Organization: J. Lücke and J. Triesch
Type: Course, 3 SWS
Time and place: Friday 11:15 - 14:00, FIAS 200
Open-minded students from a number of disciplines including Physics, Mathematics, Computer Science, Biology, and Psychology are welcome. Good analytical skills are essential.
In this course we study unsupervised learning, i.e. learning without a teacher. The course combines perspectives from modern machine learning and neuroscience. After a review of fundamental concepts from probability, statistics, and information theory, we study various architectures and algorithms for different aspects of unsupervised learning. We will be interested in the fundamental computational problems associated with various unsupervised learning tasks, algorithms for (approximately) solving these problems, and neurally plausible implementations of such algorithms.
- Elementary Probability Theory
- Overview and Introduction
- Information Theory
- Classification, Density Estimation, Maximum Likelihood and EM
- EM and Mixture Models
- Kernel Density Estimation
- Hebbian Learning 1
- Probabilistic PCA and Sparse Coding