Anant Sahai's areas of interest are machine learning, communications, control, and wireless spectrum. Within machine learning, he is particularly interested in how to get artificial agents to cooperate as well as general issues on the fundamental limits of learning, and how assumptions interact with inference procedures. In particular, he is very interested in the foundations of overparameterized systems. He leads the "Data and Machine Learning" Working Group for SpectrumX, a new multi-university collaboration in the area of wireless spectrum.
On the communications side, his interests are particularly in the areas of wireless and information theory. Within information theory, one of his current areas of interest is in developing conceptual tools needed to understand the fundamentals of "quality of service"; beyond classical Shannon data rate. To that end, he is interested in distributed control systems as they provide well formed mathematical models that do not mesh with the classical notions from information theory. In addition, distributed control is a setting in which we can explore the role of implicit communication/signalling. He is also interested in how multi-scale heterogenous wireless systems can coexist peacefully within the context of cognitive radios. This is the general problem of spectrum sharing. Besides being an interesting case of distributed control, this brings legal and economic concerns together with machine learning, big data, and communications in a non-trivial way.