[Lisa_teatalk] Practice talk, Combining Modality Specific Deep Nets for Emotion Recognition in Video