BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Microsoft AI &amp\; Pizza - Speaker to be confirmed
DTSTART:20230705T163000Z
DTEND:20230705T180000Z
UID:TALK203008@talks.cam.ac.uk
CONTACT:Kimberly Cole
DESCRIPTION:Dear Cambridge AI & Machine learning enthusiast\, \n \nWe are 
 looking to host our second AI & Pizza event at 5:30 pm on 5th July 2023\, 
 where you can get your slice of the latest AI and machine learning researc
 h in Cambridge!\n \nJoin us at the auditorium\, 21 Station Rd for an engag
 ing evening featuring two 15-minute talks on cutting-edge research in AI a
 nd ML from both academia and industry (speaker details below). After the t
 alks\, we will be providing free pizza and refreshments following the talk
 s. \n \nStay tuned for more information\, and we look forward to seeing yo
 u at the event!\n\nP.S. We also welcome volunteer speakers from various ba
 ckgrounds. Contact (chaoma@microsoft.com) if you are interested in giving 
 a talk for our future events! \n \nCheers\,\nChao\n \nLocation: The Audito
 rium\, 21 Station Rd\n \nTime: 17:30 – 18:00 (talks)\, 18:00 – 19:00 (
 pizza).\n \n \nSpeakers\n \n17:30 – 17:45: Austin Tripp\, University of 
 Cambridge\nTitle: Synthesizing molecules with Machine Learning\n Abstract:
  Synthesizing novel molecules is a key task in chemistry and drug discover
 y\, occupying a significant portion of medicinal chemists' time. In this t
 alk I present a summary of recent efforts to automate this process with ma
 chine learning\, with a particular focus on reinforcement learning. I also
  highlight some issues with benchmark practices in this area which we have
  tried to address by releasing an open source library called "syntheseus".
 \n \n \n17:45 – 18:00: Divyat Mahajan\, MILA & Université de Montréal/
 Microsoft Research Cambridge\n \nTitle: Interventional Causal Representati
 on learning\nAbstract: Causal representation learning seeks to extract hig
 h-level latent factors from low-level sensory data. Most existing methods 
 rely on observational data and structural assumptions (e.g.\, conditional 
 independence) to identify the latent factors. However\, interventional dat
 a is prevalent across applications. Can interventional data facilitate cau
 sal representation learning? We explore this question in this paper. The k
 ey observation is that interventional data often carries geometric signatu
 res of the latent factors' support (i.e. what values each latent can possi
 bly take). For example\, when the latent factors are causally connected\, 
 interventions can break the dependency between the intervened latents' sup
 port and their ancestors'. Leveraging this fact\, we prove that the latent
  causal factors can be identified up to permutation and scaling given data
  from perfect do interventions. Moreover\, we can achieve block affine ide
 ntification\, namely the estimated latent factors are only entangled with 
 a few other latents if we have access to data from imperfect interventions
 . These results highlight the unique power of interventional data in causa
 l representation learning\; they can enable provable identification of lat
 ent factors without any assumptions about their distributions or dependenc
 y structure\n
LOCATION:The small lecture theatre\, 21 Station Rd
END:VEVENT
END:VCALENDAR