intermediate
Recurrent Neural Networks
Processing sequences — from vanilla RNNs through the vanishing gradient problem to LSTMs and GRUs.
Hidden StateBackpropagation Through TimeVanishing GradientLSTM GatesGRUState Space ModelsS4 & Mamba
ML project loop2/6 stages →
DataHypothesis spaceObjectiveOptimizationEvaluationDeployment feedback
5 lessons
01Recurrent Neural Networks
Networks with memory — how a hidden state lets a model process sequences one step at a time.
Concept22m
02BPTT & the Vanishing Gradient
How RNNs learn through time — and why long-range dependencies are so hard.
Concept14m
03LSTM & GRU
Gated cells that give RNNs long memory — the cell state, the three gates, and the leaner GRU.
Concept24m
04State Space Models (S4 & Mamba)
Sequence models that are recurrent in theory but convolutional in training — linear state-space recurrence, the parallel-scan speedup, and how Mamba's selectivity rivals attention at sub-quadratic cost.
Concept18m
05RNNs Quiz
Test your understanding of recurrence, BPTT, vanishing gradients, and LSTM gating.
Quiz10m