Neural Networks and Learning Systems

Aim

The aim is that students after passing the course will be able to design and apply artificial neural networks and similar methods for signal, image and data analysis that learn from previous experience and data. Students will also be able to apply such methods to find meaningful relations in multidimensional signals where the degree of complexity makes traditional model-based methods unsuitable or impossible to use. Specifically, students should be able to:

• Explain the difference between particular learning paradigms
• Implement and use common methods in those paradigms
• Select an appropriate method for solving a given problem

Content

Machine learning, classification, pattern recognition, kernel methods and high-dimensional data analysis. Supervised learning: neural networks, linear discriminants, support vector machines, ensemble learning, boosting, deep learning. Unsupervised learning: patterns in high-dimensional data, dimensionality reduction, clustering, principal component analysis. Reinforcement learning: TD-learning, Q-learning.