The NeSy workshop series celebrates the integration of connectionist and symbolic thinking, technologies, theories, and techniques in Artificial Intelligence systems.

NeSy is the annual workshop of the Neural-Symbolic Learning and Reasoning Association.

Statistical machine learning and connectionist systems have achieved industrial relevance in a number of areas, from retail to health, by their state-of-the-art performance in language modelling, speech recognition, graph analytics, image, video and sensor data analysis. On the other hand, symbolic AI systems are challenged by such unstructured data, but are recognized as being in principle transparent, in that reasoned facts from knowledge bases can be inspected to interpret how decisions follow from input. Connectionist and symbolic methods also contrast in the problems they excel at: scene recognition from images appears to be a problem still outside the capabilities of symbolic systems for example, while neural nets are not yet sufficient for industrial-strength complex planning scenarios and deductive reasoning tasks.

Neural-symbolic computation aims to build rich computataional models and systems by combining both connectionist and symbolic learning paradigms. This combination hopes to form synergies between their strengths while overcoming their complementary weaknesses.

The NeSy workshop series is a major venue for the presentation and discussion of key topics related to neural-symbolic computing. NeSy has provided an atmosphere for the free exchange of ideas and networking in the community of scientists that straddle the line between statistical and symbolic methods in AI.

The NeSy'19 Unofficial Proceedings including all accepted papers are available for download here!

NeSy'19 Keynote Speakers

Marco Gori, Professor of Computer Science, University of Siena

9:00am Keynote: Learning and reasoning with constraints

Learning and inference are traditionally regarded as the two opposite, yet complementary and puzzling components of intelligence. In this talk we point out that a constrained-based modeling of the environmental agent interactions makes it possible to unify learning and inference within the same mathematical framework.The unification is based on the abstract notion of constraint, which provides a representation of knowledge granules gained from the interaction with the environment. The agents are based on a deep neural networks that can also naturally deal with logic constraints, that are properly translated into real-valued functions by opportune t-norms. The theory prescribes how to incorporate and extract contraints, which opens the doors to truly new mechanisms of explainability. Interestingly, computational models like graph neural networks can also be incorporated thanks to the expression of structured domains by constraints. The basic ideas are presented by simple case studies ranging from learning and inference in social nets, missing data, checking of logic constraints, and pattern generation. The theory offers a natural bridge between the formalization of knowledge and the inductive acquisition of concepts from data.

Bio: Dr. Marco Gori is a leading researcher in machine learning with applications to pattern recognition, Web mining, and game playing. He is especially interested in bridging logic and learning and in the connections between symbolic and sub-symbolic representation of information. He is co-author of “Web Dragons: Inside the myths of search engines technologies,” Morgan Kauffman (Elsevier), 2006, and “Machine Learning: A Constrained-Based Approach,” Morgan Kauffman (Elsevier), 2018. Dr. Gori serves (has served) as an Associate Editor of a number of technical journals related to his areas of expertise, he has been the recipient of best paper awards, and keynote speakers in a number of international conferences. He was the Chairman of the Italian Chapter of the IEEE Computational Intelligence Society, and the President of the Italian Association for Artificial Intelligence. He is a fellow of the IEEE, ECCAI, IAPR. He is in the list of top Italian scientists kept by the VIA-Academy.

Luc De Raedt , Professor of Computer Science, KU Leuven

2:00pm Keynote: DeepProbLog: integrating probability, logic and neural networks.

This keynote will give an overview of our work on DeepProbLog [Manhaeve et al, NeurIPS 2018] that extends the probabilistic logic programming language with neural predicates in order to address the challenge of neuro-symbolic integration.

Bio: Dr. Luc De Raedt's research interests are in Artificial Intelligence, Machine Learning and Data Mining, as well as their applications. He is well known for his contributions in the areas of learning and reasoning, in particular, for his contributions to statistical relational learning and inductive programming. Today he is working on the next generation of programming languages, which can automatically learn from data, on combining probabilistic and logical reasoning and learning, on the automation of (data) science, and on verifying learning artificial intelligence systems and robotics. Luc De Raedt is currently head of the lab for Declarative Language of AI at KULeuven. He has coordinated several EU projects and was awarded an ERC Advanced Grant in 2015. He is a EurAI and AAAI fellow, he is on the editorial board of journals including Artificial Intelligence, Machine Learning and the Journal of Machine Learning Research.

George Cybenko, Dorothy and Walter Gramm Professor of Engineering, Dartmouth

4:00pm Keynote: Machine learning and symbolic processing in cyber security

Computer systems and attacks against them are increasing in complexity and have already surpassed the point where humans can comprehend the composition of large systems and the data streams required to manage and secure them. This talk will cover recent progress on "out-of-band" security monitoring of system execution using machine learning for classification of RF emissions and finite automata to track the classification results within the context of allowable control flow transitions within a program. This example will be followed by some general observations about hybrid neural and symbolic processing possibilities.

Bio: Dr. George Cybenko has made key research contributions in machine learning, information security, parallel processing and computational behavioral analysis, including the discovery of the universal approximation theorem for neural networks. He was the Founding Editor-in-Chief of IEEE Security & Privacy, which is the largest professional society publication focused on security. Professor Cybenko is a Fellow of the IEEE, has served on the Defense Science Board, the US Air Force Science Advisory Board and the US Army Cyber Institute Advisory Board. He was also Professor of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign before joining Dartmouth. He received his BS (University of Toronto) and PhD (Princeton) degrees in Mathematics.

Workshop Schedule

Keynote: Marco Gori, 9:00am

10:00 - 10:30am coffee break

10:30am Revisiting Neural-Symbolic Learning Cycle

10:50am T-PRISM: A tensorized logic programming language for data modelling

11:10am Injecting Prior Knowledge for Transfer Learning into Reinforcement Learning Algorithms using Logic Tensor Networks

11:30am Local ABox Consistency Prediction with Transparent TBoxes Using Gated Graph Neural Networks

11:50am Neural Variational Inference For Estimating Knowledge Graph Embedding Uncertainty

12:10pm Tensor Product Representations of Subregular Formal Languages

12:30-2pm - Lunch

Keynote: Luc de Raedt, 2:00pm

2:50pm One-shot Information Extraction from Document Images using Neuro-Deductive Program Synthesis

3:10pm Self-Organized Inductive Reasoning with NeMuS

3:30-4pm Coffee break

Keynote: George Cybenko, 4:00pm

5pm: Poster Spotlights [2 minute talks]

Until 6pm: Poster session

Related Events

Knowledge Representation & Reasoning Meets Machine Learning (KR2ML)

Presentation

All papers listed under the program will be presented during the workshop. The workshop will include extra time for audience discussion of the presentation allowing the group to have a better understanding of the issues, challenges, and ideas being presented. A presentation may selected for oral or poster presentation depending on the number of accepted papers and on the ideal presentation style for the submitted work.

Authors of papers that are well-reviewed or incites interest during the workshop will be invited to submit a revised and extended version of their papers to the Journal of Applied Logics, College Publications.