KRHCAI 2021

Why KRHCAI

AI research over the past few decades has highlighted the strengths and weaknesses of both symbolic and machine learning approaches to AI, with some weaknesses in one being strengths in the other. Much is also known about the fragmentation of the field along these two approaches. However, recent discussions in AI have highlighted the need to integrate both symbolic and sub-symbolic methods, in a hybrid approach to AI, to create novel techniques that leverage both reasoning and learning. This workshop seeks to contribute to this discussion by exploring the following:

a systems approach to AI,
the use of composable AI components,
leveraging the best of both symbolic and sub-symbolic techniques in hybrid reasoning and learning architectures.

To achieve the above, there is the need to explore novel knowledge representation techniques that allow the seamless flow of computations between symbolic and sub-symbolic AI components. There is also the need to develop new datasets that evaluate the capabilities highlighted above, especially focusing on problems that cannot be solved by end-to-end differentiable neural network architectures or purely symbolic reasoning methods alone.

Join Sessions

Zoom meeting and Slack links can be found on the KR website
https://kr2021.kbsg.rwth-aachen.de/static/program/.

Program

Time (UTC)	Item
08:30 - 08:45	Welcome & Introduction
08:45 - 09:45	Invited Talk by Alessandra Russo Symbolic Machine Learning and its role in Neuro-symbolic AI
09:45 - 09:55	Lightning Talk by Chloé Mercier Ontology as Neuronal-Space Manifold: Towards Symbolic and Numerical Artificial Embedding [Paper]
09:55 - 10:55	Invited Talk by Antoine Bosselut Symbolic Scaffolds for Neural Commonsense Representation and Reasoning
10:55 - 11:00	Coffee Break
11:00 - 11:10	Lightning Talk by Kwabena Nuamah Deep Algorithmic Question Answering: Towards a Compositionally Hybrid AI for Algorithmic Reasoning [Paper]
11:10 - 12:10	Invited Talk by Alexander Gray Logical Neural Networks
12:10 - 12:50	Break
12:50 - 13:00	Lightning Talk by Patrick Kage Class Introspection: A Novel Technique for Detecting Unlabeled Subclasses by Leveraging Classifier Explainability Methods [Paper]
13:00 - 14:00	Invited Talk by Pasquale Minervini From Complex Query Answering to Neural Theorem Proving
14:00 - 15:00	Panel Discussion Alexander Gray, Antoine Bosselut, Alessandra Russo, Pasquale Minervini, Vaishak Belle, Jaehun Lee
15:00 - 15:20	General Discussion & Wrap Up