Neurosymbolic Diffusion Models

Edoardo Maria Ponti (University of Edinburgh) · Emile van Krieken (University of Edinburgh) · Pasquale Minervini (University College London) · Antonio Vergari (University of Edinburgh)

conditional independence discrete diffusionhigh-dimensional visual path planninginteractionsneurosymbolic diffusion modelsneurosymbolic predictorsout-of-distribution generalisationoverconfident predictionsrule-based autonomous drivingscalable learning state-of-the-art accuracystrong calibrationsymbol dependenciessymbolic reasoning uncertainty modeling uncertainty quantification

Neurosymbolic (NeSy) predictors combine neural perception with symbolic reasoning to solve tasks like visual reasoning. However, standard NeSy predictors assume conditional independence between the symbols they extract, thus limiting their ability to model interactions and uncertainty