Build Your Quiz
Choose topics, difficulty, timing, and feedback behavior.
Interactive GRU Learning Lab
Explore state blending, parameter counts, tensor shapes, equations, and architecture choices before starting.
Gate and Hidden-State Simulator
This scalar demonstration uses ht=(1-z)hprevious+z·hcandidate. The reset gate previews its effect on earlier memory.
Reset-weighted memory
0.40
New hidden state
0.30
Parameter-Count Calculator
Uses 3H(I+H+2), matching implementations with two bias vectors per gate.
Trainable GRU parameters
1,248
Tensor-Shape Explorer
Input:
Output:
Final state:
GRU, LSTM, and Vanilla RNN Comparison
| Model | Main state design | Gates | Typical strength | Trade-off |
|---|---|---|---|---|
| Vanilla RNN | Single hidden state | None | Simple short-sequence baselines | Long dependencies are difficult |
| GRU | Single gated hidden state | Update and reset | Efficient sequence modeling | Less explicit memory control than LSTM |
| LSTM | Hidden plus cell state | Input, forget, output | Flexible long-memory control | More parameters and computation |
Question 1Answered 0
Remaining
Untimed
Performance Report
Score
Correct
Average time
Grade