A live, interactive 3D visualization of a feed-forward neural network. Drag to orbit, scroll to zoom, click a neuron to inspect it. Adjust the network below and watch signals propagate through the layers in real time.
Network Architecture
3
Number of hidden layers between input and output.
7
Width of each hidden layer. More neurons = more connections.
0.05
Controls how strongly weights shift color/thickness after each simulated update — higher η = bigger, faster swings.
Simulation
1.0×
Activation Function
ReLU: outputs 0 for negative input, linear for positive. Fast, avoids vanishing gradients.
View
Input neuron
Hidden neuron
Output neuron
Positive weight
Negative weight
Active signal pulse
◎ Inspector
Click any neuron to see its details — layer, activation value, and incoming/outgoing connections. This panel updates live as signals flow through the network.