Machine Learning

Recommended Prerequisites

Probabilities and Statistics, Linear Algebra, Calculus, Programming.

Teaching Methods

The curricular unit consists of theoretical-practical classes (TP) that provide a detailed presentation of concepts, principles, and fundamental theories using audiovisual media. Basic practical exercises are solved during these classes to consolidate the theoretical concepts. Additionally, practical laboratory classes (PL) are conducted, where students work on exercises that require the integration of various theoretical concepts and aim to foster critical reasoning skills. The evaluation process encompasses all the topics covered in the course and focuses on assessing both the understanding of the fundamental theoretical concepts as well as the ability to solve complex problems from the real world.

Learning Outcomes

The student should be able to understand the Machine Learning pipeline and the distinction between supervised and unsupervised learning, mastering feature engineering techniques to optimize data. The student must apply rigorous performance evaluation methods (metrics, partitioning) to quantify performance and diagnose issues such as the bias-variance dilemma and the curse of dimensionality. Another core objective is to master the main techniques of unsupervised learning (Clustering) and supervised learning (kNN, SVM, Discriminants), understanding their theoretical foundation and application. Finally, the student must apply advanced classifier combination strategies (Boosting, Bagging) to increase the robustness and accuracy of predictive models.

Work Placement(s)

No

Syllabus

1. Introduction

-Pipeline review

-Supervised vs unsupervised learning

-Classification, regression and generation

- Topics on features engineering

-Generalized decision functions

-Approaches for multi-class classification

2.Model evaluation

-Performance assessment measures

-Data partitioning

-The variance-bias dilemma.

-The problem of the curse of dimensionality

3.Unsupervised learning

-Partition clustering

-Hierarchical clustering

-Density based clustering

-Clustering based on combined gaussian models

4.Supervised learning

-Linear discriminants

-Probabilistic classification

-Non-parametric classification: kNN and SVM

-Rule-based models

5. Combination of classifiers

-Product and sum rule

-Boosting, Bagging and Stacking

Head Lecturer(s)

César Alexandre Domingues Teixeira

Assessment Methods

Assessment
Project: 40.0%
Exam: 60.0%

Bibliography

-Bishop, C.M., Pattern Recognition and Machine Learning, Springer, 2006.

-Duda, R. O., Pattern Classification, 2nd ed. Wiley Interscience (2001).

-J.P. Marques de Sá, Pattern Recognition: Concepts, Methods and Applications, 2001, XIX, 318 p., 197 illus., Springer (2001).

-M. N. Murty, Pattern Recognition: An Algorithmic Approach, Springer, 1st Edition., XII, 263 p. (2011).

-Peter Flach, Machine Learning: the art and science of algorithms that make sense of data, Cambridge University Press, 2012.

-Trevor Hastie, The Elements of Statistical Learning (2nd Edition), Springer, 2008.

-Introduction to Machine Learning with Python, Andreas C. Muller and Sarah Guido, O'Reilly, 2017.

-Deisenroth, Marc Peter, Mathematics for machine learning. Cambridge UP, 2020.

-Géron, Aurélien. Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow,O'Reilly, 2022.

-Murphy, Kevin P. Probabilistic machine learning: an introduction. MIT press, 2022.