Workbench

Neural Network Visualizer

Interactive NN Builder

Network Architecture

Presets

Forward Pass Input

Network Stats

Ready — click Forward Pass

Concepts

Neuron

A computing unit that takes inputs, multiplies by weights, adds bias, applies activation.

output = activation(Σ(w·x) + b)
Weights & Bias

Weights control the strength of connections. Bias shifts the activation threshold.

z = w₁x₁ + w₂x₂ + ... + b
Activation Function

Choose an activation function to see its graph and formula.

Activation Graphs

Sigmoid

Backpropagation

Error flows backward through the network to update weights. Chain rule is used to compute gradients.

∂L/∂w = ∂L/∂y · ∂y/∂z · ∂z/∂w