Ising Machine: Largest Light-based Computer to Resolve Complex Problems
Finding the shortest route amongst a number of cities, comparing the many different roads, is a difficult task which increases in complexity based on the number of cities to visit. Combinatorial optimization problems such as these are very frequent in everyday life, in science and in engineering, but they are hard to tackle on large scale with traditional computers.
The development of new hardware system that can efficiently solve complex optimisations is one of the challenges of modern science. A procedure that looks promising is to codify these problems into Ising Models, mathematical-physical models that are defined by a finite set of variables (spin), which can be solved by specific optical processors that are referred to as Ising Machines. These computers codify the state of the variables and their connections in the bands and phases of the electromagnetic field. Calculating this data at the speed of light through a series of spatial and frequency channels will most likely be far more rapid that using electronic devices. This is the main task for which quantum computers are being developed at the NASA or Google.
The research team from the Sapienza Department of Physics and the Complex Systems Institute at the National Research Council (CNR), has designed and experimentally implemented the largest Ising Machine ever created. The results of the project, which have been published on the Physical Review Letters Journal, sheds fundamental new light on the development of future technology.
The researchers implemented the first device with a potentially unlimited number of spins at the Sapienza Photonics Laboratory. They were able to code and process tens of thousands of variables by using optical beams that were spatially modulated by active displays. The new optical prototype is based on high-resolution display technology and could allow scientists to face optimization problems that are currently considered computationally impossible, such as the determination of protein sequences from amino-acid sequences, the generation of bit-coins or that of secure passwords.
“We started with an original idea,” explains Claudio Conti, “that employs individual pixels in modern displays both as computational units and as memory units without needing to assemble any further optical devices. More specifically, we used a spatial light modulator to act on the phase and bandwidth of the optical field, codifying all the variables necessary for the calculations and their interactions. Electromagnetic fields modulated in this way are developed to simulate the problem at hand. And measuring the final profile will allow us to identify the optimal solution.”
The study, which demonstrates a large-scale computational procedure, nears Ising Machines to the solution of optimisation problems that are fundamental for the economic and social development and its sustainability. Its applications range from the development of ultra-rapid optical processers to photonic devices for artificial intelligence and automatic learning.
Large-Scale Photonic Ising Machine by Spatial Light Modulation - Pierangeli, D., Marcucci, G., & Conti, C. - Physical Review Letters 2019, 122, 213902.
Department of Physics, Sapienza University of Rome