Reset gate
A reset gate is a GRU mechanism that controls how much of the previous hidden state contributes to computing a candidate state.
Natural Language Processing in Action, Second Edition
A reset gate is a gated recurrent mechanism that controls which portions of the previous hidden state are discarded as irrelevant to the current output.
- How it is used
- Controls which previous hidden-state information a GRU treats as irrelevant to the current output.
- How deep it goes
- Explains the gate’s role and gives its sigmoid and hidden-state multiplication formulas, without detailing training dynamics.
- Examples
- Uses sigmoid and tanh matrix formulas showing how the reset vector modulates the recurrent hidden-state contribution.
- What you can do after
- Build it from scratch by computing a sigmoid reset vector and applying it to the hidden-state term in a GRU activation.
Read how Natural Language Processing in Action, Second Edition explains reset gate →
Graph Neural Networks in Action
A gated recurrent unit component that controls how much prior hidden-state information is discarded when computing a new state.
- How it is used
- Controls forgetting within GRU-based sequence models and graph decoders that predict evolving node states.
- How deep it goes
- Explains the gate’s forgetting role and GRU context, but stops short of deriving its activation equations.
- Examples
- Uses pose-sequence GRUs and a neural relational inference decoder with node states, edge messages, and future-frame predictions.
- What you can do after
- Apply via PyTorch by configuring a GRU or custom GRU for sequential or graph-based state updates.
Read how Graph Neural Networks in Action explains reset gate →
Related concepts
- GRUpart of · 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.
- sigmoiduses · A sigmoid is a smooth, bounded, S-shaped function that maps real-valued inputs between two limiting values, commonly zero and one.
- forget gatecontrasts with · A forget gate is an LSTM component that controls how much information from the previous cell state is retained or discarded.
- gated updatecontrasts with · A gated update is a state or representation update in which learned gates control how new information is combined with or replaces existing information.
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.