Abstract
This paper proposes a tentative and original survey of meeting points betweenKnowledge Representation and Reasoning (KRR) and Machine Learning (ML), twoareas which have been developing quite separately in the last three decades.Some common concerns are identified and discussed such as the types of usedrepresentation, the roles of knowledge and data, the lack or the excess ofinformation, or the need for explanations and causal understanding. Then somemethodologies combining reasoning and learning are reviewed (such as inductivelogic programming, neuro-symbolic reasoning, formal concept analysis,rule-based representations and ML, uncertainty in ML, or case-based reasoningand analogical reasoning), before discussing examples of synergies between KRRand ML (including topics such as belief functions on regression, EM algorithmversus revision, the semantic description of vector representations, thecombination of deep learning with high level inference, knowledge graphcompletion, declarative frameworks for data mining, or preferences andrecommendation). This paper is the first step of a work in progress aiming at abetter mutual understanding of research in KRR and ML, and how they couldcooperate.