OneClassMaxMinOver (OMMO) is a simple incremental algorithm for one-class support vector classification. We propose several enhancements and heuristics for improving model select...
We study the problem of learning kernel machines transductively for structured output variables. Transductive learning can be reduced to combinatorial optimization problems over a...
In least squares support vector (LS-SVM), the key challenge lies in the selection of free parameters such as kernel parameters and tradeoff parameter. However, when a large number ...
We examine the so-called rigorous support vector machine (RSVM) approach proposed by Vapnik (1998). The formulation of RSVM is derived by explicitly implementing the structural ris...
Based on studies and experiments on the loss term of SVMs, we argue that 1-norm measurement is better than 2-norm measurement for outlier resistance. Thus, we modify the previous 2...