University of Cambridge > > Machine Learning @ CUED > Inference of a partially observed kinetic Ising model

Inference of a partially observed kinetic Ising model

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Zoubin Ghahramani.

In this talk we will describe some results on inferring the interactions between variables in a point process, the kinetic Ising model, in which only parts of the spins are visible. We focus on two cases. First, one in which the hidden spins do not interact with each other, but there are directed interactions to and from them to observed spins, as well as between the observed spins. In this case we develop an algorithm based on a combination of belief propagation and replica approximation for inferring the state of hidden variables as well as connections between all the spins in the system from observing the spin history of the set of observed spins. Second, we consider a fully connected kinetic Ising model and study mean-field and TAP approximations for inference and learning.

The talk is based on the two following papers

1. Battistin, C, et al. “Belief propagation and replicas for inference and learning in a kinetic Ising model with hidden spins.” J Stat Mech 2015.5 (2015): P05021 .

2. Dunn, B, and Roud, Y. “Learning and inference in a nonequilibrium Ising model with hidden nodes.” Phys Rev E 87 .2 (2013): 022127.

This talk is part of the Machine Learning @ CUED series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity