SMILE lab focuses on the frontier research of applied machine learning, social media analytics, human-computer interaction, and high-level image and video understanding. Our research is driven by the explosion of diverse multimedia from the Internet, personal or publicly available photos and videos. We start by treating fundamental theory from learning algorithm as the soul of machine intelligence and arm it with visual perception. What follows is a synergetic media learning system that not only actively collects massive visual information from the environment, but also processes and responds human interactively with precise analysis and possible suggestions. We approach visual problem "what is it?" but we are more interested in "what should we do?" or "what does it mean?" and hopefully tackle them by interacting with machines when data are in large-scale and parsing process is extremely complex, e.g., mining social relations among billions of images, recommending through visual cues, total understanding and early prediction of events and activities under social context.