What question did this study set out to answer?

This research aims to establish a benchmark for reinforcement learning techniques in maze navigation tasks.

May 9, 20260 citationsOpen Access

A Reinforcement Learning Calibration Benchmark of Tabular Q-Learning and One-Hot DQN on Enumerable Maze Navigation

Puntos clave

This research aims to establish a benchmark for reinforcement learning techniques in maze navigation tasks.
Evaluated tabular Q-Learning and one-hot DQN in deterministic grid-world environments.
Matched interaction budgets and conducted multi-seed evaluations for reliability.
Performed ablation studies on factors such as depth, replay buffers, target networks, and state representation.
Tabular Q-Learning achieved equivalent final policy quality while demonstrating superior sample efficiency and lower wall-clock cost.
Optimization overhead from deep RL infrastructure outperformed benefits from function approximation in small enumerable MDPs.
Repository provides fully reproducible implementation, statistical evaluation utilities, and benchmark scripts.

Resumen

This paper presents a reinforcement learning calibration benchmark comparing tabular Q-Learning and one-hot Deep Q-Networks (DQN) on enumerable maze navigation tasks. Across deterministic grid-world environments with matched interaction budgets and multi-seed evaluation, tabular Q-Learning achieves equivalent final policy quality with superior sample efficiency and lower wall-clock cost. Additional ablations isolate the contribution of depth, replay buffers, target networks, and state representation. Results suggest that, in small enumerable MDPs, optimization overhead from deep RL infrastructure can dominate any benefit from function approximation. The repository includes a fully reproducible single-file NumPy implementation, statistical evaluation utilities, and benchmark scripts.

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo