A Computationally Efficient Method for Learning Exponential Family Distributions

Neural Information Processing Systems 

We consider the question of learning the natural parameters of a $k$ parameter \textit{minimal} exponential family from i.i.d.