- Patrick Perez (Technicolor, Rennes – France) – On visual comparison
- Vladimir Vapnik (Facebook Research – USA) – Learning with Intelligent Teacher : Similarity Control and Knowledge Transfer
Deciding if two pieces of visual content are somehow related, or to which degree they are related, is an ubiquitous problem when searching, processing and analyzing images. Such visual comparisons can be conducted among small fragments (e.g., point matching for tracking or 3D reconstruction, patch-based image processing, mid-level feature mining), object-level fragments (e.g., face verification or face clustering) or whole images (copy detection, image retrieval, picture linking). To this end, visual content is usually turned into a fixed size, high-dimensional vector representation and a suitable similarity measure is defined between such vectors. Focusing on large scale example-based search and on face verification, we shall discuss how parts of this “description-comparison” pipeline can be learned, with or without supervision, in order to speed up comparisons or to make them more meaningful.
In the talk, I will introduce a model of learning with Intelligent Teacher. In this model, Intelligent Teacher supplies (some) training examples with additional (privileged) information. Privileged information is available only for training examples and not available for test examples. Using privileged information it is required to find a better training processes (that use less examples or more accurate with the same number of examples) than the classical ones. In this lecture, I will present two additional mechanisms that exist in learning with Intelligent Teacher :
– The mechanism to control Student’s concept of examples similarity,
– The mechanism to transfer knowledge that can be obtained in space of privileged information to the desired space of decision rules.
Privileged information exists for many inference problem and Student-Teacher interaction can be considered as the basic element of intelligent behavior