Defeasible Reasoning in Formal Concept Analysis

Aus International Center for Computational Logic
Wechseln zu:Navigation, Suche

Defeasible Reasoning in Formal Concept Analysis

Vortrag von Sergei Obiedkov
Formal Concept Analysis (FCA) provides a framework for understanding dependencies between attributes through implications. However, classical implications fall short when dealing with noisy or exception-prone data. In this talk, we explore how the well-established KLM framework for defeasible reasoning can be adapted to FCA. We introduce defeasible conditionals—statements like "normally, A implies B"—into the FCA setting via a typicality ordering over objects. This ordering can be either specified upfront or derived from background knowledge. The resulting framework preserves the rationality postulates of KLM and supports context-sensitive inference grounded in the structure of the data. Joint work with Lucas Carr, Nicholas Leisegang, and Thomas Meyer (University of Cape Town and CAIR, South Africa)