Backprop
step by step
ReLU
Sigmoid
← Back
Next →
Original numbers
Random weights
0.5
editable: click a number and type a new one, ↑/↓ change it by 0.1
1.5
value (forward pass)
∂ −2
gradient ∂L/∂(·) (backward pass)
dashed: inactive ReLU neuron
Keys: → / ← / PageDown / PageUp