Combinatorial Optimization Search With Dynamic Parallel Bit Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing combinatorial optimization methods face inefficiencies in solving large-scale problems due to insufficient solving performance, particularly when the number of state variables for parallel trials is reduced, leading to difficulties in selecting appropriate state variables for updates and reduced state transitions.
Innovation Solution
A data processing device and method that dynamically adjusts the parallel trial bit number based on search information to optimize the number of state variables updated in each trial, improving solving performance by promoting appropriate state transitions and reducing unnecessary calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If the parallel trial bit number is reduced, then the arithmetic operation amount is reduced, but the solving performance deteriorates due to insufficient state variable updates
Solution Approach 1:
The patent dynamically adjusts the parallel trial bit number during the optimization process based on the current energy landscape and search progress. Instead of using a fixed parallel trial bit number, the system adapts this parameter in real-time to balance computational efficiency with solution quality, allowing more aggressive parallel updates when beneficial and more conservative updates when appropriate.
Solution Approach 2:
The system changes the parallel trial bit number as a controllable parameter during the simulated annealing process. By modifying this parameter based on search statistics and energy changes, the patent optimizes the trade-off between performing sufficient state variable updates (for solution quality) and limiting arithmetic operations (for efficiency).
2Productivity
If the parallel trial bit number is increased, then the solving performance is improved through more state variable updates, but the arithmetic operation amount increases
Solution Approach 1:
The patent employs dynamic adjustment of the parallel trial bit number, increasing it when the search benefits from more diverse state explorations and decreasing it when computational efficiency is prioritized. This dynamic control allows the system to respond to the actual needs of the optimization process rather than committing to a fixed parameter setting.
Solution Approach 2:
The system uses feedback from the optimization process (such as energy changes, acceptance rates, and search progress) to adjust the parallel trial bit number. By monitoring these metrics, the patent determines when to increase parallel updates to improve solution quality and when to reduce them to save computational resources.
3Device complexity
If a fixed parallel trial bit number is used, then the device complexity is reduced, but the adaptability to different problem scales and stages deteriorates
Solution Approach 1:
The patent implements a dynamic control mechanism that adjusts the parallel trial bit number based on problem characteristics and search stage. This allows the same device to adapt to different problem scales and optimization phases without requiring multiple specialized configurations, achieving versatility through a single adaptable system.
Solution Approach 2:
The system changes the parallel trial bit number parameter according to problem scale and search progress. By making this parameter adjustable and problem-dependent, the patent enables the device to handle various problem sizes and types effectively, improving adaptability while maintaining reasonable control complexity through systematic adjustment rules.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A data processing program including instructions which, when executed by a computer of searching for a solution for a combinatorial optimization problem represented by an energy function including state variables, cause the computer to execute processing including: executing search processing of searching for the solution by performing determination whether to accept a change of each value of a plurality of first state variables, for the plurality of first state variables selected from among the state variables in parallel and executing processing of changing the value of one state variable of which the change of the value is determined to be accepted while changing the plurality of selected first state variables; and specifying the number of the plurality of selected first state variables, based on a search status of the search processing or search information indicating a search record of another combinatorial optimization problem and repeating the search processing.