Ising Model Sampling Using Genetic Algorithms and Digital Annealers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sampling techniques using Ising models in the QUBO format often result in a limited sampling region, leading to an increased number of sampling iterations required to achieve a good solution with high accuracy.
Innovation Solution
The proposed solution involves an operation program that performs processing including creating an Ising model, searching for recommended points using both genetic algorithms and digital annealers, and updating the learning data group to reduce the number of sampling iterations. This is achieved by converting matrices into bit arrays and applying constraints such as the 2way1hot constraint to optimize the search process.
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 iterations increases
Solution Approach 1:
The patent combines two different search approaches: (1) converting matrices to bit arrays and applying constraints like 2way1hot constraint for structured search, and (2) using genetic algorithms for exploratory search. This merging allows the system to explore a broader sampling region while maintaining solution accuracy, thereby reducing the number of iterations needed.
Solution Approach 2:
The patent transforms the problem representation by converting matrices into bit arrays and applying dimensional constraints (2way1hot constraint where only one variable per row and column can be 1). This dimensional transformation expands the effective sampling region and enables more efficient navigation of the solution space.
2Measurement precision
If the number of sampling iterations is increased to achieve high accuracy, then better solutions can be found, but the search efficiency decreases
Solution Approach 1:
The patent dynamically adjusts the search strategy by alternating between constraint-based structured search and genetic algorithm-based exploratory search. This dynamic approach allows the system to efficiently navigate the solution space without requiring excessive iterations, thereby maintaining both accuracy and search efficiency.
Solution Approach 2:
The patent introduces intermediate representations (bit arrays with 2way1hot constraints) that serve as mediators between the original matrix problem and the sampling process. These intermediaries structure the search space in a way that improves both solution accuracy and search efficiency.
Data Source
AI summary
A medium storing an operation program for causing a computer to perform processing including: in repeatedly executing 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, generating types of a sequence as an initial point of each piece of learning data of the learning data group; and when the second recommended points are searched for, applying the genetic algorithm to a form of the sequence.