Time adaptive recurrent neural network

Date
2021
Authors
Kag, Anil
Saligrama, Venkatesh
Version
Accepted manuscript
OA Version
Citation
A. Kag, V. Saligrama. 2021. "Time Adaptive Recurrent Neural Network." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, https://doi.org/10.1109/CVPR46437.2021.01490
Abstract
We propose a learning method that, dynamically modi- fies the time-constants of the continuous-time counterpart of a vanilla RNN. The time-constants are modified based on the current observation and hidden state. Our proposal overcomes the issues of RNN trainability, by mitigating ex- ploding and vanishing gradient phenomena based on placing novel constraints on the parameter space, and by suppress- ing noise in inputs based on pondering over informative inputs to strengthen their contribution in the hidden state. As a result, our method is computationally efficient overcoming overheads of many existing methods that also attempt to improve RNN training. Our RNNs, despite being simpler and having light memory footprint, shows competitive per- formance against standard LSTMs and baseline RNN models on many benchmark datasets including those that require long-term memory.
Description
License