Optimization Device Balancing State Variable Flipping Rates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing optimization methods for combinatorial problems using Ising models often result in biased search efficiency towards partial subspaces of the state space, leading to significant degradation in solution search efficiency due to uneven flipping of state variables.

Innovation Solution

An optimization device and method that includes a state holding unit, a calculation unit to evaluate energy changes based on weight values and susceptibility correction values, and an updating unit to adjust state variables, ensuring balanced flipping rates across all state variables through dynamic adjustment of correction values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If conventional MCMC method is used to solve combinatorial optimization problems, then the search process is simple to implement, but the search efficiency deteriorates due to trapping in local solutions and biased flipping of state variables

Engineering Contradiction:
Improvesimplicity of implementationVSAvoidsolution search efficiency
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The patent introduces correction values that modify the energy change calculation for each state variable. These correction values dynamically adjust the effective energy landscape, preventing the system from getting trapped in local minima and ensuring more uniform flipping rates across all state variables, thereby improving search efficiency without complicating the basic MCMC framework

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent implements a feedback mechanism where correction values are calculated based on the flipping rates of state variables. This feedback loop continuously monitors and adjusts the energy changes to balance the flipping rates, preventing bias towards partial subspaces and maintaining efficient exploration of the state space

Inventive Principle:
Principle #23Feedback

2Productivity

If correction values are introduced to balance flipping rates, then search efficiency improves by preventing bias towards partial subspaces, but calculation complexity increases

Engineering Contradiction:
Improvesolution search efficiencyVSAvoidcalculation complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent modifies the energy change calculation by introducing correction values that are added to the conventional energy change. This parameter change allows the system to balance flipping rates without fundamentally altering the MCMC algorithm structure, adding only a computationally manageable layer of complexity

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The correction values are calculated and prepared in advance before the state variable flipping decision is made. This preliminary calculation of correction values based on current flipping rates allows the main MCMC loop to proceed with minimal additional computational overhead during the critical decision-making phase

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4258134A1Optimization device, optimization method, and optimization program
Publication Date: 2023.10.11 FUJITSU LTD
  • EP4258134A1 patent drawingFigure 1
  • EP4258134A1 patent drawingFigure 2A
  • EP4258134A1 patent drawingFigure 2B

AI summary

An optimization device including: a state holding unit that holds each of values of a plurality of state variables included in an evaluation function; a calculation unit that calculates change values of the evaluation function when any one of the values of the plurality of state variables changes with a probability based on a weight value of each of the plurality of state variables, and calculates evaluation values that evaluate which state transition to accept among the plurality of state variables, based on the calculated change values and correction values that correspond to susceptibility to change in the state variables of which the values have been changed; and an updating unit that changes one of the held values of any one of the state variables among the plurality of state variables, based on the calculated evaluation values.