We propose an approach to various sensor network applications via learning theory. As a first step, we put forward a well-known algorithm from learning theory, applying it to environmental monitoring, tracking moving objects and plumes, and localization. We give a brief overview of some of the basic concepts of learning theory and present some preliminary simulation results.