[10a-Z09-1~9] 23.1 Joint Session N "Informatics"

▲ [10a-Z09-6] Enhanced Monte Carlo Tree Search for Materials Design and Discovery

〇Sae Dieb¹, Masashi Ishii¹ (1.MaDIS, NIMS)

Keywords:Monte Carlo tree search

Materials design and discovery is often formulated as the selection of optimal solution from a space of candidates. Monte Carlo tree search (MCTS) has shown efficiency in solving this inverse design for several applications; however, randomization technique in MCTS can limit efficiency. In this work , we present an enhanced model of MCTS with policy gradient, a reinforcement learning algorithm.