On Networked Source Coding in the Presence of Exploration versus Exploitation Tradeoff
Motivated by the problems in sensor networking, primarily the well-known exploration versus exploitation tradeoff in data gathering, we propose a successive source coding scheme based solely on binning. We show that proposed scheme provides significant storage reduction compared to the conventional conditional codebook encoding scheme, without any rate loss. In particular, the codebook storage requirement of the conventional approach grows exponentially in the number of gathered sources and successive encoding stages while that of the proposed is linear in both quantities.
The problem at hand, when extended to distributed settings, includes several problems as special cases. When specialized to the scalable coding with decoder side information, our approach provides a unified encoding scheme for the two important cases concerning the degraded side information where source $X $ and the side information variables ($Y_1, Y_2$) form a Markov chain in the order of either $X-Y_1-Y_2$ or $X-Y_2-Y_1$. Our coding scheme essentially performs scalable and Wyner-Ziv coding simultaneously by binning. This unified coding scheme achieves the rate-distortion regions of prior results, while for the $X-Y_1-Y_2$ case, it suggests possible improvement.
Finally, motivated by the proposed encoding scheme, and the Gray-Wyner network model, we consider a new interpretation of scalable coding. We show the potential benefits of networked scalable coding and derive an achievable region. We discuss its connections with the lossy common information, the subset typicality lemma and the rate transfer arguments.
Emrah Akyol received the Ph.D. degree in 2011 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 USC. His current research is on the interplay of networked information theory, communications and control.