Neural Network Visualizer
Build a multilayer perceptron and watch it learn — with real forward propagation, backpropagation, and gradient descent. Every number shown is exactly what the network computes.
Architecture2 → 4 → 1
H14
Epoch
0
Loss
0.1258
Network — sample [0, 0] → target [0], predicted [0.500]
Show sample:
positive weight negative weightedge thickness ∝ |weight|
Decision boundary
The exclusive-or problem — not linearly separable, so it needs a hidden layer.
Loss over epochs
Train to see the loss curve
Calculation inspector
Step – / 5
Click a neuron in the network to inspect its exact computation.