Compressive sensing (CS) is new method for sensing and signal recovery from far-fewer measurements than what is suggested by traditional sampling theories. This has immense benefits in applications where we seek to sense at faster speeds, higher resolution, lower power consumption and reduced sensor costs. At the heart of CS is the idea that we can solve under-determined linear systems using sparse signal representations and specialized convex optimization techniques.
In this course, we will develop a formal study of techniques required to solve under-determined linear systems. We will discuss the theory and practice of sparse optimization with a two-fold objective: building novel imaging architectures and solving computer vision problems in classification, structure-from-motion, and face recognition. This will involve a rich interplay of ideas from linear algebra, probability, and convex optimization. Finally, we will explore broader applications in computational imaging, neural signal processing, and sparse regression that benefit from the theory and optimization methods developed in the course.
Pre-requisites: 36-217 or Undergraduate linear algebra, probability. An introduction to convex optimization would be useful but not necessary as the necessary material will be taught as part of the course.