An approximation approach to network information theory (Q2809806)

The main aim of this book is advocating approximation approach to network communication problems (which are difficult) and presenting together results obtained with this approach in earlier publications. The main idea of the approach is step-wise: for any network information problem, first develop a simple deterministic model that captures the main features of the problem and features of information sources and communication channels (these models can be usually be analyzed with generalizations of the well-known Ford-Fulkerson max-flow min-cut graph theory algorithm), then use these models to obtain approximate characterizations of the original problem, but with a guarantee on the gap to optimality, i.e. develop lower/upper bounds for optimal values which are independent of problem parameters (but may depend on the problem structure). This last step requires many quite non-trivial insights in coding scheme, channel capacity, inference estimation etc. The authors demonstrate use of this method in four problem areas: information flow in unicast/multicast relay networks, interference channels, multiple descriptions source coding and joint source-channel coding; for all are discussed also many extensions and future research directions and problems; at the end of the book also other deterministic models are briefly described which can be used to obtain tighter bounds. NEWLINENEWLINEThe presentation is fluent and requires only basic knowledge of probability theory and proficiency in algebraic manipulations. Development of computer networks, especially wireless/mobile communication networks usually outplays the theory, thus theory is not yet providing many directly applicable results for practice, e.g. in this book the lossless multiple description problem (the deterministic model) is solved only for the case of three descriptions and description of the achievable rate region of the multi-level diversity problem (the `full' problem) is a system of 11 linear algebraic inequalities. But the book presents many theoretical aspects of information flow in multi-user networks, thus might be useful besides researchers in this area also to graduate and post-graduate students interested in communication and signal processing theory.