Obtaining scalable and accurate classification in large-scale spatio-temporal domains

Abstract

We present an approach for learning models that obtain accurate classification of data objects, collected in large-scale spatio-temporal domains. The model generation is structured in three phases: spatial dimension reduction, spatio-temporal features extraction, and feature selection. Novel techniques for the first two phases are presented, with two alternatives for the middle phase. We explore model generation based on the combinations of techniques from each phase. We apply the introduced methodology to data-sets from the Voltage-Sensitive Dye Imaging (VSDI) domain, where the resulting classification models successfully decode neuronal population responses in the visual cortex of behaving animals. VSDI is currently the best technique enabling simultaneous high spatial (10,000 points) and temporal (10 ms or less) resolution imaging from neuronal population in the cortex. We demonstrate that not only our approach is scalable enough to handle computationally challenging data, but it also contributes to the neuroimaging field of study with its decoding abilities. The effectiveness of our methodology is further explored on a data-set from the hurricanes domain, and a promising direction, based on the preliminary results of hurricane severity classification, is revealed.