IPAB Seminar-13/02/2020

 

Title: Beyond Supervised Learning

 

Abstract: Supervised learning is a powerful technique to build models that can associate data to given targets. Today this is a successful method that is widely adopted in the industry. However, supervised learning comes with limitations: It relies on costly, time-consuming and error-prone manual labeling of a large set of examples. Moreover, animals do not seem to learn about objects through a teacher during their lifetime. It seems possible that much of the learning occurs in an unsupervised manner. I will illustrate some general ideas that show a path towards learning without annotation: self-supervised learning and learning through generative models. Self-supervised learning has emerged as a successful method to learn useful features without manual labeling. Such features can then be repurposed to other tasks through fine-tuning.

Another idea I will touch upon is on using generative models to learn structure in the data such as the 3D surface of objects or their 2D segmentation.

 

Short Bio: 

Paolo Favaro is full professor at the University of Bern, Switzerland, where he heads the Computer 

Vision Group. He received the Laurea degree (B.Sc.+M.Sc.) from Università di Padova, Italy in 1999, 

and the M.Sc. and Ph.D. degree in electrical engineering from Washington University in St. Louis 

in 2002 and 2003 respectively. He was a postdoctoral researcher in the computer science department 

of the University of California, Los Angeles and subsequently in Cambridge University, UK. 

Between 2004 and 2006 he worked in medical imaging at Siemens Corporate Research, Princeton, USA. 

From 2006 to 2011 he was Lecturer and then Reader at Heriot-Watt University and Honorary Fellow at 

the University of Edinburgh, UK. His research interests are in computer vision, machine learning, 

computational photography, and image processing.

Feb 13 2020 -

IPAB Seminar-13/02/2020

Paolo Favoro (University of Bern)

IF, G.03