Index Ai Academy
Glossary · Phase 4 · Deep learning

Output gate

Also written: output gates

The part of an LSTM that decides how much of its memory to show as its state at a step.

Try it now

Gates at 1 and 0 hold a memory. A gate at 0.5 halves it.

Learn it properly

LSTMs and GRUs, built and trained · Build an LSTM and a GRU from a linear layer, train them on the adding problem beside a plain RNN, read what the gates learned, measure the gradient's path, and carry state across a stream. About 40 minutes.

Other terms from the same lesson

  • adding problem: A test of memory for sequence models: add up the two marked numbers in a long list of numbers.
  • forget gate: The part of an LSTM that decides, at each step, how much of its stored memory to keep.
  • input gate: The part of an LSTM that decides how much new content to write into its memory at a step.
  • reset gate: The part of a GRU that decides how much of the old state the new candidate may use.
  • update gate: The part of a GRU that decides how much of its old state to keep, and so how much to replace.
Keep learning

More in plain English.

1286 terms in the glossary, each with the lesson that teaches it.

1286 terms

The glossary

Every word you'll meet in AI and machine learning, explained in everyday language.

Phase 4 · lesson 40 min

LSTMs and GRUs, built and trained

Build an LSTM and a GRU from a linear layer, train them on the adding problem beside a plain RNN, read what the gates learned, measure the gradient's path, and carry state across a stream.

Never trained an AI? Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account. Teach it to write