Understanding GRU(Gated Recurrent Unit)

GRU is the simplified version of LSTM and performs as well as LSTM. GRU has less parameters than LSTM and is faster than LSTM. Both state vectors are merged into one sinlge vector h(t) in GRU.

A single gate controller z(t) controls both forget and input gate. If the gate controller outputs 1, the forget gate is open(=1) and the input gate is closed(1-1=0). If it’s output is 0, the opposite happens. In other word whenever a memory must be stored, the location where it will be stored is erased first.

References

• Hands on machine learning by scikit-learn and keras by geron