Shinichi Goto, Terumasa Aoki: TUDCL at MediaEval 2013 Violent Scenes Detection: Training with Multi-modal Features by MKL. MediaEval 2013