How is it different from a standard computer?
The CMOS annealing machine is, in a word, a non-von Neumann computer which can process combinatorial optimization problems quickly.
To solve a problem with a conventional von Neumann computer, a solution corresponding to the problem (algorithm) is considered, the algorithm is described as an instruction sequence for the CPU, and the CPU sequentially executes the instructions. Although techniques for simultaneously executing multiple instructions and accelerating the process are also used, it is a feature of conventional computers to execute sequential instructions.
Prior to the widely used von Neumann computer (which is what the current digital circuit technology is based on), there was an era when solving a differential equation with an electronic circuit mainly composed of an electronic part called an operational amplifier. It is a device called an analog computer, the origins of the name are not only because it uses analog electronic circuitry, but also because it is based on the analogy between the two. This is based on the fact that the relationship between the problem of differential equations and the behavior of electronic circuits by operational amplifiers is similar.
Using this knowledge, we pondered if computers that efficiently handle combinatorial optimization problems could be made based on the idea of analog computers.
We describe the combinatorial optimization problem in the model of statistical physics called the Ising model and perform an operation called annealing to find the optimum condition. Instead of solving these actions based on programs, we are trying to solve these behaviors with the hardware of a structure imitating the Ising model.
The von Neumann architecture
The architecture we propose