Learning Symbolic Models of Stochastic Domains