Neural-Network in C++

This is a basic neural network built entirely from scratch in C++ to recognize handwritten digits using the MNIST dataset.

It uses:

• 🧮 Eigen library for matrix operations
• 3 layers:
  • Input layer: 784 neurons
  • Hidden layers: 128 → 64 (with ReLU)
  • Output layer: 10 neurons (with Softmax)
• 📉 Cross-entropy loss
• 🔁 Mini-batch gradient descent
• ✅ Accuracy and confusion matrix evaluation

---

🔧 How to Use

1. Run the preprocessing script to generate CSV files:

   python preprocessing.py