Hyperspectral Image Clustering with Spatially-Regularized Ultrametrics

Abstract

We propose a method for the unsupervised clustering of hyperspectral images based on spatially regularized spectral clustering with ultrametric path distances. The proposed method efficiently combines data density and spectral-spatial geometry to distinguish between material classes in the data, without the need for training labels. The proposed method is efficient, with quasilinear scaling in the number of data points, and enjoys robust theoretical performance guarantees. Extensive experiments on synthetic and real HSI data demonstrate its strong performance compared to benchmark and state-of-the-art methods. Indeed, the proposed method not only achieves excellent labeling accuracy, but also efficiently estimates the number of clusters. Thus, unlike almost all existing hyperspectral clustering methods, the proposed algorithm is essentially parameter-free.

Clustering and visualization of single-cell RNA-seq data using path metrics

“…These metrics balance density and geometry considerations in the data via computation of a density-weighted geodesic distance, making them useful for many machine learning tasks such as clustering and semi-supervised learning [40][41][42][43][44][45][46][47][48]. They have performed well in applications such as imaging [46,47,49,50], but their usefulness for the analysis of scRNAseq data remains unexplored.…”

Clustering and visualization of single-cell RNA-seq data using path metrics

“…These metrics balance density and geometry considerations in the data via computation of a density-weighted geodesic distance, making them useful for many machine learning tasks such as clustering and semi-supervised learning (40)(41)(42)(43)(44)(45)(46)(47)(48). They have performed well in applications such as imaging (46,47,49,50), but their usefulness for the analysis of scRNAseq data remains unexplored. Because these metrics are density-sensitive, they reduce cluster variance; in addition, these metrics also capture global distance information, and thus preserve global geometry; see Figure 4b.…”

“…These metrics balance density and geometry considerations in the data via computation of a density-weighted geodesic distance, making them useful for many machine learning tasks such as clustering and semi-supervised learning (40–48). They have performed well in applications such as imaging (46, 47, 49, 50), but their usefulness for the analysis of scRNAseq data remains unexplored.…”