Curiosity, unobserved rewards and function Approximation in RL - Csaba Szepesvari