Encoder-Decoder Inference Models for Combinational Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing transportation plan preparation assistance methods, such as those described in Patent Literature 1, are limited by fixed input and output data, which is not adaptable to various combinational optimization problems.

Innovation Solution

An information processing apparatus utilizing reinforcement learning to generate an inference model with an encoder and decoder for extracting features and generating solutions to combinational optimization problems, allowing for adaptable input data and flexible output formats.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If fixed input data and fixed output data are used in optimization calculation, then the calculation process is simple and deterministic, but the adaptability to various combinational optimization problems is poor

Engineering Contradiction:

Improveadaptability to various combinational optimization problemsVSAvoidcomplexity of input data processing

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent changes the parameter format from fixed structured data to variable format data (including images, text, tables, etc.). The encoder processes this variable input data and transforms it into a standardized feature representation, allowing the system to adapt to different optimization problem types without changing the core calculation architecture.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The encoder acts as an intermediary between the variable input data and the optimization calculation. It extracts features from diverse data formats (images, text, tables) and converts them into a unified feature representation that can be processed by the optimization algorithm, thereby decoupling the diversity of input data from the complexity of processing.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If reinforcement learning is used to generate inference models, then the inference accuracy and solution quality improve, but the model training time and computational resources increase

Engineering Contradiction:

Improveinference accuracy of solutionVSAvoidmodel training time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary training of the encoder and decoder models using reinforcement learning before actual optimization calculations. During training, the model learns to extract features and generate solutions for various optimization problems. Once trained, the model can quickly infer solutions for new problems without requiring time-consuming real-time training, thus separating the time investment between training phase and inference phase.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250307342A1Information processing apparatus, inference model generation method, inference method, and storage medium

Publication Date: 2025.10.02 NEC CORP

AI summary

In order to make it possible to achieve improvement in inference of a solution to a combinational optimization problem, an information processing apparatus includes: a data acquisition section that acquires input data pertaining to a combinational optimization problem for which a solution is to be obtained; and an inference section that infers a solution in accordance with the input data using an inference model which has been generated by reinforcement learning for inferring a solution to the optimization problem, the inference model including an encoder that extracts a feature of the input data and a decoder that generates information indicating a solution to the optimization problem using the feature extracted by the encoder.