Spectral Clustering Based on Local PCA

Abstract

We propose a spectral clustering method based on local principal components analysis (PCA). After performing local PCA in selected neighborhoods, the algorithm builds a nearest neighbor graph weighted according to a discrepancy between the principal subspaces in the neighborhoods, and then applies spectral clustering. As opposed to standard spectral methods based solely on pairwise distances between points, our algorithm is able to resolve intersections. We establish theoretical guarantees for simpler variants within a prototypical mathematical framework for multi-manifold clustering, and evaluate our algorithm on various simulated data sets.

References in corpus (1)

• Robust Multiple Manifolds Structure Learning

Cited by in corpus (17)

• Nonparametric Density Estimation for High-Dimensional Data - Algorithms and Applications
• Learning by Unsupervised Nonlinear Diffusion
• A New Approach To Two-View Motion Segmentation Using Global Dimension Minimization
• Beyond Evolutionary Algorithms for Search-based Software Engineering
• Spectral Convergence of the connection Laplacian from random samples
• Causal Inference in Possibly Nonlinear Factor Models
• Influential Feature PCA for high dimensional clustering
• Recovering the homology of immersed manifolds
• K-Deep Simplex: Deep Manifold Learning via Local Dictionaries
• Large sample spectral analysis of graph-based multi-manifold clustering
• The Shape of Data and Probability Measures
• Adversarial Manifold Estimation
• Cubical Covers of Sets in ${\mathbb{R}}^n$
• Vector Quantized Spectral Clustering applied to Soybean Whole Genome Sequences
• Measuring inter-cluster similarities with Alpha Shape TRIangulation in loCal Subspaces (ASTRICS) facilitates visualization and clustering of high-dimensional data
• Conformal Prediction based Spectral Clustering
• Connecting Dots -- from Local Covariance to Empirical Intrinsic Geometry and Locally Linear Embedding