Scalar activation

a = tanh(z)
Weights and bias
Input x₁ = 2 x₂ = −1
w₁ 1.5
w₁x₁ = 1.5 × 23.0
w₂ 0.5
w₂x₂ = 0.5 × (−1)−0.5
b bias −0.5
offset, independent of x−0.5
z = w₁x₁ + w₂x₂ + b. Each term shifts the running total.
w₁x₁ 3.0
w₂x₂ −0.5
b bias −0.5
z = ∑ 2.0
a = φ(z) tanh 0.96
z = 0 the contributions cancel