2011/09/16 by Misha Denil, Denil, Misha, Thomas Trappenberg +1
Computer Science · #Anomaly Detection Techniques and Applications #Artificial Intelligence (cs.AI) #Digital Media Forensic Detection #FOS: Computer and information sciences #Imbalanced Data Classification Techniques #Machine Learning and Data Classification
paper · pdf · doi:10.48550/arxiv.1109.3532
openalex publication_date 2011/09/16 · openalex created_date 2025/10/24 · openalex updated_date 2026/07/28
In this paper we demonstrate that two common problems in Machine\nLearning---imbalanced and overlapping data distributions---do not have\nindependent effects on the performance of SVM classifiers. This result is\nnotable since it shows that a model of either of these factors must account for\nthe presence of the other. Our study of the relationship between these problems\nhas lead to the discovery of a previously unreported form of "covert"\noverfitting which is resilient to commonly used empirical regularization\ntechniques. We demonstrate the existance of this covert phenomenon through\nseveral methods based around the parametric regularization of trained SVMs. Our\nfindings in this area suggest a possible approach to quantifying overlap in\nreal world data sets.\n