Optimization Apparatus Local Minimum Escape via Offset Energy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The simulated annealing method for discrete optimization problems is hindered by long calculation times due to the need to slowly lower temperature to avoid getting trapped in local minima, which is inefficient and difficult to adjust according to the specific problem, and there is a need to promote escape from local minima without dynamically changing the evaluation function or state transition generation.
Innovation Solution
An optimization apparatus with a transition control section that adds an offset to the energy change calculation, allowing state transitions at local minima to be more probable, thereby increasing the absolute values of allowance probabilities while maintaining the relative ratio of state transition probabilities, allowing for quicker escape from local minima without affecting convergence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If simulated annealing method is used to solve discrete optimization problems, then optimal solutions can be obtained, but calculation time becomes excessively long due to slow temperature lowering
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing acceptance probabilities for various energy change values in a lookup table before the optimization process begins. During the actual simulated annealing execution, the algorithm simply retrieves pre-computed probabilities from the table rather than calculating them in real-time, significantly reducing calculation time while preserving solution optimality.
Solution Approach 2:
The patent changes the parameter representation by transforming continuous probability calculations into discrete pre-computed values. By converting the acceptance probability function into a lookup table with discrete energy change values and their corresponding probabilities, the system achieves faster retrieval operations while maintaining the mathematical relationship required for optimal solutions.
2Reliability
If temperature is lowered slowly to avoid local minima, then solution quality improves, but calculation time increases significantly
Solution Approach 1:
The patent prepares acceptance probability tables in advance for various temperature and energy change combinations. This preliminary computation allows the algorithm to quickly determine whether to accept transitions from local minima without performing time-consuming probability calculations during the temperature lowering process, enabling faster escape from local minima.
Solution Approach 2:
The patent replaces the mechanical probability calculation process with a lookup table retrieval system. Instead of computationally evaluating probability functions during temperature changes, the system substitutes this with direct table lookups, dramatically reducing the computational overhead associated with temperature lowering and local minimum escape.
3Loss of time
If approximation algorithms are used instead of exact algorithms, then calculation time decreases, but solution quality becomes approximate rather than optimal
Solution Approach 1:
The patent creates a simplified copy of the probability calculation process by pre-computing acceptance probabilities and storing them in lookup tables. This copy contains the essential information needed for decision-making without requiring the full computational complexity of real-time probability evaluation, thus achieving fast approximate lookups that maintain solution quality.
Data Source
AI summary
When a transition control section stochastically determines on the basis of a temperature, an energy change, and a random number whether to allow one of a plurality of state transitions according to a relative relationship between the energy change and a thermal excitation energy, the transition control section adds an offset to the energy change and exercises control so as to set the offset at a local minimum, at which an energy is locally minimized, to be larger than an offset at the time of the energy not being minimized.