Lifelong Learning
Why do learning systems become less capable of changing as they accumulate experience? I study the mechanisms behind forgetting, interference, and loss of plasticity, and how these phenomena shape learning over long trajectories.
PhD Researcher · Computer Science · University of Vermont
Lute Lillo
My research focuses on how learning systems can remain adaptable under continual change. In particular, I study: (1) the mechanisms behind forgetting and loss of plasticity as learning systems accumulate experience; (2) how to design systems that preserve future learnability—their capacity to acquire new knowledge and behaviors over long learning trajectories; and (3) continual reinforcement learning as a setting for studying agents that must learn, retain, reuse, and revise knowledge while interacting with changing worlds.
Research
What I work on
I approach continual adaptation from three connected directions: understanding why plasticity is lost, developing mechanisms that preserve it, and building agents that can exploit diversity when the world changes.
Future learnability
How can we design learning systems that remain capable of acquiring new knowledge and behaviors in the future? I am interested in the roles of optimization, representations, architecture, and learning dynamics in preserving adaptability over time.
Continual
Reinforcement Learning
Reinforcement learning provides a setting in which continual adaptation becomes unavoidable: agents act over long horizons while their experience, knowledge, and environments evolve. I study how agents can retain, reuse, revise, and expand what they know as they continue interacting with the world.
Selected work
Publications
Recent work spans continual learning, structural plasticity, and continual reinforcement learning.