Venue: TS 127
Lecturer: Bhaskar D. Rao, Jacobs School of Engineering, UC San Diego, Nokia-Fulbright Scholar 2015-2016
Duration: four lectures, each lecture of 90 minutes duration
Time: Tuesday 10th of May – Wednesday 11th of May, 2016:
Lecture #1: Tuesday 10:15-12:00
Lecture #2: Tuesday 13:15-15:00
Lecture #3: Wednesday 09:15-11:00
Lecture #4: Wednesday 13:15-15:00
Prerequisites: The students must have at least master level, with a signal processing or applied mathematics background. Some knowledge of sparse representations is an advantage, although not necessary to follow the course.
Contents: In the recent decade, sparse representations have shown to be useful in many signal processing applications such as medical imaging, communications, among other applications. This course will provide an overview of the algorithms and available theoretical results in this emerging area.
The short course will cover the following topics:
- Compressed Sensing, Sparse representations and the Sparse Signal Recovery (SSR) problem: an introduction to the SSR problem and an analysis of its properties and potential difficulties. (one lecture)
- Matching Pursuit and recovery algorithms. The main basic matching pursuit algorithms, and regularization based algorithms will be discussed along with the performance guarantees. Recovery conditions based the dictionary coherence and the restricted isometry property will be discussed. (one lecture)
- Beyond l1: Regularization framework of SSthodsR, Majorization Minimization (MM) methods reweighted l1 and reweighted l2 (one lecture)
- Bayesian Methods: Maximum Aposteriori (MAP) techniques will be discussed followed by hierarchical Bayesian methods such as Sparse Bayesian learning. (one lecture)
- Extensions: Useful extensions such as block sparsity and the multiple measurements problem will be discussed.
- Bibliographical pointers will be given for each type of problem and research topic.
Last updated: 9.5.2016