Gated Recurrent Units
A simplified gated cell that merges the forget and input gates into one update gate and folds the cell state into the hidden state, cutting parameters by roughly a quarter. GRUs usually match LSTM accuracy on short-to-medium sequences and train faster, but the LSTM's separate cell state can still win on very long dependencies.
This Concept is waiting for its first lesson!
A simplified gated cell that merges the forget and input gates into one update gate and folds the cell state into the hidden state, cutting parameters by roughly a quarter. GRUs usually match LSTM accuracy on short-to-medium sequences and train faster, but the LSTM's separate cell state can still win on very long dependencies.
Are you a teacher? Sign in to start contributing.
Sign In