Modules

Prerequisites

Contents

Brief introduction to statistical learning theory and empirical risk minimization. Supervised and unsupervised learning framework with an optional part on reinforcement learning. Supervised Learning models include linear predictors, support vector machines and neural networks. Unsupervised Learning models include dimensionality reduction via matrix decompositions (singular value decomposition, principal component analysis) or autoencoders, clustering algorithms and empeddings (t-distributed stochastic neighbour embedding).

Learning outcomes

At the end of the module the students are able to build and apply a collection of machine learning models, ranging from simple linear predictors to deep neural networks to common engineering problems. Selection of case studies is focused on manufacturing processes. Students are able to implement data-driven algorithms of increasing complexity directly in Python. Advanced neural network architectures for image processing or time series will be based on Pytorch.

Assessment

• Oral exam (30 min) including presentation with discussion (15 slides)