Information Generation Device for Optimization Problem Search

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing machine learning models are not designed to address combinatorial optimization problems, limiting their application in decision-making processes.

Innovation Solution

An information generation device and method that accepts input of optimization problems, including objective functions and constraints, to generate search information associating these data elements, enabling the search for optimization solutions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning models are used for predictive tasks, then prediction accuracy is improved, but applicability to optimization problems deteriorates

Engineering Contradiction:

Improveprediction accuracyVSAvoidapplicability to optimization problems

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent extends machine learning models to perform both predictive tasks and optimization tasks through a unified framework. The system accepts optimization problems with objective functions and constraints, generates search information by associating problem data with features, and enables the same model architecture to handle both prediction and combinatorial optimization, thereby achieving multi-functionality

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent transforms the application parameters of machine learning models by introducing optimization-specific parameters such as objective functions and constraints. By changing how the models process and utilize parameters, the system enables optimization capabilities while maintaining the core predictive functionality, allowing models to adapt to different task types through parameter transformation

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If machine learning models are specialized for prediction, then prediction performance is improved, but flexibility for decision-making deteriorates

Engineering Contradiction:

Improveprediction performanceVSAvoidflexibility for decision-making

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal machine learning framework that can perform both prediction and decision-making through optimization. The system maintains specialized prediction capabilities while adding optimization functionality by introducing a unified interface that accepts both predictive queries and optimization problems, allowing the same model to serve multiple decision-making needs

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces search information generation as an intermediary process between the optimization problem and the machine learning model. This intermediary layer transforms optimization problems into a format that the predictive model can process, enabling flexible decision-making while preserving the model's core prediction performance through the mediating transformation step

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20240320570A1Information generation device, information generation method, and information generation program

Publication Date: 2024.09.26 NEC CORP

AI summary

The input means 81 accepts input of first data indicating an optimization problem including an objective function and constraints, and second data indicating a feature of the optimization problem. The generation means 82 generates search information that associates the first data with the second data.