Machine Learning

• Neural Networks and information theory

The aim of the course is to familiarize the student with the modern concepts of machine learning at the international research level. In particular:

• The student understands the concepts of Bayesian inference and use it to derive a number of different machine learning methods, such as regression models, kernel methods, classification models, graphical models 
• The student is familiar with stochastic networks of interacting variables, thermodynamic concepts, and sampling methods
• The student is familiar with a number of approximate inference methods, such as the variational method, belief propagation
• The student is capable to write computer programs to implement the above methods
  

This course is an advanced course on machine learning and neural networks from a probabilistic point of view. Starting in 2010/2011, the course is a continuation of the course Neural Networks and information theory and will contain advanced topics. The course is intended for master students in physics and mathematics. Students with a background in computer science, AI or cognitive science are recommended to follow the course Statistical Machine Learning instead.

The course provides a good preparation for a Masters' specialisation in Theoretical Neuroscience or Machine Learning.

See http://www.snn.ru.nl/~bertk/machinelearning/ for futher information.

• David MacKay, Information Theory, Inference and Learning Algorithms, Cambridge University press. The entire book can be viewed on-screen at http://www.inference.phy.cam.ac.uk/mackay/itila/book.html
• several handouts will be distributed during the course

• 28 hrs lecture
• 28 hrs problem session

Part of the course will be presented by the students. Part of the course will make use of videolectures. 

Examination is weighted average of oral exam, homework assignments and presentations.

http://www.snn.ru.nl/~wimw/collegeML.html