Learning Device for Optimization Problem Solution Selection

Solution Overview

Problem

Users face challenges in determining the final solution for optimization problems as they often need to compare multiple solutions, and existing systems do not effectively allow for presenting solutions with variations based on user preferences.

Innovation Solution

A learning device and method that generate multiple sets of solutions for an optimization problem and train a model to determine the preferred solutions based on user input, allowing for presentation of solutions with variations tailored to user preferences.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a single optimal solution is generated by optimization, then the solution satisfies constraint conditions, but the user cannot compare multiple solutions to determine the final solution to adopt

Engineering Contradiction:

ImproveAbility to present multiple solutions with variationsVSAvoidComplexity of solution presentation system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system performs preliminary action by generating multiple candidate solutions in advance through the solution generation unit, and pre-training the learning model with user preference data before actual solution presentation. This allows the system to have multiple solutions ready for comparison and can quickly present appropriate solutions when needed, rather than generating solutions on-demand which would be time-consuming.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements feedback by using the learning unit to train a model based on user selections and preferences. When users interact with presented solutions (selecting preferred ones or indicating preferences), this feedback is used to retrain and improve the learning model, enabling the system to progressively better understand and adapt to user preferences for solution variations.

Inventive Principle:

Principle #23Feedback

2Ease of operation

If multiple sets of solutions are generated and a learning model is trained, then solutions can be presented according to user preference, but the processing time and computational resources increase

Engineering Contradiction:

ImproveEase of selecting preferred solutionsVSAvoidTime for generating and training solutions

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The system performs preliminary action by generating multiple candidate solutions in advance through the solution generation unit, and pre-training the learning model with user preference data before actual solution presentation. This allows the system to have multiple solutions ready for comparison and can quickly present appropriate solutions when needed, rather than generating solutions on-demand which would be time-consuming.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system applies partial action by generating a limited number of solution sets (e.g., 3-5 sets) rather than exhaustively generating all possible solutions. The learning model is trained on this partial dataset of solution sets, which is sufficient to capture user preferences without the computational burden of processing all possible solution combinations.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250103672A1Learning device, presentation device, learning method, and storage medium

Publication Date: 2025.03.27 NEC CORP

AI summary

The learning device 1X mainly includes a solution generation means 15X and a learning means 16X. The solution generation means 15X generates solutions of an optimization problem. The learning means 16X generates a plurality of sets of solutions and train a model configured to determine a set of solutions to be outputted, based on one or more sets selected by an external input from the plurality of sets.