Quantum-inspired Ising machine using sparsified spin connectivity.

Area of Science:

• Computational physics
• Quantum-inspired algorithms
• Combinatorial optimization

Background:

• NP-hard combinatorial optimization problems are computationally intractable at scale.
• Simulated annealing (SA) is a metaheuristic for ground-state search, mimicking thermal dynamics.
• Extraction-type majority voting logic (E-MVL) is a quantum-inspired algorithm using digital logic circuits.

Purpose of the Study:

• To investigate the performance potential of E-MVL through systematic optimization.
• To benchmark E-MVL comprehensively against SA for the Sherrington-Kirkpatrick (SK) model.
• To explore E-MVL's impact on SA's temperature scheduling.

Main Methods:

• Systematic optimization of E-MVL parameters.
• Benchmarking E-MVL against SA on the SK model with bimodal and Gaussian couplings.
• Equilibrium state analysis to evaluate solution space search consistency.
• FPGA implementation for speed comparison.

Main Results:

• E-MVL demonstrates consistent solution space search across different coupling distributions and sizes.
• E-MVL solves exact solutions up to 1600 spins, significantly outperforming SA (limited to 400 spins).
• E-MVL provides insights that improve SA's temperature scheduling, leading to enhanced performance.
• FPGA implementation of E-MVL shows a 6-fold speedup compared to SA.

Conclusions:

• E-MVL is a highly efficient quantum-inspired optimizer for large-scale combinatorial problems.
• E-MVL offers a practical methodology for enhancing the performance of simulated annealing.
• The algorithm's sparsity control mechanism ensures robust performance across diverse problem instances.