∇ Backprop + Training

2-input sigmoid neuron · MSE loss
output—
target—
loss—
① Forward
② Backward
③ Update w
w₀ x₀ w₁ x₁ w₂ × × + dot ×−1 exp +1 1/x σ(dot)=ŷ ŷ target MSE (ŷ−y)² Loss L ── σ(x) = 1/(1+e⁻ˣ) ──
Loss Curve
✓ Converged epoch 0
L = — min —