Past increases in information consumption have been fueled mainly by increasing the bit-carrying capacity of our telecommunications infrastructure, including capital improvements such as laying fiber as well as technological advances that enable this infrastructure to carry more bits. This approach of simply moving more bits now appears to be facing diminishing returns, however. In contrast, there has been less attention paid to how data compression can reduce the number of bits that must be communicated. Thus future increases in per-capita information consumption could be driven by improvements in compression. This project will also contribute to improved STEM education through the development of adoptable course materials for a dedicated graduate-level course on information theory and data compression.
This research examines how to harness gains in compression, especially network gains. The work will focus on lossy compression for two simple networks that are not well understood: (1) the problem in which several encoders observe separate sources and send compressed versions to a centralized decoder, who seeks to reproduce the sources, and (2) the problem in which a single encoder broadcasts a compressed version of a source to multiple decoders, who reproduce the source with the benefit of side information. For both problems the PIs seek to determine the information-theoretic rate-distortion tradeoffs for a practically-important class of instances. For the first problem, the PIs will also measure the network gains that can be obtained in practice by constructing a system to compress synchrophasor data obtained from the power grid.
