The problem of automatically tuning multiple parameters for pattern recogni
tion Support Vector Machines (SVMs) is considered. This is done by minimizi
ng some estimates of the generalization error of SVMs using a gradient desc
ent algorithm over the set of parameters. Usual methods for choosing parame
ters, based on exhaustive search become intractable as soon as the number o
f parameters exceeds two. Some experimental results assess the feasibility
of our approach for a large number of parameters (more than 100) and demons
trate an improvement of generalization performance.