A Comparative Study of Machine Learning Methods for Persistence Diagrams

Abstract

Many and varied methods currently exist for featurization, which is the process of mapping persistence diagrams to Euclidean space, with the goal of maximally preserving structure. However, and to our knowledge, there are presently no methodical comparisons of existing approaches, nor a standardized collection of test data sets. This paper provides a comparative study of several such methods. In particular, we review, evaluate, and compare the stable multi-scale kernel, persistence landscapes, persistence images, the ring of algebraic functions, template functions, and adaptive template systems. Using these approaches for feature extraction, we apply and compare popular machine learning methods on five data sets: MNIST, Shape retrieval of non-rigid 3D Human Models (SHREC14), extracts from the Protein Classification Benchmark Collection (Protein), MPEG7 shape matching, and HAM10000 skin lesion data set. These data sets are commonly used in the above methods for featurization, and we use them to evaluate predictive utility in real-world applications.

Persistent homology reveals strong phylogenetic signal in 3D protein structures

“…Filling this gap will require filtration algorithms that consider additional information such as the nature and the physicochemical features of amino acids, which represents a promising direction for the future development of topological analysis in phylogenetics. Finally, another line of research is exploring the potential contribution of approaches combining PH with machine learning to study protein structure evolution, as PH has been shown to significantly improve classical machine learning ( 89 , 90 ). Indeed, PH-descriptors provide vectorizations of the complex geometry of protein structures, facilitating the application of learning methods.…”

TDAExplore: Quantitative analysis of fluorescence microscopy images through topology-based machine learning

“…We use persistence landscapes 7 to convert persistence diagrams into vectors, which are then used for machine learning. There are many other ways to do this 41 and each of these may be used with our method in place of persistence landscapes. We chose persistence landscapes for two advantages.…”

Persistent Homology Analysis for Solar Magnetograms

“…This cropping value (i.e., the filtration value) is systematically varied from the lowest to the highest values of the image, or vice versa, thus generating a different subspace for each value. As a result, the persistence homology analysis captures and examines the evolution of topological features across different thresholds, enabling insights into the structural properties of the image at various scales (Barnes et al 2021).…”