Policy-Guided MCTS for near Maximum-Likelihood Decoding of Short Codes
Published in IEEE International Conference on Communications (ICC), 2025
This paper introduces a policy-guided Monte Carlo Tree Search (MCTS) decoder that achieves near maximum-likelihood decoding (MLD) performance for short block codes. The method uses a neural network policy trained via MCTS-based learning to guide the search process, eliminating the need for Gaussian elimination while significantly reducing computational complexity compared to existing near-MLD decoders such as ordered statistics decoding (OSD). Simulation results demonstrate that the proposed decoder achieves near-MLD performance across various short block codes with substantially lower decoding complexity.