The effort to integrate logic with deep learning has intensified in recent years and has the potential to give rise to a new computational paradigm in which symbolic knowledge is used to assist deep learning systems or extend their capabilities, while offering, at the same time, a path towards the grounding of symbols and the induction of knowledge from low-level sensory data.
The synergy of these two different worlds is the topic of the workshop “When deep learning meets logic”. This workshop aims at the following: (i) present the applications that are enabled by this computational paradigm; (ii) explore the state-of-the-art and understand its level of maturity and adoption by the industrial sector; in particular, techniques and theory developed from both the deep learning and the logic communities will be presented; and (iii) identify some of the big questions that are open in this area and single out problems that require further investigation.
The workshop will take place from the February 15 to February 17, 2021, and all the recorded talks will be available in the workshop’s website.
Registration is free. To register, please follow the link
For any inquiries, please contact Efi Tsamoura at firstname.lastname@example.org or Alfie Hammali at email@example.com.
Samsung AI, Cambridge. Introduction/Greeting
Leslie Valiant, Harvard University.
Title: How to Augment Supervised Learning with Reasoning?
Martin Grohe, RWTH Aachen.
Jiajun Wu, Stanford University.
Daisy Zhe Wang, University of Florida.
Title: Neural-Symbolic models for Knowledge Graph Extraction and Reasoning.
Madhusudan Parthasarathy, University of Illinois.
Title: Exploring Combinations of Neural and Logic Learning
Balder ten Cate, Google.
Title: Providing deep models with access to structured knowledge - motivations, techniques, challenges.
Efi Tsamoura, Samsung AI Center, Cambridge.
Title: Neural-Symbolic Integration: A Compositional Perspective
Jacob Andreas, MIT.
Le Song, Georgia Institute of Technology.
Title: Efficient probabilistic logic reasoning with graph neural networks.
Ryan Riegel, IBM Research.
Christos Papadimitriou, Columbia University.