The DIY models were built mainly using NumPy for its matrix/vector operations, adhering to OOP concepts, utilising abstract classes and inheritance where appropriate and benchmarked against existing implementations provided by common machine learning libraries.
Models were used in 3 different projects:
- KNN & DT on Hepatitis & Diabetic Retinopathy Debrecen datasets
- Logistic & Multiclass Regression on IMDB Reviews & 20 Newsgroup datasets
- MLP on Fashion-MNIST dataset