Ising Model Sampling Efficiency via Coefficient Averaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing technique for achieving optimization by sampling using an Ising model in a QUBO format often results in a limited sampling region, leading to an increased number of sampling times required to obtain a good solution with high accuracy.
Innovation Solution
An arithmetic program that generates multiple training data sets, creates Ising models for each set, combines these models based on the average and deviation of their coefficients, searches for recommended points in the combined model, and updates the training data with these points and their evaluation values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single Ising model is used for sequential sampling, then the sampling process is simple, but the sampling region is limited and the number of sampling times increases
Solution Approach 1:
The patent combines multiple Ising models into a single integrated model. Specifically, it creates multiple Ising models from different training data sets, then merges them by averaging their coefficients to form a combined Ising model. This combined model expands the sampling region and reduces the number of sampling iterations needed, thereby improving sampling efficiency while reducing time loss.
Solution Approach 2:
The patent performs preliminary actions by pre-processing training data into multiple training data sets and creating corresponding Ising models before the actual sampling process. This preparation phase includes dividing training data, creating multiple models, and computing their average coefficients in advance, which enables more efficient sampling execution and reduces the number of iterations required.
2Productivity
If multiple Ising models are created and combined, then the sampling region expands and sampling efficiency improves, but the processing complexity increases
Solution Approach 1:
The patent segments the training data into multiple training data sets, which are then used to create separate Ising models. This segmentation allows the system to process data in manageable chunks and combine the results systematically. The segmentation approach balances the increased processing complexity with improved sampling efficiency by organizing the computational work into distinct, manageable segments.
Solution Approach 2:
The patent changes parameters by computing average coefficients from multiple Ising models to create a combined model. This parameter transformation approach systematically integrates multiple models by averaging their coefficients, providing a structured method to handle the complexity of multiple models while achieving expanded sampling coverage and improved efficiency.
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
An arithmetic program comprising instructions which, when executed by a computer, cause the computer to repeatedly execute processing including: generating a plurality of training data sets from a training data group; creating a plurality of Ising models by creating the Ising models for each of the plurality of training data sets; creating a combined Ising model by combining the plurality of Ising models according to an average and a deviation of coefficients of the plurality of Ising models; searching for a recommended point for the combined Ising model; and adding the recommended point and an evaluation value of the recommended point to the training data group.