The coherent Ising machine with quantum feedback: the total and conditional master equation methods

TL;DR

This paper derives and compares two quantum master equations for the coherent Ising machine, demonstrating that unconditional simulations are more efficient and scalable for solving large combinatorial problems.

Contribution

It introduces a scalable unconditional simulation method for the quantum master equation of the coherent Ising machine, improving efficiency over conditional approaches.

Findings

Unconditional simulations are more efficient and scalable.

Both master equations agree well but differ in computational efficiency.

Full quantum simulations up to 1000 nodes are feasible.

Abstract

We give a detailed theoretical derivation of the quantum master equation for the coherent Ising machine. This is a quantum computational network with feedback, that solves NP hard combinatoric problems, including the traveling salesman problem and various extensions and analogs. There are two types of master equation that are obtained, either conditional on the feedback current or unconditional. We show that both types can be accurately simulated in a scalable way using stochastic equations in the positive-P phase-space representation. This depends on the nonlinearity present, and we use parameter values that are typical of current experiments. However, while the two approaches are in excellent agreement, they are not equivalent with regard to efficiency. We find that unconditional simulation has much greater efficiency, and is scalable to larger sizes. This is a case where too much knowledge is a dangerous thing. Conditioning the simulations on the feedback current is not essential to determining the success probability, but it greatly increases the computational complexity. To illustrate the speed improvements obtained with the unconditional approach, we carry out full quantum simulations of the quantum master equation with up to 1000 nodes.