Calibration of One-Class SVM for MV set estimation

Abstract

A general approach for anomaly detection or novelty detection consists in estimating high density regions or Minimum Volume (MV) sets. The One-Class Support Vector Machine (OCSVM) is a state-of-the-art algorithm for estimating such regions from high dimensional data. Yet it suffers from practical limitations. When applied to a limited number of samples it can lead to poor performance even when picking the best hyperparameters. Moreover the solution of OCSVM is very sensitive to the selection of hyperparameters which makes it hard to optimize in an unsupervised setting. We present a new approach to estimate MV sets using the OCSVM with a different choice of the parameter controlling the proportion of outliers. The solution function of the OCSVM is learnt on a training set and the desired probability mass is obtained by adjusting the offset on a test set to prevent overfitting. Models learnt on different train/test splits are then aggregated to reduce the variance induced by such random splits. Our approach makes it possible to tune the hyperparameters automatically and obtain nested set estimates. Experimental results show that our approach outperforms the standard OCSVM formulation while suffering less from the curse of dimensionality than kernel density estimates. Results on actual data sets are also presented.

Self-adaptive SVDD integrated with AP clustering for one-class classification

“…The parameters selection influences the performance of SVM classifiers greatly. For this reason, some researchers had tried to improve the performance of their classifiers by optimizing the method of parameters selection (Thomas et al, 2015;Xiao et al, 2013;Lian, 2012). KM_SVDD (K-Means SVDD) was proposed by Tax in (Tax, 2001), and since then it was subsequently improved (Wang and Lai, 2013;Xu et al, 2011;Wang and Tan, 2013;Xiao, 2013;Hua et al, 2009;Wu, 2013).…”

Quickest physical‐layer MGD anomaly detection for jamming attacks in centralized modulated wideband converter‐based ROC curve

“…There are many anomaly detection methods including machine‐learning methods that can play an important role in the field of computer networking security and wireless communication systems security such as robust support vector machine, one‐class support vector machine, hidden Markov model . Using statistical characteristics of the compressed sampling matrix and employ statistical measures to find weather the observations deviate significantly from the normal behavior.…”

How to Evaluate the Quality of Unsupervised Anomaly Detection Algorithms?

“…Here Φ(s) denotes either the mass-volume curve M V s of s or its excess-mass curve EM s , which are defined in the next section. While such quantities have originally been introduced to build scoring functions via Empirical Risk Minimization (ERM), the MV-curve has been used recently for the calibration of the One-Class SVM (Thomas et al, 2015). When used to attest the quality of some scoring function, the volumes induced become unknown and must be estimated, which is challenging in large dimension.…”