The workshop “Learning Data Representation: Hierarchies and Invariance” will be held at the McGovern Institute for Brain Research at MIT (Bldg. 46) on November 22-24, 2013. The workshop will start Friday Nov. 22 afternoon and run until Sunday Nov. 24 at lunch. A reception dinner has been planned for Friday night.

All workshop activities will take place at McGovern Institute for Brain Research, MIT Bldg 46-3002 (Singleton Auditorium), 43 Vassar Street, Cambridge, MA 02139 (directions).

Workshop Agenda

There will be four sessions, each one with a set of talks and a panel discussion.



Session 1: Early Features in Vision

Session 2: Learning Features and Representations

Session 3: Learning Invariances and Hierarchies

Session 4: Beyond Feedforward Architectures

Schedule

pdf

Videos/slides







Friday November 22, 2013

: Use the links below for videos of the talks/panels ( MIT TechTV ) and presentation slides (kindly made available by all participants).

MIT Bldg 46-3002, Singleton Auditorium

14:30 Introduction to the workshop | T. Poggio ( video ), L. Rosasco ( video ), ( slides

Session: Early Features in Vision



14:50 Filters and other potions for early vision and recognition | P. Perona (

15:30 Transformations in early vision from a symmetry argument | C. Stevens (

16:10 Break

16:40 Hierarchical models of the visual cortex | T. Serre (video*) (

17:20 Panel discussion (

14:50| P. Perona ( video ) ( slides 15:30| C. Stevens ( video ) ( slides 16:10 Break16:40| T. Serre (video*) ( slides 17:20 Panel discussion ( video ) | C. Tomasi [moderator] ( video ) ( slides ), W. Freeman ( video ) ( slides ), G. Sandini ( video ) ( slides ) + Speakers

8:00 Reception at "Catalyst Restaurant"

* video is not available due to technical issues with the recording







Saturday November 23, 2013

| Chair: C. Tomasi

MIT Bldg 46-3002, Singleton Auditorium

8:30 am Breakfast

Session: Learning Features and Representations



9:30 Latent structure beyond sparse codes | B. Recht (

10:10 Dictionary learning: 3+ snippets on learning features | G. Sapiro (

10:50 Break

11:20 Panel discussion ( [ moderator], L. Rosasco (



12:20 Lunch 9:30| B. Recht ( video ) ( slides 10:10| G. Sapiro ( video ) ( slides 10:50 Break11:20 Panel discussion ( video ) | A.Verrimoderator], L. Rosasco ( video ) ( slides ), M. Belkin ( video ) ( slides ), J. Buhmann ( video ) ( slides ) + Speakers12:20 Lunch

Session: Learning Invariances and Hierarchies



14:00 Learning invariant feature hierarchies | Y. LeCun (

14:50 Scattering bricks to build invariants for perception | S. Mallat (video

15:40 Break

16:10 M-theory | T. Poggio (

17:00 Panel discussion (



18:30 Dinner (up to individual participants) 14:00| Y. LeCun ( video ) ( slides 14:50| S. Mallat (video pt.1 pt.2 ) ( slides 15:40 Break16:10| T. Poggio ( video ) ( slides 17:00 Panel discussion ( video ) | J. DiCarlo [moderator] ( video ) ( slides ), P. Baldi ( video ) ( slides ), M. Riesenhuber + Speakers18:30 Dinner (up to individual participants)

Sunday November 24, 2013

| Chair: A. Verri| Chair: J. DiCarlo

MIT Bldg 46-3002, Singleton Auditorium

8:30 am Breakfast

Session: Beyond Feedforward Architectures | Chair: J. Tenenbaum

9:00 The role of mobility and control in the inference of representations | S. Soatto (video) (slides)

9:40 Atoms of recognition | S. Ullman (video) (slides)

10:20 Break

10:50 Compositional models: complexity of representation and inference | A. Yuille (video) (slides)

11:30 Panel discussion (video) | J. Tenenbaum [moderator] (video) (slides), R. Salakhutdinov (video) (slides) + Speakers

12:30 Final remarks | T. Poggio (video) (slides)