KITT

Modelling sequences: hidden state, backpropagation through time and the vanishing gradient, LSTM and GRU, why transformers displaced recurrence for sequence-to-sequence work — and why recurrence came back as state-space models and linear attention for inference-heavy deployment.