Balancing Interpretability and Performance in Reinforcement Learning: An Adaptive Spectral Based Linear Approach