Dynamic Neural Network

Visualize how weights and biases shape decision boundaries.

Positive Weight
Negative Weight
Input Layer
0.50
0.80
Hidden Layer
0.00
b: 0.1
0.00
b: -0.2
0.00
b: 0.05
Output
b: -0.1
0.00
Prediction
System ready. Adjust inputs or randomize parameters, then execute forward pass.

The Mechanics

Weight Polarity & Magnitude

Lines represent weights. Thickness equals importance (magnitude). Color equals polarity.

  • Green paths amplify the signal.
  • Red paths suppress the signal, driving the next neuron toward zero.

Biases

Displayed above each hidden and output neuron. Biases shift the activation function left or right. A negative bias means the incoming weighted sum must be strong enough to overcome the deficit before the neuron activates.

ReLU Activation

Formula: max(0, x)

If the sum of (Inputs × Weights) + Bias is negative, ReLU violently clamps it to exactly 0. This non-linearity is what allows neural networks to learn complex, curving boundaries rather than just straight lines.