Đorđe Božić

I am a reinforcement learning researcher pursuing a PhD at the Bath Reinforcement Learning Laboratory (BRLL) under the supervision of professor Özgür Şimşek.

My main research focus is symbol grounding and knowledge generalisation through the use of language. Learning to properly associate language to behaviours and environment observations would allow artificial agents to learn from human experiences and inherit commonsense prior beliefs about the world. I am currently looking into learning multimodal representations able to generate both behaviours and their natural language descriptions.

Speaking more broadly, I am interested in areas bridging the gap between humans and artificial agents. The associated research quesitons are how to build behaviour hierarchies, how to equip agents with intrinsic motivation, and how to make agents continually learn without forgetting past knowledge. Common to all my research pursuits is the same underlying motivation to build agents with the capacity to generalise to to new tasks and transfer knowledge to unfamiliar environments.

Beyond research, I am currently a teaching assistant on several undergraduate and postgraduate courses at the University of Bath, including reinforcement learning and Bayesian machine learning. Since I like teaching, I am additionally involved in the organisation of the PSIML machine learning summer school.

Prior to my postgraduate academic pursuits, I was a machine learning engineer for five years. I worked on computer vision problems including classification, object detection and tracking, action detection, outlier detection, pose estimation, and expert systems in the retail industry.