Projects
NumPy Deep Neural Network
Machine Learning · Neural Network from Scratch
About
A custom-built deep neural network implemented entirely from scratch in Python, without relying on standard machine learning frameworks like PyTorch or TensorFlow. This project demonstrates a fundamental understanding of neural network mechanics by manually implementing forward and backward propagation, weight initialization, and gradient descent. The resulting program is a highly configurable, foundational model capable of executing both classification and regression tasks.
Key Highlights
—Engineered the entire neural network architecture from scratch, including forward and backward propagation and parameter updates, supporting arbitrary network depths and configurable hidden unit activations (ReLU, sigmoid, tanh).
—Executed both classification and regression tasks using minibatch gradient descent, successfully building out the math and matrix operations manually.