HS2012

Course resources are available here: http://scg.unibe.ch/teaching/cp

The goal of the computer graphics seminar is to introduce students to current topics in computer graphics research. The seminar includes lectures by the instructor and external guests to introduce selected topics. In addition, we will read and discuss recent research papers in these areas. The seminar also provides the opportunity for students pursuing Bachelors and Masters projects to present their work.

The objective of the computer vision seminar is to create a research background and culture (e.g., awareness of what has been done and is being done in vision and machine learning) and to develop critical analysis skills. Each week a presenter will lead the discussion by giving an informal presentation of the chosen paper, or her own research, during the first half of the meeting. In the second half, she will concentrate on an open discussion of the merits of the paper, or project, and how it compares to other work in the literature. The reading material is selected from best-paper awards and most cited papers of the top computer vision and machine learning conferences and journals. The seminar also provides the opportunity for students pursuing Bachelors and Masters projects to present their work. The webpage to the seminar is also http://www.cvg.unibe.ch/cv%20seminar.html

The main objective of this course is to introduce the students to the various techniques and strategies that can be used to • to discover pertinent relationships (or correlations) between variables • to evaluate such relationships and machine learning approaches • to know how to conduct a machine learning process based on available data and to interpret the results. Practical exercises will complete the theoretical presentation.

The natural language proecssing seminar is a first step to working and evaluating linguistic tools.

Seminar on Pattern Recognition and Machine Learning for Human Activity

Coq Proof Assistant enables to produce verified formal proofs interactively by the means of automated “tactics.” With Coq, it is possible to describe precise specifications for programs and obtain code that is provably correct with respect to those specifications. Coq’s rich language also makes it suitable for use in formalizing mathematical statements and theories. Coq is currently one of the most important and influential computer environments for such tasks. The seminar intends to give a gentle introduction to Coq, both from the practical and the theoretical perspective. Textbook: Bertot Yves, Castéran Pierre. "Interactive Theorem Proving and Program Development". Springer, 2004. ISBN 978-3-540-20854-9 Copies of this book will be provided in the seminar.

Optimization problems often involve many simple and interrelated decisions which – combined - yield a set of feasible solutions that is finite, but astronomically large. Applications are numerous, e.g. finding optimal paths or tours in a road network, determining cost optimal production and distribution in a supply chain, scheduling tasks, assigning personnel, etc. Combinatorial optimization deals with modeling and solving such problems, and a core of standard problems, models (often related to graphs), and methods has developed. In this course, we address some of these standard problems, including optimal trees in graphs and independent sets in matroids, maximum and minimum cost flows in networks, maximal and optimal matchings in bipartite and general graphs, optimal postman tours, traveling salesman tours, maximum stable sets and minimum colorings. Literature (mandatory): Cook J. C., Cunningham W. H., Pulleyblank W. R., Schrijver, A., Combinatorial Optimization, Wiley-Interscience 1998 Further recommended: - Lee, J., A First Course in Combinatorial Optimization, Cambridge University Press, 2004 - Korte B., Vygen J., Combinatorial Optimization: Theory and Algorithms, 4th Ed., Springer, Berlin, 2007