Abstract:A family of probability density functions (pdfs) is defined on the unit hypersphere $$S^{n}$$ . The parameter space for the pdfs is $$G(d, n+1)\times \mathbb {R}_{\ge 0}$$ , for $$1\le d\le n$$ , where $$G(d,n+1)$$ is the Grassmannian of d-dimensional linear subspaces in $$\mathbb {R}^{n+1}$$ and $$\mathbb {R}_{\ge 0}$$ is the range of values for a concentration parameter. This family of pdfs generalises the Watson distribution on the sphere $$S^{2}$$ . It is shown that the pdfs are tractable, in that (i) a given pdf can be sampled efficiently, (ii) the parameters of a pdf can be estimated using maximum likelihood, and (iii) the Kullback–Leibler divergence and the Fisher–Rao metric on $$G(d, n+1)\times \mathbb {R}_{\ge 0}$$ have simple forms. A wide range of shapes of the pdfs can be obtained by varying d and the concentration parameter. The pdfs are used to model clusters of feature vectors on the hypersphere. The clusters are compared using the Kullback–Leibler divergences of the associated pdfs. Experiments with the mnist, Human Activity Recognition and Gas Sensor Array Drift datasets show that good results can be obtained from clustering algorithms based on the Kullback–Leibler divergence, even if the dimension n of the hypersphere is high.

Universal Approximation on the Hypersphere

Optimal Reduction of Multivariate Dirac Mixture Densities

On the best approximation by finite Gaussian mixtures

Differentiability and Approximation of Probability Functions under Gaussian Mixture Models: A Bayesian Approach

On Uniform Approximations to Hypersingular Finite-Part Integrals

Adaptive Bayesian multivariate density estimation with Dirichlet mixtures

A multivariate normal approximation for the Dirichlet density and some applications

Generalised Watson Distribution on the Hypersphere with Applications to Clustering.

Multivariate Approximating Averages

Parameter Estimation and Model-Based Clustering with Spherical Normal Distribution on the Unit Hypersphere

Infinite-dimensional integration and $L^2$-approximation on Hermite spaces

A Short Note on the Volume of Hypersphere

Geometric and arithmetic aspects of approximation vectors

Approximation of Hardy Space on the Unit Sphere

Approximation in Hölder Spaces

Uniform decomposition of probability measures: quantization, clustering and rate of convergence

On the capacity of deep generative networks for approximating distributions

Approximate independence of permutation mixtures

Risk Bounds for Mixture Density Estimation on Compact Domains via the $h$-Lifted Kullback--Leibler Divergence

Approximation of plurisubharmonic functions by logarithms of Gaussian analytic functions

Universal Approximation of Symmetric and Anti-Symmetric Functions.