Concept
LSTM
LSTM (Long Short-Term Memory) is a type of recurrent neural network (RNN) designed to model sequences while avoiding the vanishing-gradient problem that cripples plain RNNs on long inputs. Like a vanilla RNN, it processes tokens one at a time and carries a hidden state…
The rest of “LSTM” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→