We consider a class of non-convex problems, with application to robust regression and robust support vector machines. We propose an algorithm that computes the exact solution using a branch and bound approach in parameter space. Numerical experiments show that, in some cases, the time complexity of the algorithm is linear with respect to the number of samples, while it is exponential with respect to the number of regressors.
A branch and bound approach for a class of non-convex problems with applications to robust regression / Denaro, Francesco; Consolini, Luca; Locatelli, Marco. - ELETTRONICO. - (2017), pp. 382-387. (Intervento presentato al convegno 25th Mediterranean Conference on Control and Automation, MED 2017 tenutosi a University of Malta, Valletta Campus, mlt nel 2017) [10.1109/MED.2017.7984148].
A branch and bound approach for a class of non-convex problems with applications to robust regression
DENARO, FRANCESCO;Consolini, Luca;Locatelli, Marco
2017-01-01
Abstract
We consider a class of non-convex problems, with application to robust regression and robust support vector machines. We propose an algorithm that computes the exact solution using a branch and bound approach in parameter space. Numerical experiments show that, in some cases, the time complexity of the algorithm is linear with respect to the number of samples, while it is exponential with respect to the number of regressors.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
