UNIL
Vous êtes ici: UNIL > L'enseignement > Fiche de cours
Français | English   Imprimer   

Fiche de cours

Hands-on introduction to machine learning for biomedical data - Spring session - 38a/Série 3

Faculté de gestion: Ecole doctorale (FBM-DOCT)

Responsable(s): Jonas Richiardi

Période de validité: 2020 ->

Horaires du cours (Apériodique)

Date Lieu Remarque Thématique Intervenant(s)
16.04.2024 de 13:00 à 16:00 salle 315.1, bâtiment Amphipôle, UNIL-Sorge Linear supervised learning: linear algebra essentials, matrix inversion, dimensionality, regularization (L1, L2, Elastic Net). Packages: numpy, seaborn, scikit-learn. Model: Logistic Regression. Open dataset: Allen Institute brain gene expression. Costa Georgantas, Jonas Richiardi
23.04.2024 de 13:00 à 16:00 salle 315.1, bâtiment Amphipôle, UNIL-Sorge Model evaluation: bias/variance, cross-validation, metrics. Kernel methods. Model: support vector machine (SVM). Open dataset: Cleveland heart disease. Julien Bodelet, Jonas Richiardi
30.04.2024 de 13:00 à 16:00 salle 315.1, bâtiment Amphipôle, UNIL-Sorge Feature selection: wrappers, filters, double-dipping. Ensembling: bootstrapping, bagging. Hyperparameter optimization: search, nested cross-validation. Model: random forest. Oliver Yibing Chen, Costa Georgantas
07.05.2024 de 13:00 à 16:00 salle 315.1, bâtiment Amphipôle, UNIL-Sorge Clustering: distance metrics, k-means, hierarchical clustering and dendrograms, quality indices. Costa Georgantas, Jonas Richiardi
14.05.2024 de 13:00 à 16:00 salle 315.1, bâtiment Amphipôle, UNIL-Sorge Neural networks and deep learning overview. ML4biomed Micro-hackaton. Costa Georgantas, Jonas Richiardi

Cours (optionnel)

Annuel
Apériodique
Langue(s) d'enseignement: anglais, français
Public: Oui
Crédits: 1.5

Objectif

Machine learning approaches are gaining importance in all fields of medicine and biology. They offer multivariate predictive modelling tools focusing on individual subjects rather than group-level inference. The main goal of this tutorial is to equip students with enough knowledge and practical experience with Python to start applying machine learning techniques to their own research project.

Contenu

This hands-on, flipped classroom course will start by linear supervised learning methods, focusing on essential linear algebra notions, matrix inversion, and regularization. We will cover basic decision theory, binary and multi-class problems, linear and non-linear discriminative models (in particular logistic regression, support vector machine, random forest), feature selection, model evaluation. We will also cover clustering and briefly introduce deep learning. In each session we will implement the discussed algorithms using standard libraries such as scikit-learn, and use open biomedical data from genomics and clinical domains. The course will culminate with a group micro-hackaton using data of your choice, possibly your own!     

Evaluation

Travail personnel : Oui
Présentation : Oui
Test final : Non
Évaluation de la participation par le tuteur: Oui

Bibliographie

1) Hastie et al, The Elements of Statistical Learning, 2nd ed, Springer (available as free PDF)

2) Duda et al., Pattern Recognition, 2nd ed, Wiley

3) Bishop, Pattern Recognition and Machine Learning, Springer (available as free PDF) 4) Deisenroth et al., Mathematics for machine learning, Cambridge University Press (available as free PDF)

4) Deisenroth et al., Mathematics for machine learning, Cambridge University Press (available as free PDF)

Exigences du cursus d'études

Basic linear algebra and calculus; Basic statistical literacy; Demonstrable Python programming experience including Jupyter notebooks; Previous use of packages numpy, pandas, and matplotlib. Relatively recent personal laptop (any OS).

Conditions d'octroi

Evaluation positive de la participation active par le tuteur.

Conditions d'accès

Inscription: auprès de l'Ecole doctorale. Série 3

Unicentre - CH-1015 Lausanne - Suisse
Tél. +41 21 692 11 11
Canton de Vaud
Swiss University