An observation which makes sense in our day to day activities is that filtering or processing of data can only reduce the amount of information available. A nice formal analogue of this statement in information theory is the data processing inequality for mutual information on Markov chains. More general dependency structures among random variables arise in a variety of fields such as digital communications, inference, and machine learning. This work aims at identifying suitable functionals for which data processing inequality analogues hold over a wider range of dependency structures among random variables. This work also aims at exploring interesting consequences of such data processing inequality analogues

I worked on probabilistic graphical models in the context of seismic image analysis with Professor Alan Willsky’s group over IAP and Spring 2013. Over IAP and Spring 2014, I worked on problems in information theory with Professor Polyanskiy.