A general method for deriving backpropagation algorithms for networks with recurrent and higher order networks is introduced. The propagation of activation in these networks is ...
In the work described here, the backpropagation neural network learning procedure is applied to the analysis and recognition of speech. This procedure takes a set of input/outpu...
Consensus algorithms for networked dynamic systems provide scalable algorithms for sensor fusion in sensor networks. This paper introduces a distributed filter that allows the n...
We present SpecAugment, a simple data augmentation method for speech\nrecognition. SpecAugment is applied directly to the feature inputs of a neural\nnetwork (i.e., filter bank ...
In this paper, we extend the work done on integrating multilayer perceptron (MLP) networks with HMM systems via the Tandem approach. In particular, we explore whether the use of...
Social media, news, blog, policy document mentions