Data Search by a Laser Ising Machine with Gradual Pumping or Coupling

We study two operational schemes for a coherent Ising machine based on an injection-locked laser network. These schemes gradually increase the pumping rate or the mutual coupling among the slave lasers. We numerically simulate the two schemes against a data search problem implemented with the Ising model in cubic graphs without frustration. We show the machine can achieve a better success probability and effective computational time to find a target/ground state with these gradual schemes than those with the abrupt introduction of the mutual injection which has been studied previously. The computational time simulated with typical parameters is almost constant up to the problem size M = 200 and turns into a nearly linear scale holding up to M = 1000.