Speaker: Juan Andrés Morales Cordovilla (post-doctoral fellow)
Date: March 3, 2016
Abstract:
Sum Product Networks (SPN) are a new kind of probabilistic models that have the advantages of Deep learning of Neural Networks (DNNs) and of exact marginalization of Gaussian Mixture Models (GMMs). These two properties are very useful to do Missing Data or Uncertainty Decoding on the acoustic models of a Automatic Speech Recognition (ASR) system. In this seminar we explain these advantages, how to train SPNs, previous SPN applications (mainly on vision problems) and our initial results on noisy ASR.
