EE 520
Sparse and Low-rank Statistical Signal Processing
Spring 2012
M W 2:10-3:30, 1219 Coover
- Instructor: Aleksandar Dogandzic
- Room: 3119 Coover
- Office hours: M W 1-2
- Email:
- Reference book:
-
C. M. Bishop,
Pattern Recognition and Machine Learning,
Springer,
2006.
- Course Outline:
- Introduction;
- Sampling signals with finite rate of innovation;
- Basics of estimation theory, expectation-maximization (EM)
algorithm;
- Applications of EM algorithm:
- sparse Bayesian learning,
- hard and soft thresholding for sparse
signal reconstruction and matrix completion,
- reweighted basis pursuit schemes,
- overrelaxation schemes for accelerating EM-type algorithms and
their application to sparse signal reconstruction,
- reconstruction from quantized noisy measurements;
- Relevance vector machines and Bayesian compressed sensing;
- Model selection and selecting regularization parameters;
- An introduction to graphical models, inference on graphs, belief
propagation;
- Applications of graphical models:
Approximate message passing (AMP) and
model-based compressed sensing;
- Topics of students' interests, e.g.,
- robust principal component
analysis,
- applications: X-ray computed tomography, magnetic
resonance imaging, sensor networks.
- Handouts