Skip to main content.
Advanced search >
<< Back to previous page Print

<< Thursday, May 02, 2013 >>

Remind me

Tell a friend

Add to my Google calendar (bCal)

Download to my calendar

Bookmark and ShareShare

[Dissertation Talk] Modern Low-Complexity Capacity-Achieving Codes For Network Communication

Presentation: Departmental: BWRC | May 2 | 1-2 p.m. | 400 Cory Hall

Naveen Goela, PhD Candidate, U.C. Berkeley, EECS, Wireless Foundations Center

Electrical Engineering and Computer Sciences (EECS)

Communication over unreliable, interfering networks is one of the current challenges in engineering. For point-to-point channels, Shannon established capacity results in 1948, and it took more than forty years to find coded systems approaching the capacity limit with feasible complexity. Significant research efforts have gone into extending Shannon’s capacity results to networks with many partial successes. By contrast, the development of codes of reasonable complexity for networks (codes implemented by modern signal processing circuits) has received limited attention to date.

Using both algebraic structure and probability theory, I will present main ideas about the following low-complexity codes we’ve developed for certain classes of networks: i) A broadcast code which achieves rates on the capacity boundary of several classes of multi-user broadcast channels. The code is based on Arikan’s polarization theory of random variables and provides insight into information-theoretic concepts such as random binning, superposition coding, and Marton’s construction. Experimental evidence and simulations are provided; ii) A network code which achieves the computing capacity for a countably infinite class of simple noiseless interfering networks. The code separates a network into constituent irreducible networks over which a function alignment strategy is applied.

In my dissertation thesis, I also discuss low-complexity transmission of sources across noisy networks based on dimensionality reduction and convex optimization. Sometimes simple un-coded strategies achieve a performance which is exactly optimal, or close to optimal in the low signal-to-noise regime applicable for sensor networks.