Outlier Detection via Deep Learning Architecture
Irina Kakanakova, Stefan Stoyanov · 2017
An important issue in processing data from sensors is outlier detection. Plenty of methods for solving this task exist - applying rules, Support Vector Machines, Naive Bayes. They are not computationally intensive and give good results where border between outliers and inliers is linear. However, when the border's shape is highly non-linear, more sophisticated methods should be applied, with the requirement of not being computationally intensive. Deep learning architecture is applied to solve this problem and results are compared with the ones obtained by applying shallow architectures.