Scenario optimization: new schemes for the non-convex case