Ulderico Fugacci (05/07/25) : Topology or Learning? A Naive Analysis

Abstract: Putting ourselves in the shoes of newcomer users of learning methods and topological data analysis, we investigate which of the two classes of approaches is the one to be preferred. Specifically, we compare persistent homology methods with traditional machine learning and deep learning techniques for label-efficient classification. Simple topological approaches - including persistence thresholding and Bottleneck distance classification - are evaluated against conventional learning algorithms and hybrid methods on two binary classification tasks: surface crack detection and malaria cell identification.

Associated paper: https://diglib.eg.org/items/1d10fe7b-...