“…TDA and tools of persistent homology (PH) have recently emerged as powerful approaches for ML, allowing us to extract complementary information on the observed objects, especially, from graph-structured data. In particular, PH has become popular for various ML tasks such as clustering, classification, and anomaly detection, with a wide range of applications including material science [68, 43], insurance [99, 46], finance [55], and cryptocurrency analytics [33, 4, 73]. (For more details see surveys [6, 22] and TDA applications library [34]) Furthermore, it has become a highly active research area to integrate PH methods into geometric deep learning (GDL) in recent years [41, 100, 19, 23].…”