Deep learning is becoming popular in recent years, due to the huge improvement in the computation power and cost of GPU. Deep learning is widely applied in various areas such as networking, audio and graphics. It also attracted a lot of attention in the gaming field. Serious game is highly data intensive in nature as it requires human computer interaction which generates a lot of user input/output and bio-feedback. Classification and recognition technology also take a very important role in serious game. In this presentation, I will talk about the relationship between deep learning and serious game in the existing state of art, and how deep learning can be applied to improve serious game design.