LSTM
In plain terms
A recurrent network architecture good at long sequences.
Going deeper
Long Short-Term Memory networks use gated memory cells to retain information over long sequences, mitigating the vanishing-gradient problem. They remember relevant context.
Under the hood
Gates control what to keep, forget, and output; largely superseded by transformers for many tasks.
Examples
Time series, speech recognition, text generation (pre-transformer), sequences.
Think of it like...
Like a note-taker who decides what to keep, discard, and recall as a story unfolds.