Multimedia Meets Deep Reinforcement Learning

Shu-Ching Chen

Abstract: Multimedia data analysis methods based on artificial intelligence and machine learning have achieved great success in the past decades. However, it remains challenging to generate automated decision-making based on the multimedia data directly. Recent advances in deep reinforcement learning (DRL) have made it a practical framework for solving various sequential decision-making problems. However, it lacks the capability to be effectively used in real-world applications. Multimedia data are useful sources of information to support reliable decision-making. By incorporating DRL with multimedia content analysis, it is promising to develop reliable and effective decision-making systems and frameworks for critical applications, such as disaster management, autonomous driving, healthcare, and so on. Hence, it creates the opportunities for the multimedia community to develop novel techniques to address the existing challenges to further improve the usability and performance of automated decision-making based on multimedia data.