hdetach: Modifying the LSTM Gradient Towards Better Optimization
Abstract
Recurrent neural networks are known for their notorious exploding and vanishing gradient problem (EVGP). This problem becomes more evident in tasks where the information needed to correctly solve them exist over long time scales, because EVGP prevents important gradient components from being backpropagated adequately over a large number of steps. We introduce a simple stochastic algorithm (\textit{h}detach) that is specific to LSTM optimization and targeted towards addressing this problem. Specifically, we show that when the LSTM weights are large, the gradient components through the linear path (cell state) in the LSTM computational graph get suppressed. Based on the hypothesis that these components carry information about long term dependencies (which we show empirically), their suppression can prevent LSTMs from capturing them. Our algorithm\footnote{Our code is available at https://github.com/bhargav104/hdetach.} prevents gradients flowing through this path from getting suppressed, thus allowing the LSTM to capture such dependencies better. We show significant improvements over vanilla LSTM gradient based training in terms of convergence speed, robustness to seed and learning rate, and generalization using our modification of LSTM gradient on various benchmark datasets.
 Publication:

arXiv eprints
 Pub Date:
 October 2018
 arXiv:
 arXiv:1810.03023
 Bibcode:
 2018arXiv181003023A
 Keywords:

 Statistics  Machine Learning;
 Computer Science  Machine Learning
 EPrint:
 First two authors contributed equally. Published in ICLR 2019