Chowdhury, Oliveira, 2022. Value Function Approximations via Kernel Embeddings for No-Regret Reinforcement Learning.