Statistical learning

ECTS : 4

Volume horaire : 39

Description du contenu de l'enseignement :

• 1 Examples and machine learning framework: applications, supervised and non-supervised learning
• 2 Useful theoretical objects: predictors, loss functions, bias, variance 
• 3 K-nearest neighbors (k-NN); Higher dimensions and Curse of dimensionality
• 4 Regularization in high dimensions: ridge and lasso (for linear and logistic models)
• 5 Stochastic Optimization Algorithms used in machine learning: Stochastic Gradient Descent, Momentum, Adam, RMSProp
• 6 Naive Bayesian classification
• 7 Deep learning through neural networks : introduction, theoretical properties, practical implementations (Tensorflow, PyTorch depending on acumen)
• 8 Generative and non-supervised learning: k-means

Compétence à acquérir :

Introduction to statistical learning, particularly in a high-dimensional context, including baseline algorithms (k-NN, ...) and modern approaches in deep learning (neural networks). 

Mode de contrôle des connaissances :

cf. CC

Bibliographie, lectures recommandées :

See site of the course (site of the teacher); also see textbook by G. Turinici (cf. Amazon)