ELEN0062 - Introduction to Machine Learning (2024-2025)

This page contains practical information concerning projects for the Introduction to Machine Learning course.

Teachers: Pierre Geurts, Louis Wehenkel.

Teaching Assistants: Sacha Lewin (sacha.lewin@uliege.be), Yann Claes (y.claes@uliege.be)

Please prioritize eCampus for asking questions, and check for previously asked questions there.

Projects

Three projects are planned throughout the semester. They will be uploaded here as soon as they are available. For submitting, please refer to the information at the beginning of the statement of the project.

Project 1 - Classification Algorithms

Updates:

• Oct. 3: Typo in kNN question (make_data1 -> make_dataset)

Project 2 - Bias & Variance

Project 3 - Competition

Please see Gradescope for the leaderboard.
Updates:

• Dec. 12: Fixed typo in Section 4 (LS -> TS), and updated the end of competition (13 -> 15 Dec).

Resources

Here are some resources that might be useful for the projects.

Running your Code

Some calculations might be too heavy to run locally on your own computer. Here are some free resources that you can use to run Jupyter notebooks on powerful machines.

Colab is free, quite popular and can run quite big machine learning models, which should be more than enough for you.

Feel free to contact us if you need technical help, we are here for that!

Writing your Report (LaTeX)

LaTeX is the standard method for writing academic papers. It allows to create well-formatted documents, and is especially useful for writing mathematical equations. We strongly advise, although it is not compulsory, to use LaTeX for writing your reports.

We provide a ready-to-use template for writing your reports. We recommend cloning the project on Overleaf for collaborating online with your groupmates. You can find how to copy the Overleaf project on this page. Feel free to check the Overleaf documentation for learning the basics.

Python

Python is a programming language that is widely used in machine learning. It is easy to learn and has a lot of libraries that are useful for machine learning.

Installing Python

There are many ways to download and run Python, with its related libraries. One common and handy manner consists in using virtual environments, where packages can be installed individually. This can be achieved, for instance, with Anaconda, which comes with hundreds of installed packages as well as with a Python IDE (Spyder). If you don't want to install all these packages, prefer using your own text editor and running code from the command line, you can also install Miniconda, which only comes with conda (package manager), its dependencies and the selected version of Python. In any case, make sure to download the installer that corresponds to your system, and a 3.6+ version of Python! Also, note that when using virtual environments, you will need to install Python for each newly created environment.

Managing Environments & Packages

The projects require working with packages, mainly Scikit-learn, Matplotlib and Numpy. In order to install them, it is a common practice to create a new virtual environment, for instance within the Anaconda Prompt:

conda create --name iml python=3.8

Then, activate the environment:

conda activate iml

Finally, install the packages:

conda install scikit-learn

Note that Scikit-learn requires Numpy and Matplotlib, which will be installed automatically. You will then be able to use this package in any future Python code, as long as you remain in this environment. To import this package, begin your Python code with the following line:

import sklearn

Alternatively, you can write code in Jupyter notebooks. However, we require you to submit Python files (.py)!

Tutorials

Here are some tutorials that might be useful.