Online Learning and Learning from Non-stationary Distributions research projects
Autonomous Learning in a Dynamic World
Developing new machine learning algorithms that adapt autonomously to change.
Project lead: Prof.Geoff Webb
Change is constant but the systems we model rarely remain static. Most machine learning research focuses on learning static models. Such models inevitably become irrelevant to current circumstances especially with change as a constant variable. Those algorithms that do adjust to changes assume that the change is uniform, whereas in practice it is likely to differ in form and rate in different data subspaces. Just like men’s purchasing behaviour changes may differ from those of a women’s or the young and the old. We are developing new techniques that can adapt autonomously and dynamically to different types and rates of change, and deliver effective autonomous capacity for continuous learning in a dynamic world.