Harvard's CS109A course was used as reference to model Neural Networks from scratch (https://harvard-iacs.github.io/2020-CS109A/)
- Simple and multi linear Perceptron (MLP)
- MLP from first principles on Iris dataset
- MLP using Keras
- Review of Keras
- Use Keras on Fashion MNSIT dataset
- Gradient Descent methods and plots
- Back propagation from first principles
- Regularization methods to improve neural networks - L1 and L2 norm penalties, early stopping, data augmentation, and dropout.