Ising Optimization Circuit for 1-Hot Constraint State Transitions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optimization devices face challenges in reducing calculation time for optimization problems with 1-hot constraints, as they often transition to states not satisfying the constraint, leading to a larger search space and increased computation time due to energy barriers.
Innovation Solution
The optimization device includes calculation circuits to determine energy changes for bit value updates, a selection circuit to identify permitted bit updates based on thermal excitation energy and random numbers, and an updating circuit to update bit values, allowing for simultaneous changes of two bits to satisfy the 1-hot constraint, thereby reducing the search space and computation time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the optimization device changes only one bit at a time to perform state transition, then the energy change can be calculated simply, but the search space becomes larger and calculation time increases due to transitions to states not satisfying the 1-hot constraint
Solution Approach 1:
The patent segments the bit change process into two coordinated changes: selecting a bit to change from 1 to 0 and simultaneously selecting a bit to change from 0 to 1 within the same group. This segmentation ensures the 1-hot constraint is maintained while reducing the search space. The calculation circuits compute energy changes for both bits simultaneously, avoiding the need to handle invalid intermediate states.
Solution Approach 2:
The patent performs preliminary selection of bit pairs that satisfy the 1-hot constraint before executing the state transition. The selection circuit identifies valid bit combinations in advance, ensuring that only transitions maintaining the constraint are performed. This preliminary action prevents entering invalid states and reduces the effective search space.
2Ease of operation
If the optimization device performs state transition with Hamming distance=1, then the transition rule is simple, but the number of states to search increases and energy barriers are generated for 1-hot constraint terms
Solution Approach 1:
The patent dynamically adjusts the transition rule based on the current state and constraint requirements. Instead of a fixed Hamming distance=1 rule, the system adaptively selects bit pairs where one bit changes from 1 to 0 and another from 0 to 1, maintaining the 1-hot constraint. This dynamic transition rule eliminates energy barriers associated with constraint violations while preserving operational simplicity through the selection circuit.
3Adaptability or versatility
If the optimization device searches in a larger search space including states not satisfying the 1-hot constraint, then more states are available for transition, but the calculation time increases due to energy barriers
Solution Approach 1:
The patent extracts and eliminates states that do not satisfy the 1-hot constraint from the search space. By using the selection circuit to identify only valid bit transitions, the system removes invalid states from consideration. This extraction reduces the search space to only constraint-satisfying states, eliminating energy barriers while maintaining sufficient transition flexibility through the bit pair selection mechanism.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces calculation time by focusing on transitions that satisfy the 1-hot constraint, decreasing the search space and lowering the energy barrier, resulting in faster convergence to the ground state.
Implementation Method 1
a magnitude relationship between thermal excitation energy, determined based on an input temperature parameter and a random number, and the first energy change output by each of the plurality of calculation circuits
Data Source
AI summary
An optimization device includes a plurality of calculation circuits; a selection circuit; an identification information calculation circuit, and an updating circuit. Each of the plurality of calculation circuits calculates, for a plurality of bits corresponding to a plurality of spins included in an Ising model obtained by converting a problem to be calculated, a first energy change of the Ising model due to a value of a first bit having the value of 1 being changed from 1 to 0 and a value of a second bit having the value of 0 being changed from 0 to 1. The selection circuit outputs first bit identification information identifying one second bit having a value permitted to be updated from 0 to 1, based on a magnitude relationship between thermal excitation energy and the first energy change output by each of the plurality of calculation circuits.