Objectives:

We introduce the principles and the critical analysis of the main paradigms for learning from data and their applications.
The concepts are progressively introduced starting from simpler approaches up to the state-of-the-art models in the general conceptual framework of modern machine learning. The course focuses on the critical analysis of the characteristics for the design and use of the algorithms for learning functions from examples and for the experimental modelization and evaluation.

Syllabus

- Introduction: Computationl learning tasks, prediction, generalization.

- Basic concepts and models: structure of the hyothesis space, discrete and continuous spaces, linear models, nearest neighbor, prepositional models, inductive bias.

- Neural models: Perceptron and computational properties. Introduction to multilayer feedforward Neural Networks architectures and learning algorithms.

- Rule based models.

- Principles of learning processes and general practical aspects:
Validation, Bias-Variance analysis. Elements of Statistical Learning Theory, VC-dimension. Ensemble learning.

- Support Vector Machines: linear case, kernel-based models.

- Bayesian and Graphical models.

- Unsupervised learning.

- Introduction to Applications and advanced models.