Ising Model Recommendation Sampling with FMDA and GA

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing binary variable sampling technologies, such as those using QUBO-format Ising models, face challenges in efficiently adjusting the number of recommendations, leading to potential oversampling and reduced accuracy due to limitations in sampling area and dependence on training data quality.

Innovation Solution

A hybrid approach combining FMDA (Factorization Machine with Digital Annealer) and GA (Genetic Algorithm) to dynamically adjust the number of recommendations based on the accuracy of the FM model, using FMDA for high-accuracy areas and GA for broader sampling when FM model accuracy is low, thereby optimizing the search for optimal solutions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If FMDA is used for sampling, then manufacturing precision of recommendation points is improved, but productivity decreases due to limited sampling area and difficulty in adjusting recommendation numbers

Engineering Contradiction:

Improveaccuracy of recommendation pointsVSAvoidefficiency of sampling process

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent combines FMDA and GA into a hybrid recommendation system. FMDA is used to generate initial recommendation points with high manufacturing precision, while GA is used to expand the sampling area and adjust the number of recommendations. This merging allows the system to maintain high accuracy while improving productivity and flexibility.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent implements dynamic adjustment of the number of recommendations by using GA to generate additional recommendation points based on the initial FMDA results. The system can adaptively increase or decrease the number of recommendations depending on the complexity of the optimization problem and the desired sampling coverage, thereby improving productivity without sacrificing precision.

Inventive Principle:

Principle #15Dynamics

2Productivity

If the number of recommendations is increased to improve sampling coverage, then productivity is improved, but manufacturing precision decreases due to oversampling and resource dilution

Engineering Contradiction:

Improvesampling coverageVSAvoidaccuracy of recommendation points

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The patent segments the recommendation generation process into two stages: first, FMDA generates a small number of high-precision recommendation points; second, GA generates additional recommendation points to expand sampling coverage. This segmentation ensures that the core high-precision recommendations are not diluted by excessive sampling, while still achieving broad coverage through the GA-generated points.

Inventive Principle:

Principle #1Segmentation

3Productivity

If GA is used for broader sampling, then productivity is improved, but manufacturing precision decreases due to random sampling nature

Engineering Contradiction:

Improvesampling area coverageVSAvoidaccuracy of recommendation points

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The patent applies preliminary action by using FMDA to generate high-precision recommendation points first, which serve as a foundation for the subsequent GA process. The GA algorithm then uses these pre-generated points as starting points for broader sampling, ensuring that even though GA is random in nature, it begins from high-precision locations, thereby maintaining overall manufacturing precision while expanding productivity.

Inventive Principle:

Principle #10Preliminary action

4Manufacturing precision

If FMDA is used for all recommendations, then manufacturing precision is improved, but device complexity increases due to difficulty in adjusting recommendation numbers

Engineering Contradiction:

Improveaccuracy of recommendation pointsVSAvoidcomplexity of controlling recommendation numbers

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent introduces GA as an intermediary component that handles the adjustment of recommendation numbers. Instead of making FMDA complex to accommodate variable recommendation counts, the system uses GA to generate the appropriate number of additional recommendations based on FMDA's high-precision results. This intermediary approach maintains FMDA's simplicity and precision while adding the flexibility to adjust recommendation numbers.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20250322284A1Non-transitory computer-readable recording medium, calculation method and information processing device

Publication Date: 2025.10.16 FUJITSU LTD

AI summary

A calculation program causes a computer to execute a process including determining a first set number and a second set number according to accuracy of an Ising model, in repeating of a process of creating the Ising model based on a learning data group, searching for the first set number of a first recommendation point for the Ising model using an Ising machine, searching for the second set number of a second recommendation point for a learning data by a genetic algorithm, and adding the first recommendation point and a first evaluation value of the first recommendation point, and the second recommendation point and a second evaluation value of the second recommendation point, respectively, to the learning data group as a learning data.