Special Session: KR And Machine Learning

Over the last two decades Machine Learning (ML) has made incredible advancements showing to be very effective in solving specific tasks and robust in many experimental learning applications. Deep learning, statistical (relational) learning, reinforcement learning and (logic-based and/or probabilistic) learning are among the many ML approaches that are witnessing such advancements. On the other hand, Knowledge Representation and Reasoning (KR) has continued to be at the core of Artificial Intelligence (AI) research providing solutions for explicit declarative representation of knowledge and knowledge-based inference, which have theoretical and practical relevance in many aspects of AI as well as in new emerging fields outside AI. The synergy between these two areas of AI has the potential to unlock new advancements on foundations of AI that offer new insights into open fundamental challenges included, but not limited to, learning symbolic generalisations from raw (multi-modal) data, using knowledge to facilitate data-efficient learning, supporting interpretability of learned outcomes, federated multi-agent learning and decision making.

This year, for the first time, KR2020 will host a special session on "Knowledge Representation and Machine Learning". This special session aims at providing researchers and industrial practitioners with a dedicated forum for presentation and discussion of new ideas, research experience and emerging results on topics related to computational learning and symbolic knowledge representation and reasoning. This special session provides the opportunity for fostering meaningful connections between researchers from these two main areas of AI and, at the same time, offering the possibility to learn about progress made on these topics, share their own views and learn about approaches that could lead to effective cross-fertilisation among research in ML and KR and new innovative solutions to key AI research challenges.

Expected contributions

The Special Session on KR and ML at KR2020 invites submissions of papers across KR and ML on advancements in one of these areas for the purpose of addressing open research challenges in the other, integration of computational learning and knowledge representation and reasoning, and the application of combined KR and ML approaches to solve real-world problems, including case studies and benchmarks.

We welcome papers on a wide range of topics, including but not limited to:

• Learning ontologies and knowledge graphs
• Learning action theories
• Learning common-sense knowledge
• Learning spatial and temporal theories
• Learning preference models
• Learning causal models
• Learning tractable probabilistic models
• Probabilistic reasoning and learning
• Graphical models for knowledge representation and reasoning
• Reasoning and learning over knowledge graphs
• Logic-based learning algorithms
• Neural-symbolic learning
• Statistical relational learning
• Multi-agent learning
• Machine learning for efficient knowledge inference
• Symbolic reinforcement learning
• Learning symbolic abstractions from unstructured data
• Machine-learning-driven reasoning algorithms
• Explainable AI
• Transfer learning
• Multi-agent learning
• Expressive power of learning representations
• Knowledge-driven natural language understanding and dialogue
• Knowledge-driven decision making
• Knowledge-driven intelligent systems for internet of things and cybersecurity
• Application of knowledge-driven ML to question answering and story understanding
• Application of knowledge-driven ML to Robotics

Submission Guidelines and Evaluation Criteria

The special session emphasizes KR and ML, and welcomes contributions that extend the state of the art at the intersection of KR and ML. Therefore, KR-only or ML-only submissions will not be accepted for evaluation in this special session.

Submissions will be rigorously peer reviewed by PC members, who are active in KR and ML. Submissions will be evaluated on the basis of the overall quality of their technical contribution, including criteria such as originality, soundness, relevance, significance, quality of presentation, and understanding of the state of the art.

In this special session, the selection process of the highest quality papers will apply the following criteria:

• Importance and novelty of using knowledge representation and reasoning to advance machine learning, or novelty of using machine learning solutions to advance knowledge representation and reasoning.
• Applicability of the proposed solutions in real-world.
• Reusability of datasets, case studies and benchmarks for systems and/or application papers.
• Proved theoretical or empirically demonstrated practical advancement of the proposed solution with respect to baseline pure KR or ML approaches.

For submissions follow the instructions in the submission page .

Chairs

Alessandra Russo (Imperial College, London)
Guy Van den Broeck (UCLA, USA)