“…They estimate a decision function that extracts the 'normal' or 'regular' data from of the training sample by maximising the distance of the 'regular' data to the origin. In recent years extensions to this method have been proposed (see for example Bicego & Figueiredo, 2009;Li, 2011;Sjörstrand, Hansen, Larsson, & Larsen, 2007;Tran, Li, & Duan, 2005), and they have been used in various fields (for example audio surveillance Rabaoui, Davy, Rossignol, & Ellouze, 2008; or portfolio selection Gotoh and Takeda, 2012).…”