Abstract:The Gaussian kernel and its traditional normalizations (e.g., row-stochastic) are popular approaches for assessing similarities between data points. Yet, they can be inaccurate under high-dimensional noise, especially if the noise magnitude varies considerably across the data, e.g., under heteroskedasticity or outliers. In this work, we investigate a more robust alternative -- the doubly stochastic normalization of the Gaussian kernel. We consider a setting where points are sampled from an unknown density on a low-dimensional manifold embedded in high-dimensional space and corrupted by possibly strong, non-identically distributed, sub-Gaussian noise. We establish that the doubly stochastic affinity matrix and its scaling factors concentrate around certain population forms, and provide corresponding finite-sample probabilistic error bounds. We then utilize these results to develop several tools for robust inference under general high-dimensional noise. First, we derive a robust density estimator that reliably infers the underlying sampling density and can substantially outperform the standard kernel density estimator under heteroskedasticity and outliers. Second, we obtain estimators for the pointwise noise magnitudes, the pointwise signal magnitudes, and the pairwise Euclidean distances between clean data points. Lastly, we derive robust graph Laplacian normalizations that accurately approximate various manifold Laplacians, including the Laplace Beltrami operator, improving over traditional normalizations in noisy settings. We exemplify our results in simulations and on real single-cell RNA-sequencing data. For the latter, we show that in contrast to traditional methods, our approach is robust to variability in technical noise levels across cell types.

Computationally Efficient Learning of Statistical Manifolds

Manifold learning with approximate nearest neighbors

Deep Manifold Computing and Visualization

Deep Manifold Computing and Visualization Using Elastic Locally Isometric Smoothness

Histogram-based Embedding for Learning on Statistical Manifolds

Learning on manifolds without manifold learning

Adaptive Manifold Learning

Manifold Learning on Empirical Probability Distributions

Manifold Approximation by Moving Least-Squares Projection (MMLS)

Robust Inference of Manifold Density and Geometry by Doubly Stochastic Scaling

Learning High-Dimensional Correspondence Via Manifold Learning and Local Approximation

Entropy-based Closure for Probabilistic Learning on Manifolds

Manifold learning through locally linear reconstruction based on Euclidean distance

Manifold Learning for the Emulation of Spatial Fields from Computational Models.

Manifold learning in Wasserstein space

Manifold Optimization Assisted Gaussian Variational Approximation

Manifold-based learning and synthesis.

A geometric viewpoint of manifold learning

Distributed Learning via Filtered Hyperinterpolation on Manifolds

Efficient Isometric Multi-Manifold Learning Based on the Self-Organizing Method.

Hierarchical Structuring Of Data On Manifolds