BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Generative models for few-shot prediction tasks - Marta Garnelo (G
 oogle DeepMind)
DTSTART:20190220T134500Z
DTEND:20190220T151500Z
UID:TALK120481@talks.cam.ac.uk
CONTACT:75379
DESCRIPTION:Few-shot density estimation lies at the core of current meta-l
 earning (or ‘learning to learn’) research and is crucial for intellige
 nt systems to be able to adapt quickly to unseen tasks. In this talk we wi
 ll introduce generative query networks (GQN - published in Science last ye
 ar)\, a generative model for few-shot scene understanding that learns to c
 apture the main features of synthetic 3D scenes. In the second half of the
  talk we will cover neural processes (NPs)\, a generalisation of the GQN t
 raining regime a wider range of tasks like regression and classification. 
 NPs are inspired by the flexibility of stochastic processes such as Gaussi
 an processes\, but are structured as neural networks and trained via gradi
 ent descent. We show how NPs make accurate predictions after observing onl
 y a handful of training data points\, yet scale to complex functions and l
 arge datasets.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR