Message-passing algorithms are used for probabilistic inference in a variety of applications including communication, computer vision, statistical physics, and machine learning. These algorithms are typically successful when the dependence between the random variables of interest can be represented as a sparse, tree-like graph.
Recently, there has been much interest in applying approximations of these algorithms to situations where the dependence graph of the variables is dense. Despite empirical success in some settings, we do not have a sound theoretical understanding of approximate message-passing (AMP) yet.
This project will investigate the question of “when does approximate message-passing work”. The aim is to characterize conditions for the algorithm to converge to the correct solution. The project could also investigate applications of AMP to new settings such as sparse approximation and data compression.
Candidate Profile: An undergraduate degree in Engineering, Applied Mathematics, or Statistics with a good academic record. A strong mathematical background is essential, especially in probability and inference. Some knowledge of optimization and/or information theory is desirable. Prior research experience in any of these areas would be a plus.
An EPSRC-funded Ph.D studentship is available for this project (for UK/EU students only). The studentship is a full award (fees+stipend) if the student is from the UK, and covers fees if the student is from the EU. Overseas students are not eligible and should not apply.
Applications should be made on-line via the Cambridge Graduate Admissions Office before the deadline:
http://www.admin.cam.ac.uk/students/gradadmissions/prospec/apply/Â with Dr Ramji Venkataramanan identified as the potential supervisor. Informal enquiries about this post may be addressed to Dr Venkataramanan,Â firstname.lastname@example.org
The University values diversity and is committed to equality of opportunity.