KAIST researchers have created specialized computer hardware that can tackle difficult optimization tasks much more quickly than traditional methods, using only standard manufacturing processes already common in the chip industry.
Combinatorial optimization problems require finding the best solution from a huge number of possibilities. These tasks appear in logistics, finance, and chip design and can take thousands of years for ordinary computers to complete. The new hardware, called an oscillatory Ising machine, uses many small oscillating parts that work together to reach the best answer naturally. An oscillator produces repeating electrical signals at a steady rhythm.
In this system, multiple oscillators exchange signals and adjust their rhythms until they synchronize. This stable synchronized state corresponds to the optimal solution for the problem. Previous versions of such machines faced difficulties because it was hard to control small differences in their frequencies and to connect the parts effectively for complex tasks.
Breakthrough in silicon-based design
The researchers built both the oscillators and the connecting couplers using ordinary single transistors made with standard CMOS technology used for computer chips in smartphones and processors. This new approach greatly reduced unwanted frequency variations, improved synchronization, and allowed more precise connections that better represent the weights in real-world problems.
As a result, the hardware showed strong performance on the Max-Cut problem, a well-known optimization task that involves dividing a network into two groups to maximize connections between them. This type of problem has direct uses in route planning, investment portfolios, and placing circuits on chips. Because the device uses existing silicon production lines, it can be manufactured at large scale without new factories or special materials.
The work also proposes a broader idea for the future of transistors. After decades focused on making transistors smaller to improve switching and amplification, the researchers suggest a third role for them as oscillators. This shift could open new possibilities as traditional miniaturization reaches physical limits. The hardware operates at room temperature and demonstrates a practical path toward faster, more accurate decision-making tools for industry.
This research is published in Science Advances.