Dynamic Logic learning in cognitive-based situation models

Abstract

we present a cognitive modeling framework called Neural Modeling Fields (NMF) and its application to situation learning and categorization. We discuss how this framework is related to the perceptual symbol systems theory of cognition (PSS). Essentially, the mathematical apparatus of NMF is a way to learn the frames and simulators described qualitatively by PSS. For the purposes of this work, a situation is modeled as a set of objects and relationships that exist among them. Here we consider object recognition problem solved and demonstrate how the NMF framework is used to learn high level concepts such as situations.

Combined linguistic and sensor models for machine learning

“…This work builds on a cognitive theory proposed in [1], called dynamic logic, also referred to as neural modeling fields (NMF), and modeling fields theory (MFT) in other publications [2] [3] [4] [5] [6].This is a cognitively inspired mathematical framework providing a generic way of finding an optimal match between a set of parametric probabilistic models and sensor input data. The probabilistic models provide the ability to include a priori knowledge and to model the variability of real world data.…”