Learning to Control - an Agenda
Carl Edward Rasmussen, Cambridge
Abstract:
Learning from experience is an incredibly powerful paradigm essential to the survival of many biological systems, but surprisingly under utilized in artificial control systems. In this talk, I'll discuss how to proceed: I'll argue, that the necessary ingredients are: probabilistic inference and non-parametric models. I'll discuss some real and simulated proof of concept examples, where learning to control is achieved with unprecedented speed from essentially no prior knowledge.