// NEUROFORGE
DATASET
Circle
XOR
Spiral
Gaussians
Moons
Linear
Points
200
Noise
0.10
ARCHITECTURE
Hidden layers (neurons per layer)
Input: 2 | Output: 1 | 0 = skip layer
ACTIVATION
Sigm
ReLU
Tanh
TRAINING
Learn Rate
0.030
Batch
Full batch
16
32
64
Train
Step (1 epoch)
Reset Network
New Data
LOSS
STATS
neural network playground
hand-rolled backpropagation
real-time decision boundary
SGD with mini-batches
3 activations // 6 datasets
zero dependencies // bbobop.com
DECISION BOUNDARY
NETWORK