Back to Home

Neural Network Pro

Interactive multi-layer neural network simulator with comprehensive features

📊 Network Architecture

Input: 3Layer 1: 4Layer 2: 1Output: 1

Total parameters: 21

Input Layer

3 values
x1:
x2:
x3:

Learning Rate

For training (reference parameter)

Network Output

Layer 1 Output(relu)

n1:0.450000
n2:0.000000
n3:0.420000
n4:0.210000

Layer 2 Output(sigmoid)

n1:0.619399

Final Network Output

Output 1:

0.619399

Hidden & Output Layers

Layer 1

Neuron 1 (Weights & Bias)

Weights:

Bias:

Neuron 2 (Weights & Bias)

Weights:

Bias:

Neuron 3 (Weights & Bias)

Weights:

Bias:

Neuron 4 (Weights & Bias)

Weights:

Bias:

Layer 2(Output Layer)

Neuron 1 (Weights & Bias)

Weights:

Bias:

💡 Quick Tips

  • • Click "Add Layer" to create deeper networks (default: 3 neurons with ReLU)
  • Typical architecture: Input → Hidden Layer(s) with ReLU → Output with Sigmoid/Linear
  • • Adjust neurons per layer to control network capacity
  • Hidden layers: Use 4-10 neurons with ReLU activation
  • Output layer: Use Sigmoid for binary, Linear for regression, or multiple neurons for multi-class
  • • Randomize weights to experiment with different initializations
  • • Expand neuron details to fine-tune individual weights and biases

🏗️ Building Real Networks

For Binary Classification:

  • • Hidden: 4-8 neurons (ReLU)
  • • Output: 1 neuron (Sigmoid)
  • • Example: [2, 3] → 6 → 4 → 1

For Multi-Class:

  • • Hidden: 6-10 neurons (ReLU)
  • • Output: N neurons (Linear/Sigmoid)
  • • Example: [4] → 8 → 6 → 3

For Regression:

  • • Hidden: 4-8 neurons (ReLU)
  • • Output: 1 neuron (Linear)
  • • Example: [3] → 6 → 4 → 1

Deep Networks:

  • • Multiple hidden layers
  • • Gradually reduce neurons
  • • Example: [5] → 8 → 6 → 4 → 1