BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning to Predict and to Act - Exploring Structure in World Mode
 ls and Latent Spaces - Ingmar Posner\, Oxford University 
DTSTART:20240307T140000Z
DTEND:20240307T150000Z
UID:TALK212992@talks.cam.ac.uk
CONTACT:Fulvio Forni
DESCRIPTION:In robotics\, the ability to learn predictive models of system
 s and environments in an unsupervised\, data-driven way has emerged as a p
 romising research direction. World models in particular harbour the potent
 ial to serve as an interactive experience store for autonomous agents\, en
 abling direct trajectory optimisation as well as model-based learning in i
 magination. In this talk I will describe our recent work in creating robus
 t and versatile world models. In particular\, I will demonstrate that caus
 ally inspired inductive biases provide sufficient structure to achieve eff
 icient adaptation to intervened environments. I will then demonstrate that
  the structure encoded in a learnt latent space already provides a powerfu
 l and intuitive way to disentangle and manipulate task-relevant factors of
  variation. I will describe how this structure can be exploited to capture
  a purely data-driven model of complex robot platforms and demonstrate tha
 t this not only casts a novel light on affordance learning\, but also reve
 als a framework capable of learning a versatile unified representation for
  quadruped locomotion deployable in the real-world.\n\nThe seminar will be
  held in the JDB Seminar Room\, Department of Engineering\, and online (zo
 om): https://newnham.zoom.us/j/92544958528?pwd=YS9PcGRnbXBOcStBdStNb3E0SHN
 1UT09
LOCATION:JDB Seminar Room\, Department of Engineering and online (Zoom)
END:VEVENT
END:VCALENDAR