Note that this course will be postponed to the SS

The goal of this course is to introduce the student to key machine learning concepts and promote the development of technical skills that will enable the application of theory on selected problems from the fields of natural language processing, image analysis and audio processing. The course will mainly adopt a classification perspective and will cover traditional machine learning theory along with more recent trends, reaching up to probabilistic graphical models, dictionary learning and common deep neural networks architectures. Furthermore, clustering methods will be treated as complementary material. The presentation format will be largely based on the explanation of the underlying theory for each topic followed by practical examples in Python and Matlab that highlight the key features of the presented methods. It is assumed that the student has basic knowledge of linear algebra, statistics, function optimization theory and basic programming skills. Paper exams contribute 70% to the final grade and the remaining 30% comes from a programming assignment related to the development of a machine learning based system that analyzes a publicly available dataset.

An outline of the course structure is the following: Introduction to Machine Learning, Bayesian Theory fundamentals, Cost Function Optimization and related Classifiers, Neural Networks and Deep Learning, Data Transforms (Feature Generation/Dimensionality Reduction), Feature Selection, Template Matching, Hidden Markov Modeling, Probabilistic Graphical Models, Dictionary Learning and Clustering.

The goal of this course is to introduce the student to key machine learning concepts and promote the development of technical skills that will enable the application of theory on selected problems from the fields of natural language processing, image analysis and audio processing. The course will mainly adopt a classification perspective and will cover traditional machine learning theory along with more recent trends, reaching up to probabilistic graphical models, dictionary learning and common deep neural networks architectures. Furthermore, clustering methods will be treated as complementary material. The presentation format will be largely based on the explanation of the underlying theory for each topic followed by practical examples in Python and Matlab that highlight the key features of the presented methods. It is assumed that the student has basic knowledge of linear algebra, statistics, function optimization theory and basic programming skills. Paper exams contribute 70% to the final grade and the remaining 30% comes from a programming assignment related to the development of a machine learning based system that analyzes a publicly available dataset.

An outline of the course structure is the following: Introduction to Machine Learning, Bayesian Theory fundamentals, Cost Function Optimization and related Classifiers, Neural Networks and Deep Learning, Data Transforms (Feature Generation/Dimensionality Reduction), Feature Selection, Template Matching, Hidden Markov Modeling, Probabilistic Graphical Models, Dictionary Learning and Clustering.

the course is based on the following text books

[1] S. Theodoridis, K. Koutroumbas, “Pattern Recognition”, 4th edition, Academic Press, 2009. 

[2] S. Theodoridis, “Machine Learning: A Bayesian and Optimization Perspective”, Academic Press, 2015.

[3] S. Theodoridis, A. Pikrakis, K. Koutroumbas, D. Cavouras, Academic Press, “Introduction to Pattern Recognition: a MATLAB approach”, 2010.

[4] Christopher M. Bishop, “Pattern Recognition and Machine Learning”, Springer, 2006.

solid knowledge in Signal Processing is required, e.g. SP1 and SP2