Abstract
In this paper, we proposed a novel method for Discrete Optimization Problem based on UCB algorithm. General method like GA is easily affected by neighborhood relation structure of the problem, therefore we reconsider the point of the problem from the exploitation and exploration dilemma by dividing the search space into several grids. Selection of the grid area is one of Multi-Armed Bandit Problem, therefore UCB algorithm is directly introduced for the balancing problem. UCB-grid area search was proposed and numerical experiments of 0-1 Knapsack Problem were conducted. Our method showed stable results in different environments.
