Long short-term memory (LSTM)

LSTM is a type of recurrent neural network built to retain information across long sequences. It helps machines learn patterns that develop over time, instead of losing earlier context as new data arrives.

Example

LSTM networks are useful in applications such as speech recognition, where the model must process long sequences of spoken words. By capturing timing, context, and changes in the input, an LSTM can transcribe speech into text, even when accents, speaking speed, or background noise vary.