Tokyo scientists enhance Ising machines for efficiency

Researchers at Tokyo University of Science have developed a method to enhance the scalability and efficiency of Ising machines, which can provide faster solutions to complex optimisation problems.

Ising machines are specialised computing systems that solve optimisation problems by minimising system energy through "spins" interactions. However, their fully connected architecture poses a scalability challenge due to a large circuit footprint, which the recent advancements aim to address.

Professor Takayuki Kawahara and his team at the Tokyo University of Science have proposed a matrix-folding method to halve the number of physical interactions required between spins. Their work has been published in the journal IEEE Access.

The team's method involves visualising spin interactions as a two-dimensional matrix where each element represents a specific interaction. Since these interactions are symmetric, half of the matrix is redundant and can thus be omitted. In 2020, Kawahara's team introduced a way to reorganise the matrix to limit the circuit size. However, that approach increased the complexity of wiring, which made scaling challenging.

In their latest study, the researchers divided the matrix into four sections, halving each separately and rearranging them into a rectangular shape. This novel arrangement preserves the regularity of the setup, leading to better scalability.

The team implemented a fully coupled Ising machine using their approach on custom circuits comprising 16 field-programmable gate arrays (FPGAs). "Using the proposed approach, we were able to implement 384 spins on only eight FPGA chips. In other words, two independent and fully connected Ising machines could be implemented on the same board," explained Kawahara. "Using these machines, two classic combinatorial optimisation problems were solved simultaneously—namely, the max-cut problem and four-color problem."

The results demonstrated significant performance gains. "We found that the performance ratio of two independent 384-spin fully coupled Ising machines was about 400 times better than simulating one Ising machine on a regular Core i7-4790 CPU to solve the two problems sequentially," reported Kawahara.

These advances have implications for various fields, including faster molecular simulations to expedite drug and materials discovery, efficient data centres, and improving the electrical power grid. Such outcomes align with global sustainability aims, such as reducing the carbon footprint in emerging technology domains like electric vehicles and telecommunications.

Tokyo University of Science (TUS) is a well-known university and the largest science-specialised private research university in Japan, with four campuses in central Tokyo and its suburbs and in Hokkaido. Established in 1881, the university has continually contributed to Japan's development in science through inculcating the love for science in researchers, technicians, and educators.