1. •Stochastic modelling of mobile wireless networks
   Modern robotic and sensor networks often exhibit wireless communication behaviour heavily depend on its mobility model. We derive geometric statistics regarding the node mobility patterns to describe the consequences and hence investigate the control principles for these wireless networks.
   
   This is a collaboration with University of Massachusetts.
   

1. •nsSNPs and Systems Biology
   How does a single mutation make a difference? Does it render the protein structure invalid or does it affect protein-protein interactions or alter systemic behaviours?
   

Bongo (PLoS CB 2008) is our first argument towards this problem in which we demonstrate structure networks alone can explain a big chunk of disease related mutations.

This is a collaboration with Dept of Biochemistry (Blundell group) and Dept. Chemistry.

1. •Social Networks
   Social networks and its dynamics are gaining lots of attention these days. My interest here is to understand how networks are formed as it is now so that one can understand about future. One recent collaboration with Oxford describes our ideas about how variance can make a difference in the network scale.
   Size matters: personal network size variations and network performance (SocialCom 2009)
   
   This is a collaboration with Institute of Cognitive and Evolutionary Anthropology, University of Oxford.
   
   

2. •High dimensional sensory fusion
   Many high level concepts are hard to decide for one single signal, e.g., a rabbit jumps over a green lawn. Many signal sources are usually required and their dynamics considered to be sufficient for these high level concepts. I am working on near-real-time, robust techniques for finding these complex time series patterns.
   

See Beta Random Projection (CIKM 2008)

1. •Statistical Bio-marker for complex diseases
   
   

2. •Very Large, Distributed Databases
   

This is about how one can actually carry out the queries above in a distributed, large database system. First is the efficient algorithms and network structures for decentralised queries in large systems. The result is a system called Qube.

Concurrent database evaluations due to complex queries often end up in disasters. Either you are slow or something undeterministic happens. I work on the semantics of distributed Datalog programs (aka SQL) such that a safe (i.e. deterministic) concurrent execution plan is followed. This is done within the P2 project.

This is a collaboration with Intel Research Berkeley (Maniatis).