As the research eld of mobile computing and communica-tion advances, so does the need for a distributed, ad-hocwireless network of hundreds to thousands of microsensors,which can be randomly scattered in the area of interest.In this paper, we present two energy-e±cient algorithms toperform distributed incremental learning for the training ofa Support Vector Machine (SVM) in a wireless sensor net-work, both for stationary and non-stationary sample data(concept drift). Through analytical studies and simulationexperiments, we show that the two proposed algorithms ex-hibit similar performance to the traditional centralized SVMtraining methods, while being much more e±cient in termsof energy cost.K. Flouri, B. BeferullLozano, P. Tsakalides
