Aims

- To understand and apply standard techniques to build computer programs that automatically improve with experience

- To assess the quality of a learned model for a given task

- To assess the relative performance of several learning algorithms

- To justify the use of a particular learning algorithm given the nature of the data, the learning problem and a relevant performance measure

- To use, adapt and extend learning software

Main themes

- Leaning as search, inductive bias
- Combinations of decisions
- Loss function minimization, gradient descent
- Performance assessment
- Instance-based learning
- Probabilistic learning
- Unsupervised classification

Content and teaching methods

- Concept learning, Generalization as Search, Version Space
- Decision trees
- Multilayer Perceptrons
- Quality measures, Confidence intervals, Hypothesis testing
- K-Nearest Neighbors
- Bayesian Learning, Naïve Bayes
- Clustering techniques

Teaching method: lectures (theory), simulations on real data, extensions of learning software, problem based learning

Other information (prerequisite, evaluation (assessment methods), course materials recommended readings, ...)

- Prerequisites:
(1) SINF1121 or equivalent
(2) Basic knowledge in probability and statistics (BIR1203, BIR1304 or equivalent)

- Reference book:
Machine Learning, Tom Mitchell, McGraw Hill, 1997

- Evaluation:
Written exam

-Remarque:
WEB site: http://www.info.ucl.ac.be/notes_de_cours/INGI2262/

Other credits in programs