Accelerated Reinforcement Learning Algorithms with Nonparametric Function Approximation for Opportunistic Spectrum Access