Limitations of Deep Learning Workshop

Limitations of Deep Learning Workshop in Sestri Levante, Italy, June 25-27, 2019

Sestri Levante (photo: Jiuguang Wang)

The Center for Brains, Minds and Machines and The Hebrew University of Jerusalem, with generous support from the Haar family, are organizing a workshop on advances in and limitations of deep learning on June 25-27, 2019 in Sestri Levante, Italy. For three days we will bring together computer scientists, cognitive scientists, and neuroscientists to share and discuss recent advances in deep learning, with particular attention to its current limitations and how they might be overcome to develop intelligent systems and models of the human mind.

Organizers: Josh McDermott (MIT), Yair Weiss (Hebrew University)

Location: Fondazione Mediaterraneo, 14 Via Portobello, Sestri Levante, GE 16039

CBMM Mission:

The mission of CBMM is to make progress on the greatest problem in science — human intelligence. A new field is emerging dedicated to developing a computationally centered understanding of human intelligence and to establishing an engineering practice based on that understanding. CBMM’s long-term goals are to uncover basic principles of intelligence, both natural and artificial, and the mechanisms that can be used to implement intelligent systems, both by brains and in silicon. For the past several years, the Center for Brains, Minds, and Machines has been pursuing this goal and is now bringing the discussion to the European research community.

 As robots develop further and our lives are increasingly intertwined with machines and algorithms, understanding intelligence is becoming a crucial problem. Without this understanding, machines won’t be able to adapt to new problems, recover from errors, and learn similarly to the way humans do. But the brain is too complex to simply inspect it and understand how it works. We have to develop hypotheses and test them, and this requires models of intelligence which can benefit tremendously from machines that solve similar tasks. Neuroscience and engineering are tied together by new developments in cognitive science that characterize the concepts and algorithms that humans use. By promoting close collaborations between these domains of research, we will significantly advance our understanding of brains, minds, and machines.

Preliminary Schedule

Day 1

Tuesday, June 25

   
9:30 Coffee & Welcome
  including: Alessandro Verri, University of Genoa, Machine Learning Group
10:00 Tommy Poggio
  “Solving 3 theory puzzles about Deep Learning”
10:45 Tali Tishby
  “The information theory of deep learning and the surprising computational benefits of hidden layers”
11:30 Coffee
12:00 Lorenzo Rosasco
  “Implicit regularization: a guided tour”
12:45 Lunch
14:45 Amnon Shashua
  “Deep networks meet quantum physics”
15:30 Jim DiCarlo
  “Reverse engineering human visual intelligence”
16:15 Coffee
16:45 Daphna Weinshall
  “All deep networks are created equal”
17:30 Boris Katz
  “ObjectNet: Understanding the performance of Vision Systems”
18:15 Shai Shalev-Shwartz
  “Is deeper better only when shallow is good?”
  Dinner on own
   

Day 2

Wednesday, June 26

   
10:00 Philip Isola
  “A case for structured objectives”
10:45 Yedid Hoshen
  "Non-adversarial deep learning"
11:30 Coffee
12:00 Josh McDermott
  “Metamers of deep neural networks”
12:45 Lunch
14:45 Yair Weiss
  “Adversarial examples as optimal percepts”
15:30 Katharina Dobs
  “Face representations in deep neural networks”
16:15 Coffee
16:45 Haim Sompolinsky
  “Deep perceptual learning”
17:30 Merav Ahissar
  “The dynamics of forming perceptual priors and its importance to skill acquisition”
19:00 Workshop Dinner
   

Day 3

Thursday, June 27

   
9:00 Shmuel Peleg
  “Benefits and limitations using audio-visual analysis”
9:45 Gabriel Kreiman
  “Finding any Waldo: zero-shot invariant and efficient visual search”
10:30 Coffee
11:00 Shimon Ullman
  “Combining bottom-up and top-down processing in scene understanding”
11:45 Michael Werman
  “Learning with small memory”
13:00 Lunch

Participants:

Merav Ahissar, The Hebrew University of Jerusalem
Jim DiCarlo, Massachusetts Institute of Technology
Katharina Dobs, Massachusetts Institute of Technology
Yedid Hoshen, The Hebrew University of Jerusalem
Philip Isola, Massachusetts Institute of Technology
Boris Katz, Massachusetts Institute of Technology
Gabriel Kreiman, Children's Hospital Boston, Harvard Medical School
Josh McDermott, Massachusetts Institute of Technology
Shmuel Peleg, The Hebrew University of Jerusalem
Tomaso Poggio, Massachusetts Institute of Technology
Lorenzo Rosasco, Istituto Italiano di Tecnologia, Università degli Studi di Genova
Shai Shalev Shwartz, The Hebrew University of Jerusalem
Amnon Shashua, The Hebrew University of Jerusalem
Haim Sompolinsky, Harvard University; The Hebrew University of Jerusalem
Naftali Tishby, The Hebrew University of Jerusalem
Shimon Ullman, MIT, Weizmann Institute of Science
Daphna Weinshall, The Hebrew University of Jerusalem
Yair Weiss, The Hebrew University of Jerusalem
Michael Werman, The Hebrew University of Jerusalem