Learning Factored Representations in a Deep Mixture of Experts

  Abstract