By SPIE – International Society for Optics and Photonics Feb. 26, 2024
Collective state computation makes use of the Ising problem and light-based techniques to provide a novel approach to overcome the limitations of traditional computing.
The researchers developed a computer from a VCSEL array with optical feedback.
In our data-driven era, it's critical to solve complex problems efficiently. However, traditional computers often struggle to accomplish this task when dealing with a large number of interactive variables, leading to inefficiencies such as von Neumann bottlenecks. A new type of collective state computation has emerged to solve this problem by mapping these optimization problems to the Ising problem in magnetism.
Here's how it works: Imagine representing a problem as a graph where nodes are connected by edges. Each node has two states, +1 or -1, representing a potential solution. The goal is to find a configuration that minimizes the total energy of the system based on a concept called Hamiltonian.
In the Ising computer (represented here by 4 bits), the variables all move towards parallel solutions. **Author doi: 101117/1.jom.4.1.014501
To effectively solve for the Isinghamitonian, researchers are exploring physical systems that can surpass traditional computers. A promising approach involves the use of light-based techniques, in which information is encoded into properties such as polarization, phase, or amplitude. By taking advantage of effects such as interference and light feedback, these systems can quickly find the right solution.
In a study published in the Journal of Optical Microsystems, researchers from the National University of Singapore and the Agency for Science, Technology and Research investigated the use of a vertical-cavity surface-emitting laser (VCSEEL) system to solve the Ising problem. In this setup, the information is encoded in linear polarization states of VCSELs, with each state corresponding to a potential solution. The lasers are connected to each other, and the interaction between them encodes the structure of the problem.
The researchers tested their system on modest 2-, 3-, and 4-bit ising problems and found promising results. However, they also found challenges, such as the need for minimal VCSEL laser anisotropy, which may be difficult to achieve in practice. Nonetheless, overcoming these challenges could result in a VCSEL-based all-optical computer architecture capable of solving problems that are currently unsolvable by traditional computers.
Reference: "Linear Polarization State Encoding for Ising Computing with Optically Injection-Locked VCSELS" by Brandon Loke, Zifeng Yuan, Soon Thor Lim, Aaron Danner, December 28, 2023, Journal of Optical Microsystems.