This project was completed for the CS-433 Machine Learning course taught at EPFL in Fall 2022. The goal of this project is to learn to use basic machine learning concepts on a real-world dataset, start to finish. First, an exploratory data analysis is performed to understand your dataset and the features. Then, feature processing and engineering is done to clean the dataset and extract more meaningful information. Afterwards, linear and logistic regression are implemented with some tweaks and used on the data. Finally, the methods are analyzed and inference is performed.

The final report for this project is embedded below. It focuses on the most important points but does not exhaustively cover everything done in the project. For more information on the project implementations, visit the GitHub repository linked above.

Here are some keypoint of the project:

Higgs Boson Classification
Ridge Regression
Ridge Logistic Regression
Kernels
Accelerated Gradient Descent with Restart

The main finding from this project is that basic methods like linear or logistic regression can still be used to create a powerful non-linear model for binary classification using kernel methods. This simple approach can compete with more sophisticated models like neural networks and random forests while being much simpler to implement and understand.