BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20261005T082829EDT-1934uBwUrz@132.216.98.100
DTSTAMP:20261005T122829Z
DESCRIPTION:Title: 'Some steps towards causal representation learning'.\n\n
 Abstract:\n\nHigh-dimensional unstructured data such images or sensor data
  can often be collected cheaply in experiments\, but is challenging to use
  in a causal inference pipeline without extensive engineering and domain k
 nowledge to extract underlying latent factors. The long term goal of causa
 l representation learning is to find appropriate assumptions and methods t
 o disentangle latent variables and learn the causal mechanisms that explai
 n a system's behaviour. In this talk\, I'll present results from a series 
 of recent papers that describe how we can leverage assumptions about a sys
 tem's causal mechanisms to provably disentangle latent factors. I will als
 o talk about the limitations of a commonly used injectivity assumption\, a
 nd discuss a hierarchy of settings that relax this assumption.\n\nSpeaker
 \n\nJason Hartford is currently a postdoc at Mila with Yoshua Bengio. Prev
 iously - PhD at UBC with Kevin Leyton-Brown. His research interest is focu
 sed on using deep learning for causal inference\, and on designing deep ne
 twork architectures for permutation invariant data.\n\nMcGill Statistics S
 eminar schedule: https://mcgillstat.github.io/\n\nhttps://mcgill.zoom.us/j
 /83436686293?pwd=b0RmWmlXRXE3OWR6NlNIcWF5d0dJQT09\n\nMeeting ID: 834 3668 
 6293\n\nPasscode: 12345\n\n \n
DTSTART:20221007T193000Z
DTEND:20221007T203000Z
SUMMARY:Jason Hartford
URL:https://www.mcgill.ca/mathstat/channels/event/jason-hartford-342690
END:VEVENT
END:VCALENDAR