Timo Koski: Learning of Multicausal Interaction Networks

Timo Koski (Dept. of Mathematics, Royal Institute of Technology, Stockholm) will give a seminar in the lunch area, 8th floor Niels Henrik Abels hus at 14:15.

Title: Learning of Multicausal Interaction Networks

Abstract: We develop a model of multicausal disjunctive interaction, which includes the Noisy OR model as a special case. The model is a probabilistic representation of the pie model of disease. We propose a Bayesian network for multicausal disjunctive interaction, where the interactions are represented by a layer of mixture distributions. An alternating algorithm for learning the structure from data is introduced.

Published Apr. 11, 2016 3:04 PM - Last modified Apr. 11, 2016 3:04 PM