Networked Information Processing: New Compression, Processing and Control Paradigms for Networks
Resource (energy, bandwidth, delay, storage) constraints are the key factors in designing the current and emerging communications, networking and signal processing applications. Such applications strongly motivate foregoing the convenience of digital communication and revisiting the potential benefits, and even necessity of analog communication, albeit in the modern context of source-channel coding. In the first part of the talk, I will present the major challenges and my research efforts in analog networking, including the derivation of a theoretical foundation from estimation and information theoretic principles, optimal algorithm design based on principles derived from statistical physics, i.e., deterministic annealing. Analog networking problem is intrinsically connected to the interplay between communications and control, namely stochastic networked control. The theoretical results obtained from analog networking shed light on the open stochastic control problems, such as the jamming problem and the numerical approaches are directly applicable control problems as well. As examples, I will present the saddle point solution to the jamming problem in general (non-Gaussian) settings and a numerical solution to a well known open problem in distributed control, namely Witsenhausen's counterexample.
In the second part of the talk, I will outline challenges in the practical problem of minimum cost source coding and dispersive information routing (DIR) over a resource constrained network. This problem opens up a rich set of research problems, such as the derivation of achievable regions of multiterminal source coding under DIR paradigm, that are strongly connected to well known open problems in information theory including the multiple descriptions coding and the common information of correlated random variables. I will first outline our contributions within the DIR setting and then present two new coding schemes for L-channel multiple descriptions coding, both can be viewed as extensions of the seminal work by Zhang and Berger to multichannel. Obtained regions strictly subsume the well known counterparts due to Venkataramani, Kramer and Goyal and Pradhan, Puri and Ramchandran. Finally, I will present the extensions of the different notions of common information (due to Wyner and Gacs-Korner) to lossy settings with new connections between Gray-Wyner network and distributed storage and databases.
Joint work with Kenneth Rose, Kumar Viswanatha and in part with Tamer Basar.
Bio: Emrah Akyol received the B.Sc. degree in 2003 from Bilkent University, the M.Sc. degree in 2005 from Koc University (both in Turkey), and the Ph.D. degree in 2011 in electrical and computer engineering from UC Santa Barbara. From 2005 to 2007, he held positions at Hewlett-Packard Laboratories and NTT Docomo Laboratories (both in Palo Alto, CA) and UCLA, where he worked on several topics in multimedia compression and networking.
Currently, Dr. Akyol is a postdoctoral researcher at UC Santa Barbara. His current research focuses on networked source coding, joint source-channel coding, relations between estimation theory, information theory and statistical physics and low-delay communications with applications to optimization and control of smart grids.