Ph.D. student starting either in
Spring or Fall 2008
I have an opening for a Ph.D.student in my group starting either
Spring
or Fall 2008. Please email me if you have a good
background in
- undergraduate probability
(at the level of first 4 chapters of Bertsekas &
Tsitsiklis),
- signal processing (at the
level of first 4-5 chapters of Oppenheim
&Wilsky's book), and
- linear algebra.
- knowledge of detection and estimation theory and of stochastic
processes if you have an M.S. degree
It's good if you've done some image or signal processing research
projects or written papers, but that is not critical at all.
Two possible research topics
are (there is considerable freedom to choose other topics
also):
1. Change detection and changed
parameter estimation in noise-corrupted linear or nonlinear
systems.
Such systems first need to be tracked using Kalman or particle filters.
I foresee that the work will involve a fair mix of theoretical analysis
and simulations and possibly
some applications to biomedical imaging
problems. Initial ideas described in:
N. Vaswani, Additive Change Detection in Nonlinear
Systems with Unknown Change Parameters, IEEE
Trans. Signal Processing, March 2007.
2. Large Dimensional
Sequential Bayesian filtering (a.k.a. tracking). My preliminary
work in this area has focussed on developing particle filtering
algorithms for large dimensional state spaces. This itself is in its
preliminary stages - a lot more needs to be done to analyze their
performance and to make them practically implementable. Also, similar
ideas can be developed for large dimensional Kalman filtering and large
dimensional importance sampling. The essential idea of the approach is
that in most practical situations, "most of the state change" occurs in
a small number of dimensions, whereas the change in the rest of the
dimensions is "small". Initial ideas described in:
Namrata Vaswani, PF-EIS
and PF-MT:
New
Particle Filtering
Algorithms for Multimodal Observation Likelihoods and Large Dimensional
State Spaces, in IEEE Intl.
Conf. on
Acoustics, Speech and Signal Processing
(ICASSP), 2007
Some possible applications are in deformable contour tracking,
in tracking spatially varying temperature/pressure (or other physical
quantities) change over time, or in tracking spatially varying
illumination change of moving objects. Details in this talk
Back to Namrata
Vaswani's webpage