PROJECT 1 - TEAM MPG - COURSE OF MACHINE LEARNING 2021/2022

---

Repository for the first "Machine Learning" course project by the MPG team, whose members are Matteo Calafà (@Teocala), Giulia Mescolini (@giuliamesc) and Paolo Motta (@paolomotta).

The task proposed is to propose a solution to the Higgs Boson Challenge (see project1_description.pdf for further information), in a python environment, with the usage of the libraries numpy and, for visualization, matplotlib.

Data

Data are contained in the files train.csv and test.csv. They can be downloaded from in https://github.com/epfml/ML_course/tree/master/projects/project1/data and, to run the code, they need to be stored in the data folder.

Structure of the Code

• implementations.py contains the implementation of 6 requested methods to train the model.
• run.py generates the results using the selected model to predict test data.
• utilities.py contains all the functions implemented for pre-processing.
• parameter_selection_utilities.py contains the definitions of the functions used in parameter_selection.py for the hyper-parameters' choice.
• parameter_selection.py generates the optimal hyper-parameters with cross-validation.
• plots.py generates the exploratory plots of the training features.
• proj1_helpers.py contains useful functions for reading/writing on .csv files.

Results

The results used for the submission on AIcrowd can be found in the folder data, under the name submission.csv.

Report

The final 2-page report is in the .pdf document report.pdf.

---

USEFUL LINKS:

• https://www.epfl.ch/labs/mlo/machine-learning-cs-433/
• https://github.com/epfml/ML_course/tree/master/projects/project1
• https://www.aicrowd.com/challenges/epfl-machine-learning-higgs/leaderboards