Gated recurrent unit
Unidades recurrentes cerradas
- Date
- 2014 · confidence high
- Place
- Montreal, Quebec, Canada
- Credited to
- Kyunghyun Cho
- Field
- Knowledge & symbols
What it is
A recurrent neural network architecture with gating mechanisms that regulate the flow of information without a separate memory cell, introduced by Kyunghyun Cho and collaborators as a simpler alternative to LSTMs.
What it led to
- widespread adoption in machine translation and sequence modeling
- reduction in parameter count compared with LSTM networks
- influence on later deep-learning architectures for text and speech