Poster ID: T62

Hidden Common Cause Relations in Relational Learning
silva@statslab.cam.ac.uk

Ricardo Silva, Wei Chu, Zoubin Ghahramani Gatsby Computational Neuroscience Unit, Columbia University, University of Cambridge

X1

Y1

· An entirely new family of relational models · Relations express hidden common causes · Information propagates through chains of training points · Classifiers with state-of-the-art performance · Easy-to-implement Gaussian process model
Y1 Y2 Y3 Y4

X2

Y2

Y1 X3 Y3

Y2

Y3

Y4

How is the hidden common cause model different from a Markov network? Check the poster.