Atlas · GenAI 2026
Gated Recurrent Unit
Designing and training recurrent sequence models with reset and update gates that control how prior hidden state is used and retained.
Also searchable as: GRU, GRUs, Gated Recurrent Units
conceptNeural ArchitecturesAI consensus: 0/3
Prerequisites
Hidden state and recurrent sequence processing are the conceptual foundation.
Recommended reference
Cho et al.: Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation — https://arxiv.org/html/1406.1078v3; Keras: GRU layer — https://keras.io/api/layers/recurrent_layers/gru/
Reviewed sources
Primary and first-party material reviewed for this editorial summary. These citations are separate from the AI consensus score above.
- Cho et al.: Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
Section 2.3 introduces the gated unit with reset and update mechanisms.
- Keras: GRU layer
A concrete GRU layer and implementation variants.
- PyTorch 2.9: GRU
Official GRU API with reset and update equations, hidden-state shapes, stacked and bidirectional options, and a documented implementation difference from the original paper.
Notes from AI deep research
Related skills
- → is subcategory of: Recurrent Neural Networks