Replica Interaction in Ising Machine Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optimization methods using multiple replicas in Ising machines often fail to sufficiently explore the state space due to replicas getting trapped in local solutions, leading to inefficient search performance.
Innovation Solution
The optimization method introduces interaction between replicas by calculating the strength of interaction based on distance, using the Metropolis-Hastings method to determine proposal and acceptance probabilities, thereby influencing the state transition process and enhancing the exploration of the state space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple replicas are used to perform MCMC processes independently, then the search speed is improved, but replicas get trapped in local solutions reducing search effectiveness
Solution Approach 1:
The patent introduces interaction terms that create feedback between replicas based on their state variable configurations. Each replica's state transition probability is modified by feedback from other replicas' states, allowing the system to collectively escape local solutions while maintaining independent parallel computation. This feedback mechanism ensures that replicas influence each other's search trajectories without requiring complete synchronization.
Solution Approach 2:
The patent modifies the transition probability parameters by introducing interaction terms that depend on the distance between replica states in the state space. The proposal probability is changed from a uniform distribution to one that incorporates interaction strengths, allowing replicas to explore the state space more effectively. This parameter change enables the system to maintain high productivity while improving solution accuracy through enhanced exploration.
2Reliability
If interaction between replicas is introduced to prevent trapping in local solutions, then solution search capability is improved, but computational complexity increases
Solution Approach 1:
The patent applies local quality by introducing interaction terms that are non-zero only for replicas within a certain distance threshold in the state space. Not all replicas interact with all others; instead, each replica only considers interactions with nearby states, reducing the computational burden. This localized interaction approach maintains solution search capability while significantly lowering computational complexity compared to full pairwise interactions.
Solution Approach 2:
The patent implements partial action by considering only a subset of interaction terms rather than all possible replica combinations. The interaction strength is modulated by distance, automatically reducing the effective number of interactions as replica states diverge. This partial action strategy provides sufficient interaction to prevent local solution trapping while avoiding the excessive computational cost of complete interaction consideration.
Data Source
AI summary
An optimization method includes: identifying, for each of a plurality of replicas each of which has a plurality of state variables, an amount of change in strength of interaction that corresponds to a change in a distance between the replica and another replica in a state space in a case where a value of a first state variable among the plurality of state variables of the replica is updated, the state space indicating a space which a combination of values of the plurality of state variables is able to take; and determining whether or not to update the value of the first state variable, based on a proposal probability that corresponds to the amount of change in the strength of interaction in the case where the value of the first state variable is updated and based on an acceptance probability that corresponds to a target probability distribution.