Hybrid FMDA GA Sampling for Ising Model Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing sampling techniques using Ising models in the QUBO format often result in a limited sampling region, leading to an increased number of sampling times required to achieve high accuracy in finding optimal solutions, as they may not fully express the problem and struggle with adjusting the number of recommendations effectively.
Innovation Solution
An operation program and information processing apparatus that utilize a combination of FMDA and GA methods, where initial points are generated using D-optimal design to expand the sampling space, and the weight of a linear weighted sum of objective functions is dynamically changed to optimize the sampling process, reducing the number of sampling times by adjusting the number of DA and GA recommendations based on the accuracy of the FM model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sampling techniques using Ising models in QUBO format are used, then optimization can be performed, but the sampling region is limited and the number of sampling times increases
Solution Approach 1:
The patent combines FMDA (Factorization Machine with Digital Annealing) and GA (Genetic Algorithm) into a hybrid recommendation system. The FMDA generates initial recommended points based on learning data, while the GA searches for additional recommended points in a broader space. This merging allows the system to maintain high accuracy through FMDA while expanding the sampling region through GA, thereby reducing the total number of sampling times needed to achieve high accuracy.
Solution Approach 2:
The patent dynamically adjusts the number of recommended points generated by FMDA and GA based on the accuracy of the FM model. When the FM model accuracy is high, fewer FMDA recommendations are needed; when accuracy is low, more recommendations are generated. This dynamic adjustment optimizes the sampling process by adapting to the current state of model accuracy, reducing unnecessary sampling operations while maintaining solution quality.
2Measurement precision
If the number of recommended points is increased to expand sampling region, then solution accuracy improves, but the complexity of adjusting recommendations increases
Solution Approach 1:
The patent implements a feedback mechanism where the accuracy of the FM model is evaluated and used to control the number of recommended points generated by FMDA and GA. The system monitors model accuracy and adjusts the recommendation counts accordingly, creating a closed-loop control system that automatically optimizes the sampling strategy without manual intervention, thereby reducing complexity while maintaining accuracy.
3Loss of time
If FMDA and GA are combined to expand sampling space, then the number of sampling times reduces, but the complexity of the operation processing increases
Solution Approach 1:
The patent segments the recommendation generation process into two distinct phases: FMDA generates initial recommended points based on learning data patterns, and GA performs additional search in the expanded space. This segmentation allows each algorithm to specialize in its strength while working in sequence, making the overall complex process more manageable and efficient compared to using a single algorithm for the entire search.
Data Source
AI summary
A computer-readable recording medium stores an operation program for causing a computer to execute a process of, when operation processing is repeatedly executed, the operation processing including creating an Ising model based on a learning data group, searching for a first set number of first recommended points for the Ising model, searching for a second set number of second recommended points for the learning data group by a genetic algorithm, and adding the first recommended points and first evaluation values of the first recommended points and the second recommended points and second evaluation values of the second recommended points to the learning data group as learning data, expressing each piece of learning data of the learning data group as an objective variable obtained by a linear weighted sum of a plurality of objective functions, and changing a weight of the linear weighted sum every time the operation processing is executed.