NETWORK VISUALIZER
Live activations · weighted connections
Strong positiveNegativeNear zeroLine thickness = weight magnitude
Ready to train
Training loss—Binary cross-entropy
Accuracy—0 / 4 correct
Epoch0Optimization steps
LEARNING CURVE
Loss should generally decrease during training
Earlier epochsNo training data yetLatest epoch
TRAINING DATASET
Four examples · two inputs · one binary output
| Input A | Input B | Target | Prediction | Result |
|---|
Choose an example to inspect the network's prediction. The target is the answer we want the network to learn.
WEIGHTS & BIASES
Inspect and edit the actual parameters
A weight controls how strongly one neuron influences the next. Bias shifts the neuron's activation threshold. Edit a value and the prediction updates immediately.
| Connection | Weight |
|---|