Interpretable RL Model Matches Neural Networks on Atari Breakout

Researchers have introduced an interpretable experiential learning model that uses state history and global feedback to learn behavioral models represented as transition graphs between state sets, with attributed utility and evidence counts. Designed for resource-constrained environments, the model was evaluated on the OpenAI Gym Atari Breakout benchmark and achieved performance comparable to known neural network-based solutions. The work is published on arXiv under computer science and machine learning categories.