Exploiting Nontrivial Connectivity for Automatic Speech Recognition
Sound
2017-11-29 v1 Audio and Speech Processing
Machine Learning
Abstract
Nontrivial connectivity has allowed the training of very deep networks by addressing the problem of vanishing gradients and offering a more efficient method of reusing parameters. In this paper we make a comparison between residual networks, densely-connected networks and highway networks on an image classification task. Next, we show that these methodologies can easily be deployed into automatic speech recognition and provide significant improvements to existing models.
Cite
@article{arxiv.1711.10271,
title = {Exploiting Nontrivial Connectivity for Automatic Speech Recognition},
author = {Marius Paraschiv and Lasse Borgholt and Tycho Max Sylvester Tax and Marco Singh and Lars Maaløe},
journal= {arXiv preprint arXiv:1711.10271},
year = {2017}
}
Comments
Accepted at the ML4Audio workshop at the NIPS 2017