ASSET Seminar: “Lifelong Learning for Autonomous Systems: Progress and Challenges” (Eric Eaton, Penn)

ABSTRACT: 

Research in lifelong or continual machine learning has advanced rapidly over the past few years, primarily focusing on enabling learned models to acquire new tasks over time while avoiding catastrophic forgetting of previous tasks. However, autonomous systems still lack the ability to rapidly learn new generalizable skills by building upon and continually refining their learned knowledge—a hallmark of lifelong learning in humans and animals. In this talk, I will examine our progress toward lifelong machine learning and its application to autonomous systems, exploring recent work in compositional representations and self-supervision. We’ll then discuss how far we have come and the open problems that remain toward learning generalizable skills in lifelong settings.