Weight Update Simulator
Step through gradient descent as error and learning rate move each weight.
Epoch boundary — no row being evaluated.
Start with weights [0, 0] and bias 0, learning rate 0.2. Output is 1 when w·x + b ≥ 0.
step 1 of 31
Weights
- w1
- 0
- w2
- 0
- bias
- 0
- Epoch
- 0
- Errors this epoch
- 0
training…
Setup
Learning rate
The cyan line is the decision boundary. Each update rotates and shifts it until every point sits on the correct side.
← → to step, or use the buttons above.