Construction of Polar Codes With Reinforcement Learning

This paper formulates the polar-code construction problem for successive-cancellation list (SCL) decoder as a maze-traversing game, which can be solved by reinforcement-learning techniques. The proposed method provides novel technique that no longer depends on sorting and selecting bit-channels reliability, in most current algorithms. Instead, this decides whether input bits should frozen purely sequential manner. equivalence of optimizing SCL under maximizing expected reward traversing maze is drawn. Simulation results show standard constructions are designed optimal with respect to frame error rate (FER). In contrast, game-based finds code have similar or lower FER various lengths sizes decoder, compared state-of-the-art methods. advantage over increases channel signal-to-noise ratio size decoding. Moreover, learning highly efficient terms number required training samples computational operations.