Introduction to machine learning

"Artificial intelligence (AI) is a term that is omnipresent in the media, in political and marketing speeches... and in job offers for engineers. This is due to the recent success of machine learning, a scientific discipline that allows computers to "learn" from numerical data. In this course, we will study the scientific and technical foundations of AI. The objective is to give the keys to understand what machine learning is, its possibilities and its limits, in a historical perspective. Numerous practical exercises will illustrate the different notions and allow the implementation of the main learning models.

At the end of the module, students will be able to 

• apply the main learning models to real data from different scientific fields, and evaluate their performance;
• build on their understanding of the foundations of the discipline to identify practical limitations
• identify the ethical issues of AI.

Each session (consisting of one hour in lecture hall and two hours of tutorials in computer room) requires one hour of personal work beforehand: study of the corrections to the tutorials from the previous session and reading of about twenty pages of the handout.

Some of the concepts covered:

• fundamental difficulties of learning: curse of dimension and bias-fluctuation and bias-variance dilemmas
• data partitioning: hierarchical classification, k-means
• supervised classification and regression: logistic regression, support vector machines, perceptron, artificial neural networks, deep learning, and related learning algorithms

The course and practical work will take the form of Jupyter notebooks with Python/scikit-learn (software programs used in professional environments).

All the course material (handout, presentation materials, topics and corrections of the tutorials) is available on the University's Arche platform.

A handout is distributed. It is freely available at the following URL: https://members.loria.fr/FSur/enseignement/apprauto/poly_apprauto_FSur.pdf

to learn the vocabulary of machine learning and to understand the challenges and limitations of artificial intelligence

to know how to interpret the different performance metrics of the models

to evaluate the performance of different models