2026-05-29

On Tuesday, 2^nd June 2026, 13:00, Chaithra Umesh, PhD student at SBI, will give a presentation on “Scalable extraction and visualization of multi-attribute logical and functional dependencies in tabular data” as part of our research seminar. Afterwards we look forward to discussions. The seminar is open to anyone interested.

Abstract

Higher-order logical dependencies (LDs) and functional dependencies (FDs) capture
important structural relationships in tabular data. Existing LD discovery approaches remain
limited to pairwise relationships and are considerably less explored than FDs. Although
multi-attribute FD extraction methods exist, they become computationally expensive in high-
dimensional feature spaces. Moreover, there are no unified frameworks for jointly extracting
multi-attribute LDs and FDs from tabular datasets.
In this talk, I present a unified framework that leverages hypergraph-guided search-space
reduction and KD-tree-based approximation to extract multi-attribute LDs and FDs
efficiently. I also introduce hypergraph-based visualizations to illustrate dependency
structures.
Experimental results from simulated and real-world datasets demonstrate that this
framework effectively identifies meaningful higher-order dependencies while reducing
computational time compared to exhaustive methods. This approach aids in analyzing
purchasing patterns in e-commerce, gene expression interactions in biomedical research,
and preserving dependencies in synthetic tabular data.

Zoom: https://uni-rostock-de.zoom.us/j/67422288014