Abstract:

In comparison with the dual formulation of ε-support vector machine, the dual of v-support vector regression (v-SVR) has an extra inequality constraint. To date, there is no effective and feasible v- solution path for v-SVR. To solve the infeasible updating path problem of the v solution path for v-SVR, which was first proposed by Loosli et al, an improved v- solution path for v-SVR is proposed. Based on the modified formulation of v-SVR and the Karush-Kuhn-Tucker (KKT) conditions, the strategy of using a new introduced variable and an extra term can avoid the conflicts and exceptions effectively during the adiabatic incremental adjustments. Finally, the proposed algorithm can fit the entire v-solution path within the finite number of iterations. Theoretical analysisand simulation results demonstrate that the proposed algorithm is effective and feasible.