GRU
A gated recurrent unit (GRU) is a recurrent neural-network architecture that uses learned gates to regulate information flow and update a hidden state across sequence steps.
Inside Deep Learning: Math, Algorithms, Models
A gated recurrent neural network architecture that updates a hidden state over sequence steps with reset and update gates.
- How it is used
- Uses recurrent hidden-state updates for autoregressive text models and sequence-to-sequence encoders or decoders.
- How deep it goes
- Explains implementation choices, hidden-state handling, and layer iteration, but defers the GRU's internal gate equations.
- Examples
- Demonstrates character-level next-token prediction, bidirectional sequence encoding, variable-length decoding, and multilayer GRUCell construction.
- What you can do after
- Build it from scratch with PyTorch nn.GRU or nn.GRUCell and integrate it into autoregressive or sequence-to-sequence models.
Read how Inside Deep Learning: Math, Algorithms, Models explains GRU →
Graph Neural Networks in Action
A recurrent neural-network architecture that uses reset and update gates to control information flow through its recurrent state.
- How it is used
- Uses GRUs to carry temporal information in sensor-trajectory prediction and a graph-aware pose decoder.
- How deep it goes
- Explains gate roles and implements both a PyTorch GRU model and a custom decoder cell, without deriving GRU internals mathematically.
- Examples
- Shows a pose-estimation model using `nn.GRU` and a custom cell that incorporates edge-type-weighted messages.
- What you can do after
- Build it from scratch as a custom node-and-edge decoder cell, or apply it via PyTorch `nn.GRU` for sequence prediction.
Deep Learning with Python, Third Edition
A recurrent neural network unit designed to process sequences with gated state updates.
- How it is used
- Uses GRU for sequence forecasting and next-character prediction, with a further encoder–decoder role in translation.
- How deep it goes
- Works through Keras stacking and explicit-state generation, but does not derive GRU gating equations.
- Examples
- Trains a stacked temperature forecaster and a character-level Shakespeare language model.
- What you can do after
- Build it from scratch in Keras for stacked forecasting, then adapt trained weights for token-by-token generation.
Read how Deep Learning with Python, Third Edition explains GRU →
Deep Learning with R, Third Edition
A gated recurrent unit (GRU) is a recurrent neural-network architecture that models sequence dependencies with update and reset mechanisms.
- How it is used
- Models ordered timeseries and text sequences for temperature forecasting and next-character prediction.
- How deep it goes
- Shows working single-layer and stacked architectures, including sequence outputs and dropout, but does not derive GRU internals.
- Examples
- Uses a stacked temperature-forecasting model and a single-layer Shakespeare character language model.
- What you can do after
- Apply it via Keras layer_gru to build single-layer or stacked sequence models and evaluate them against a baseline.
Natural Language Processing in Action, Second Edition
A gated recurrent unit (GRU) is a recurrent neural-network architecture that uses reset and update gates to control information flow and hidden-state updates.
- How it is used
- Uses gated recurrence to preserve useful information when ordinary RNNs lose influence from early tokens.
- How deep it goes
- Explains gate intuition, learned thresholds, recurrent structure and reset/update calculations down to implementation expressions.
- Examples
- Uses PyTorch WikiText2 language modeling, masked `<unk>` tokens, parameter loading and prompted word generation.
- What you can do after
- Apply it via PyTorch by selecting a GRU language model, loading parameters and generating text from a prompt.
Read how Natural Language Processing in Action, Second Edition explains GRU →
Sutskever's List
A gated recurrent unit is a recurrent neural-network unit that uses learned gates to regulate state updates.
- How it is used
- Deploys GRUs as encoder and decoder units for attention-based neural machine translation.
- How deep it goes
- Explains bidirectional encoding and decoder inputs but stops short of detailing individual gate equations.
- Examples
- Uses bidirectional sentence encoding, a 1,000-unit decoder, and WMT English–French translation results.
- What you can do after
- Evaluate how GRU-based encoder-decoder systems support translation with attention and handle longer sentences.
Grokking Machine Learning
A gated recurrent unit is a recurrent neural-network unit that uses gates to control the retention and update of sequence state.
- How it is used
- Applies gated recurrent architectures to language-processing tasks involving written or spoken sequences.
- How deep it goes
- Defines GRUs as advanced recurrent architectures but stops short of explaining their gate equations or internal update procedure.
- Examples
- Uses sentiment analysis, machine translation, speech recognition, text summarization, and topic-specific chatbots as application examples.
Related concepts
- gated updatehas part · A gated update is a state or representation update in which learned gates control how new information is combined with or replaces existing information.
- reset gatehas part · A reset gate is a GRU mechanism that controls how much of the previous hidden state contributes to computing a candidate state.
- vanishing gradient problemaddresses · The vanishing-gradient problem is a neural-network training difficulty in which gradients shrink toward zero as they propagate backward, making updates to earlier layers or distant sequence steps ineffective.
- LSTMalternative to · An LSTM is a recurrent neural network architecture with gated memory mechanisms that control information flow and help retain or update information across sequences.
- recurrent neural networkaddresses · A recurrent neural network is a neural network that processes sequences by updating an internal state across successive inputs, allowing earlier inputs to influence later computations.
- bidirectional rnncontrasts with · A bidirectional recurrent neural network processes a sequence in forward and reverse directions and combines the resulting representations.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.