Given data from some experiment, inferring information from the underlying distribution is of prime importance, and has been extensively studied. However, due to the huge size of the data, traditional methods are often ...