Full-Capacity Unitary Recurrent Neural Networks

NeurIPS 2016 Scott WisdomThomas PowersJohn R. HersheyJonathan Le RouxLes Atlas

Recurrent neural networks are powerful models for processing sequential data, but they are generally plagued by vanishing and exploding gradient problems. Unitary recurrent neural networks (uRNNs), which use unitary recurrence matrices, have recently been proposed as a means to avoid these issues... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Sequential Image Classification Sequential MNIST Full-capacity uRNN Unpermuted Accuracy 96.9% # 6
Sequential Image Classification Sequential MNIST Full-capacity uRNN Permuted Accuracy 94.1% # 5