Graphical modelling and network inference
Inverse problems in statistical physics are motivated by the
challenges `big data' in different fields, especially
high-throughput experiments in biology. Key question is how to
infer parameters of a model which describes the statistics of the
data and how to link those parameters to the processes generating
data. In this seminar, we focus on network inference using Bayesian
networks and explore links to both statistical and quantum physics.
Specific topics include
This seminar is part of the StatBio module of the Master's in Physics.
Specific topics include
- Probabilistic inference and Bayesian statistics (Anton Walker, 13.11)
- Boltzmann machine learning and mean field approximation (Hooman Farhang Ranjbar, 27.11)
- Causal interpretation of Bayesian networks (Alexander Klug, 4.12)
- Information-theoretic methods for graphical model learning (Simon Fischer,11.12.)
- Bayesian networks & quantum non-locality (Leonard Fischer,18.12)
- Variational methods in graphical inference (Mario Josupeit,8.1
cancelled! ) - Belief propagation (Jonas Rzezonha, 15.1.16, discussion room first floor ETP)
- l_1 regularisation (Aditya Kela, 29.1.)
- Algebraic statistics, quantifier elimination (Daniel Suess,5.2.)
- 12.2. Guest lecture by José L. Casadiego Bastidas (MPI for Dynamics and Selforganization, Göttingen): Network Dynamics as an Inverse Problem: Reconstruction from time series
This seminar is part of the StatBio module of the Master's in Physics.
Schedule
Friday 12:00, Seminar Room I Physical Institute