Speaker
Jarek Duda
(Jagiellonian University)
Description
While machine learning is usually focused on prediction of values, on various applications I will introduce to simple family of methods to work with learned probability distributions - e.g. model joint, predict conditional, their time evolution. One proposed application direction will be multi-feature Granger causality, enhancing the standard method with evaluation of propagation speed, and automatic extraction of multiple types of dependencies for example in EEG data. There will be also proposed its application to build artificial neurons updating model of joint distribution of its connections.
Author
Jarek Duda
(Jagiellonian University)
