This repository contains my implementation of the core concepts of neural networks and deep learning. The focus is on building learning algorithms from scratch using NumPy, with an emphasis on understanding the underlying mathematics, optimization, and algorithmic structure rather than relying on high-level frameworks. The work follows a structured progression from vectorized computation and logistic regression to multi-layer neural networks, forward and backward propagation, and end-to-end model training.
The repository covers:
- Vectorized numerical computation with NumPy
- Logistic regression viewed as a single-layer neural network
- Forward and backward propagation
- Gradient-based optimization
- Construction and training of fully connected deep neural networks
The goal is to make each component of the learning pipeline explicit and interpretable.
- Python 3.x
- Numpy
- Matplotlib
Open the notebooks directly or run them in a local Jupyter environment.