17 Active Learning with Kernel Machines for Tornado Detection
-
Published:2008
Download citation file:
In this paper, active learning with kernel machines, including Support Vector Machines and the p-Center Machine, is applied to the problem of tornado detection. This method is used to discriminate which storm-scale circulations yield tornadoes based on the radar derived Mesocyclone Detection Algorithm (MDA) and near-storm environment (NSE) attributes. The main objective of active learning is to choose instances or data points that are important to be labeled and included in the training set since labeling the instances in tornado data is considered costly and time consuming. We compare active learning with passive learning where the next instances to be included to the training set are randomly selected. The preliminary results show that active learning can achieve good performance and significantly reduce the training set size.