BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:On Learning Latent Models with Multi-Instance Weak Supervision - E
 fi Tsamoura (Samsung AI\, Cambridge)
DTSTART:20231017T120000Z
DTEND:20231017T130000Z
UID:TALK206044@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:We consider a weakly supervised learning scenario where the su
 pervision signal is generated by a transition function σ of labels associ
 ated with multiple input instances. We formulate this problem as multi-ins
 tance Partial Label Learning (multi-instance PLL)\, which is an extension 
 to the standard PLL problem. Our problem is met in different fields\, incl
 uding latent structural learning and neuro-symbolic integration. Despite t
 he existence of many learning techniques\, limited theoretical analysis ha
 s been dedicated to this problem. We provide the first theoretical study o
 f multi-instance PLL with possibly an unknown transition σ. We make minim
 al assumptions on the data distributions. In fact\, we prove learnability 
 even under the “toughest" distributions that concentrate their mass on a
  single instance. In addition\, we provide learning guarantees under widel
 y used surrogate losses for training classifiers subject to logical theori
 es. We are the first to provide this theoretical analysis\, closing a gap 
 in the neuro-symbolic and latent structural learning literature. This work
  will be presented in NeurIPS 2023: "https://arxiv.org/pdf/2306.13796.pdf.
 ":https://arxiv.org/pdf/2306.13796.pdf\n \n*Bio:* Efi Tsamoura is a Senior
  Researcher at Samsung AI\, Cambridge\, UK. In 2016\, she was awarded an e
 arly career fellowship from the Alan Turing Institute\, UK\, and before th
 at\, she was a Postdoctoral Researcher in the Department of Computer Scien
 ce of the University of Oxford. Her main research interests lie in the are
 as of logic\, knowledge representation and reasoning\, and neuro-symbolic 
 integration. Her research has been published in top-tier AI and database v
 enues (SIGMOD\, VLDB\, PODS\, AAAI\, ICML\, NeurIPS\, etc.). Efi started t
 he Samsung AI neuro-symbolic workshop series “When deep learning meets l
 ogic” and has been a keynote in the 2023 Extended Semantic Web Conferenc
 e. \n\n"You can also join us on Zoom":https://cam-ac-uk.zoom.us/j/92041617
 729\n\n\n
LOCATION:Lecture Theatre 2\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR