In Advanced Concepts in Machine Learning a selected number of recent developments in the field are presented and experimented with. Machine learning deals with the prediction of labels or real values for unseen objects, based on a set of previously encountered examples, or automatically adapting behaviour according to previous experience. In the past years, topics such as Deep Neural Networks, Recommender Systems, Relational Learning, Reinforcement Learning, Support Vector Machines and Gaussian Processes have made their appearance in the course. Besides an overview of recent Machine Learning techniques, the course also highlights the importance of representations in successful applications of machine learning. In this context, propositional representations are contrasted with multi-instance and relational representations, but also automatically generated representations through Sparse Coding, Auto encoders, Deep Belief Nets and indirect representations such as distances and kernel receive substantial attention. After completion of this course, the students are able to select most suited representations and best-fits of learning techniques for a given machine learning problem and reason about the limitations of the suggested selections.
Desired Prior Knowledge: Familiarity with the basics of machine learning through a Machine Learning or Data Mining course
Recommended literature: Pattern Recognition and Machine Learning - C.M.Bishop; Bayesian Reasoning and Machine Learning - D. Barber; Gaussian Processes for Machine Learning - C.E. Rasmussen & C. Williams; The Elements of Statistical Learning - T. Hastie et al.