Skip to yearly menu bar Skip to main content

Workshop: Advances in Programming Languages and Neurosymbolic Systems (AIPLANS)

Proof Extraction for Logical Neural Networks

Thabang Lebese · Ndivhuwo Makondo · Cristina Cornelio · Naweed A Khan


Automated Theorem Provers (ATPs) are widely used for the verification of logical statements. Explainability is one of the key advantages of ATPs: providing an expert readable proof path which shows the inference steps taken to conclude correctness. Conversely, Neuro-Symbolic Networks (NSNs) that perform theorem proving, do not have this capability. We propose a proof-tracing and filtering algorithm to provide explainable reasoning in the case of Logical Neural Networks (LNNs), a special type of Neural-Theorem Prover (NTP).