Machine Learning

Linköping University

Goals

The overall aim of the course is to provide an introduction to machine learning, with special focus on regression and classification problems. Machine learning is presented from a probabilistic perspective with inference and prediction based on probability models. The course aims to give students an overview of machine learning within a unified framework and a good basis for further studies in the field.
After completing the course the student should be able to:
• use relevant concepts and methods in machine learning to formulate, structure and solve practical problems.
• infer the parameters in a number of common machine learning models.
• use machine learning models for prediction and decision making.
• evaluate and choose among models.
• implement machine learning models and algorithms in a programming language.

Prerequisites

Students entering the course should have passed at least one course in basic statistics and be familiar with linear statistical models, in particular simple and multiple regression. Also, it is a prerequisite that the students have passed courses in calculus and linear algebra.

Organization

Contents

Introduction and overview of machine learning and its applications. Unsupervised and supervised learning. Discriminative and generative models. Prediction. Generalization. Classification. Nearest neighbors. Naïve Bayes. Discriminant analysis. Cross-validation. Model selection. Overfitting. Bootstrap. Regression. Regularization. Ridge regression. Lasso. Variable Selection. Binary and multi-class regression. Dimension reduction. PCA. ICA. Kernel smoothers. Support Vector Machines. Decision trees. Neural networks. Deep learning.

Literature

- Pattern recognition and machine learning by C.M. Bishop, ISBN 9780387310732.
- The Elements of Statistical Learning by T. Hastie, R. Tibshirani and J. Friedman, Second Edition, ISBN 9780387848587, pdf available for free.

Back