliminary action by pre-defining a set of available blocks and modules that can be configured through genetic algorithms. This preparation enables the system to quickly generate accurate models by selecting and configuring from pre-established components, rather than creating everything from scratch, thus maintaining generation speed while improving accuracy through optimized configurations.
Solution Approach 2:
The patent substitutes mechanical model generation approaches with genetic algorithm-based optimization. Instead of manually or systematically generating models through fixed procedures, the system uses evolutionary computation to explore and optimize model architectures, achieving higher accuracy through adaptive search while maintaining reasonable generation speeds through efficient algorithm design.
Data Source
AI summary
An information processing method according to the present application is an information processing method executed by a computer, the information processing method including: acquiring learning data used for learning of a model having a plurality of blocks each including at least one module; and generating the model in which an input to one module is connected as an input to another module by learning using the learning data.