Multi-point search algorithm using reinforcement learning

This paper presents a new meta-heuristic algorithm using multi-point searches and reinforcement learning. The meta-heuristic algorithm has a layered structure that consists of a global search operation and a local search operation for every single point. The results of the local searches are used as the initial conditions of the global search. We demonstrate that the meta-heuristic algorithm with reinforcement learning can efficiently optimize various functions by applying this algorithm to some famous benchmark functions.