• Corpus ID: 204767309

2019 Twelfth International Conference on Contemporary Computing ( IC 3 ) August 8-10 , 2019 Abstract of the Keynotes Stable and Efficient Recurrent Neural Networks

@inproceedings{Dhillon20192019TI,
  title={2019 Twelfth International Conference on Contemporary Computing ( IC 3 ) August 8-10 , 2019 Abstract of the Keynotes Stable and Efficient Recurrent Neural Networks},
  author={Inderjit S. Dhillon},
  year={2019}
}
of the Keynotes Stable and Efficient Recurrent Neural Networks Prof. Inderjit S. Dhillon University of Texas at Austin, USA Despite having remarkable performance on many sequence learning tasks, recurrent neural networks (RNNs) are hard to train with long sequences due to limited expressive power and the vanishing and exploding gradient issues. Previous work has focused on stabilizing the gradients by encouraging orthogonality of weight matrices via re-parameterization techniques. However, two…