NeuNet is a comprehensive neural network implemented entirely from scratch using NumPy. This project demonstrates the fundamental concepts of deep learning.
git clone https://github.com/Abhinavexists/NeuNet.git
cd NeuNet
pip install numpy matplotlib plotly networkx pandasfrom src.models.neural_network import NeuralNetwork
from src.layers.core import Dense
from src.layers.activations import ReLU, Softmax
from src.layers.regularization import BatchNormalization, Dropout
from src.layers.losses import CategoricalCrossentropy
from src.layers.dataset import create_data
# Create synthetic dataset
X, Y = create_data(samples=100, classes=3, plot=True)
# Build the network
model = NeuralNetwork()
model.add(Dense(2, 128, learning_rate=0.002, optimizer='adam'))
model.add(BatchNormalization())
model.add(ReLU())
model.add(Dropout(0.1))
model.add(Dense(128, 64, learning_rate=0.002, optimizer='adam'))
model.add(ReLU())
model.add(Dense(64, 3, learning_rate=0.002, optimizer='adam'))
model.add(Softmax())
# Set loss function with regularization
model.set_loss(CategoricalCrossentropy(regularization_l2=0.0001))
# Train with advanced features
model.train(X, Y, epochs=500, batch_size=32, patience=30, verbose=True)
# Evaluate
predictions = model.predict(X)
accuracy = model.predict_proba(X)
print(f"Final Accuracy: {accuracy:.4f}")- Dense Layers: Fully connected layers with He initialization
- Activation Functions: ReLU, LeakyReLU, Sigmoid, Tanh, Softmax
- Loss Functions: Categorical crossentropy with L1/L2 regularization
- Optimizers: SGD with momentum, Adam optimizer with bias correction
- Regularization: Batch normalization, dropout, weight penalties
- High Accuracy: 95%+ on synthetic datasets
- Fast Training: 3-5x speed improvement with Adam optimizer
- Robust: Early stopping prevents overfitting
- Stable: Numerical stability built into all operations
Our comprehensive documentation is available at abhinavexists.github.io/NeuNet
- Home: Project overview and quick navigation
- Implementation Guide: Detailed technical walkthrough
- Development Timeline: Evolution of the project
- Reference: Complete component documentation
# Sophisticated architecture with all features
model = NeuralNetwork()
# Input processing with normalization
model.add(Dense(2, 128, learning_rate=0.002, optimizer='adam'))
model.add(BatchNormalization())
model.add(ReLU())
model.add(Dropout(0.1))
# Deep feature extraction
model.add(Dense(128, 64, learning_rate=0.002, optimizer='adam'))
model.add(BatchNormalization())
model.add(ReLU())
model.add(Dropout(0.1))
# Final classification
model.add(Dense(64, 32, learning_rate=0.002, optimizer='adam'))
model.add(ReLU())
model.add(Dense(32, 3, learning_rate=0.002, optimizer='adam'))
model.add(Softmax())
# Advanced loss with regularization
model.set_loss(CategoricalCrossentropy(regularization_l2=0.0001))
# Professional training setup
model.train(X, Y, epochs=500, batch_size=32, patience=30, verbose=True)from src.utils.metrics import calculate_accuracy, confusion_matrix, precision_recall_f1
# Get predictions
predictions = model.predict(X_test)
probabilities = model.predict_proba(X_test)
# Calculate metrics
accuracy = calculate_accuracy(y_true, probabilities)
cm = confusion_matrix(y_true, predictions, num_classes=3)
precision, recall, f1 = precision_recall_f1(y_true, predictions, num_classes=3)
print(f"Accuracy: {accuracy:.4f}")
print(f"Precision: {precision}")
print(f"Recall: {recall}")
print(f"F1-score: {f1}")from src.utils.network_data import export_network
from src.utils.Visualization import network_visualization
# Export network structure
dense_layers = [layer for layer in model.layers if hasattr(layer, 'weights')]
export_network(*dense_layers[:4])
# Create interactive visualization
fig = network_visualization("src/utils/network_data.json")
# Opens interactive HTML visualizationWe welcome contributions! Please see CONTRIBUTING.md for guidelines.
| Metric | Value |
|---|---|
| Classification Accuracy | 95%+ |
| Training Speed (vs SGD) | 3-5x faster |
| Components | 15+ core modules |
| Activation Functions | 5 implemented |
| Optimization Algorithms | 2 advanced |
| Regularization Techniques | 3 methods |
This project is licensed under the MIT License - see the LICENSE file for details.