Abstract:In this paper, we address the classification of instances each characterized not by a singular point, but by a distribution on a vector space. We employ the Wasserstein metric to measure distances between distributions, which are then used by distance-based classification algorithms such as k-nearest neighbors, k-means, and pseudo-mixture modeling. Central to our investigation is dimension reduction within the Wasserstein metric space to enhance classification accuracy. We introduce a novel approach grounded in the principle of maximizing Fisher's ratio, defined as the quotient of between-class variation to within-class variation. The directions in which this ratio is maximized are termed discriminant coordinates or canonical variates axes. In practice, we define both between-class and within-class variations as the average squared distances between pairs of instances, with the pairs either belonging to the same class or to different classes. This ratio optimization is achieved through an iterative algorithm, which alternates between optimal transport and maximization steps within the vector space. We conduct empirical studies to assess the algorithm's convergence and, through experimental validation, demonstrate that our dimension reduction technique substantially enhances classification performance. Moreover, our method outperforms well-established algorithms that operate on vector representations derived from distributional data. It also exhibits robustness against variations in the distributional representations of data clouds.

Combinatorial invariants of finite metric spaces and the Wasserstein arrangement

Geometric Structures on the Complement of a Projective Arrangement

Decomposition of the Kantorovich problem and Wasserstein distances on simplexes

Combinatorial invariants of metric filtrations and automorphisms; the universal adic graph

On nonexpansiveness of metric projection operators on Wasserstein spaces

Wasserstein geometry and Ricci curvature bounds for Poisson spaces

Canonical Variates in Wasserstein Metric Space

Algebraic Wasserstein distances and stable homological invariants of data

Ultralimits of Wasserstein spaces and metric measure spaces with Ricci curvature bounded from below

A Geometric Approach to Problems in Birational Geometry.

Geometry of the Space of Partitioned Networks: A Unified Theoretical and Computational Framework

Combinatorics of free and simplicial line arrangements

Coarse homological invariants of metric spaces

Gromov-Hausdorff convergence of metric pairs and metric tuples

Higher-categorical combinatorics of configuration spaces of Euclidean space

A Wasserstein-Type Distance for Gaussian Mixtures on Vector Bundles with Applications to Shape Analysis

On the geometry of Wasserstein barycenter I

On the metric spaces of lattices and periodic point sets

Maximizers of nonlocal interactions of Wasserstein type

Homogeneous isosceles-free spaces

On the Reach of Isometric Embeddings into Wasserstein Type Spaces