Energy-efficient superparamagnetic Ising machine and its application to traveling salesman problems

Area of Science:

• Materials Science
• Computer Engineering
• Artificial Intelligence

Background:

• Artificial intelligence (AI) growth intensifies computational demands for solving complex non-deterministic polynomial-time (NP)-hard problems.
• Existing Ising computers face challenges including high power consumption and limited scalability.

Purpose of the Study:

• To experimentally demonstrate an Ising annealing computer utilizing superparamagnetic tunnel junctions (SMTJs).
• To address the computational burden and limitations of current NP-hard problem-solving methods.

Main Methods:

• Construction of an Ising annealing computer with 80 superparamagnetic tunnel junctions (SMTJs) and all-to-all connections.
• Application of the system to solve a 70-city traveling salesman problem (TSP), equivalent to a 4761-node Ising problem.
• Leveraging intrinsic SMTJ randomness, a global annealing scheme, and an efficient algorithm.

Main Results:

• The SMTJ-based Ising annealer demonstrated superior performance over other Ising schemes in power consumption and energy efficiency.
• Successfully solved a complex 70-city traveling salesman problem.
• Proposed a scalable cross-bar array architecture for integration with magnetic random-access memories.

Conclusions:

• The SMTJ-based Ising computer presents a highly energy-efficient and scalable solution for complex computational problems.
• This approach offers a promising pathway for solving challenging problems with constrained hardware resources.
• SMTJ-based Ising computers are strong candidates for future unconventional computing paradigms.