Thu Dec 26 2024 00:21:27
After completing Andrew Ng’s Machine Learning course, I wanted to take my understanding of neural networks to the next level. While building one in PyTorch was rewarding, I couldn’t shake the question:
What if I built it completely from scratch using only NumPy?
That question turned into a challenge, and now I’m excited to share the result: NumPy Neural Network for MNIST Dataset.
What’s inside
This project is a pure NumPy implementation that dives into the fundamentals of forward propagation, backpropagation, and activation functions—no high-level frameworks, just the raw math and logic behind neural networks.
It was a challenging yet rewarding experience that gave me a deeper appreciation for the inner workings of these models.
