independent Tensor Knowledge Graph Reasoning Engine research

The post-doctoral research assignment at City St George’s, University of London completed successfully some months ago. The client funding the research (GCHQ, via the Alan Turing Institute) commissioned a one-year 2nd phase extension to that research project. Dr. Ernesto Jimenez-Ruiz offered me a one-year extension to the post-doctoral research assistant role for that 2nd phase of research. After some tough soul searching, I declined the opportunity. Rather than continue researching LLM-assisted ontology alignment with LogMap, I opted instead to go independent so as to have the freedom to pursue other, more personal, research priorities.

Since that time I have primarily been pursuing my vision (from my PhD research) of a Tensor Knowledge Graph Reasoning Engine tailored for use in neurosymbolic systems that seek to combine subsymbolic (neural) learning with symbolic reasoning. This journey is proving to be a fascinating exploration of the relationship between mathematics and (description) logics. My notion of a Tensor Knowledge Graph (TKG) uses an unambiguously mathematical representation of ontological knowledge — binary matrices. My notion of a Tensor Knowledge Graph Reasoner (TKGR) performs all reasoning (logical inference) using unambiguously mathematical operations — set operations and (binary) relation operations applied to the knowledge matrices of a TKG.