Forward Pass

Forward Pass

The forward pass computes the output of the network from the input values.

Formula

For each non-input neuron:

z = sum(w_i * h_i) + b

Then the activation function is applied:

h = f(z)

Where:

• w_i is an incoming weight.
• h_i is the activation from the previous layer.
• b is the bias.
• z is the weighted sum.
• f is the activation function.
• h is the output activation.

What the Canvas Shows

During the forward pass, the canvas can show:

• Active node.
• Current connection.
• Weight values.
• Bias values.
• Weighted term values.
• Node activation.

Example

For a neuron with two inputs:

z = (w1 * x1) + (w2 * x2) + b
h = sigmoid(z)

The formula panel shows the current substitution as the animation advances.

Output Layer

The final layer produces the prediction.

For a single-output binary example:

y_hat = output activation

A simple class prediction can be computed using:

class = 1 if y_hat >= 0.5 else 0