Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: MinimalRNN: Toward More Interpretable and Trainable Recurrent Neural Networks
(Submitted on 18 Nov 2017 (v1), last revised 20 Jun 2018 (this version, v2))
Abstract: We introduce MinimalRNN, a new recurrent neural network architecture that achieves comparable performance as the popular gated RNNs with a simplified structure. It employs minimal updates within RNN, which not only leads to efficient learning and testing but more importantly better interpretability and trainability. We demonstrate that by endorsing the more restrictive update rule, MinimalRNN learns disentangled RNN states. We further examine the learning dynamics of different RNN structures using input-output Jacobians, and show that MinimalRNN is able to capture longer range dependencies than existing RNN architectures.
Submission history
From: Minmin Chen [view email][v1] Sat, 18 Nov 2017 01:42:04 GMT (2772kb,D)
[v2] Wed, 20 Jun 2018 02:19:13 GMT (2772kb,D)
Link back to: arXiv, form interface, contact.