A delayed single-layer RNN with specific sparse weights reproduces the behavior of a stacked RNN, and with a large delay can match bidirectional RNNs on acausal tasks.
Scaling learning algorithms towards ai
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Approximating Stacked and Bidirectional Recurrent Architectures with the Delayed Recurrent Neural Network
A delayed single-layer RNN with specific sparse weights reproduces the behavior of a stacked RNN, and with a large delay can match bidirectional RNNs on acausal tasks.