Gossip Relay Layer

x = 0.00
Sources Relays Story Bias w>0 w<0 |w| thick
x x₂ x₃ b Relay 1 Relay 2 Story 0.00 -0.35 0.25 h₁=0.00 h₂=0.00 y=0.00

Unified notation: weights w₁…w₈, biases b₁,b₂, inputs x,x₂,x₃. Toggle “Weights” to show labels; “Show z” reveals pre-activations.

Activation Function

x ∈ [−5,5] x = 0.00
f(x)=0.000
ReLU
Range: [~−1,~1]