The Forward Pass: What a Network Can Draw
The forward pass turns one input \(x\) into one prediction \(\hat y\). Do it for every \(x\) and you get a curve: the shape the network draws. Turn the knobs and see how complex that shape can get, and how fast.
The prediction \(\hat y = f_\theta(x)\), for every \(x\)
prediction: the sum of the neurons, plus \(c\)
target points
thin lines: each neuron \(v_j\,\sigma(w_j x + b_j)\), numbered at its bend.
Drag on the plot to move \(x\).
parameters:
The knobs
The network, at \(x =\)
Blue: the values of the forward pass. Edge thickness: size of the weight; dashed: negative weight.