Variable bandwidth diffusion kernels
From MaRDI portal
Abstract: Practical applications of kernel methods often use variable bandwidth kernels, also known as self-tuning kernels, however much of the current theory of kernel based techniques is only applicable to fixed bandwidth kernels. In this paper, we derive the asymptotic expansion of these variable bandwidth kernels for arbitrary bandwidth functions; generalizing the theory of Diffusion Maps and Laplacian Eigenmaps. We also derive pointwise error estimates for the corresponding discrete operators which are based on finite data sets; generalizing a result of Singer which was restricted to fixed bandwidth kernels. Our analysis reveals how areas of small sampling density lead to large errors, particularly for fixed bandwidth kernels. We explain the limitation of the existing theory to data sampled from compact manifolds by showing that when the sampling density is not bounded away from zero (which implies that the data lies on an open set) the error estimates for fixed bandwidth kernels will be unbounded. We show that this limitation can be overcome by choosing a bandwidth function inversely proportional to the sampling density (which can be estimated from data) which allows us to control the error estimates uniformly over a non-compact manifold. We numerically verify these results on non-compact manifolds by constructing the generator of the Ornstein-Uhlenbeck process on a real line and a two-dimensional plane using data sampled independently from the respective invariant measures. We also verify our results on compact manifolds by constructing the Laplacian on the unit circle and the unit sphere and we show that the variable bandwidth kernels exhibit reduced sensitivity to bandwidth selection and give better results for an automatic bandwidth selection algorithm.
Recommendations
Cites work
- Anisotropic diffusion on sub-manifolds with application to Earth structure classification
- Consistency of spectral clustering
- Diffusion Maps, Reduction Coordinates, and Low Dimensional Representation of Stochastic Systems
- Diffusion maps
- Diffusion maps, spectral clustering and reaction coordinates of dynamical systems
- From graph to manifold Laplacian: the convergence rate
- Graph Laplacian Tomography From Unknown Random Projections
- Laplacian Eigenmaps for Dimensionality Reduction and Data Representation
- Learning Theory
- Linear theory for filtering nonlinear multiscale systems with model error
- Local kernels and the geometric structure of data
- Non-linear independent component analysis with diffusion maps
- Nonlinear Laplacian spectral analysis for time series with intermittency and low-frequency variability
- On Estimation of a Probability Density Function and Mode
- On Locally Adaptive Density Estimation
- Remarks on Some Nonparametric Estimates of a Density Function
- Spectral independent component analysis
- Time-scale separation from diffusion-mapped delay coordinates
- Variable kernel density estimation
Cited in
(59)- Eigen-convergence of Gaussian kernelized graph Laplacian by manifold heat interpolation
- Doubly stochastic normalization of the Gaussian kernel is robust to heteroskedastic noise
- Density estimation on manifolds with boundary
- Delay-coordinate maps and the spectra of Koopman operators
- Data-driven spectral decomposition and forecasting of ergodic dynamical systems
- Local kernels and the geometric structure of data
- Kernel Methods for Bayesian Elliptic Inverse Problems on Manifolds
- Localized Fourier analysis for graph signal processing
- Discovering causal structure with reproducing-kernel Hilbert space \(\epsilon\)-machines
- Graph-theoretic algorithms for Kolmogorov operators: approximating solutions and their gradients in elliptic and parabolic problems on manifolds
- Delay-coordinate maps, coherence, and approximate spectra of evolution operators
- Spectral methods for solving elliptic PDEs on unknown manifolds
- Particle-Continuum Multiscale Modeling of Sea Ice Floes
- Quantum Mechanics for Closure of Dynamical Systems
- Local and global perspectives on diffusion maps in the analysis of molecular systems
- Data-driven probability density forecast for stochastic dynamical systems
- Iterated diffusion maps for feature identification
- Diffusion forecasting model with basis functions from QR-decomposition
- Semiparametric modeling: correcting low-dimensional model error in parametric models
- Diffusion maps-aided neural networks for the solution of parametrized PDEs
- Dynamics-adapted cone kernels
- A parameter estimation method using linear response statistics: Numerical scheme
- Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques
- Gaussian bandwidth selection for manifold learning and classification
- IAN: Iterated Adaptive Neighborhoods for Manifold Learning and Dimensionality Estimation
- Reproducing kernel Hilbert space compactification of unitary evolution groups
- Diffusion maps tailored to arbitrary non-degenerate Itô processes
- Spatiotemporal pattern extraction by spectral analysis of vector-valued observables
- Robust Inference of Manifold Density and Geometry by Doubly Stochastic Scaling
- Kernel Analog Forecasting: Multiscale Test Problems
- Diffusion maps for embedded manifolds with boundary with applications to PDEs
- Spatiotemporal analysis using Riemannian composition of diffusion operators
- Approximating solutions of linear elliptic PDE's on a smooth manifold using local kernel
- Extraction and prediction of coherent patterns in incompressible flows through space-time koopman analysis
- Functional diffusion maps
- Time coupled diffusion maps
- Kernel-based prediction of non-Markovian time series
- Graph-based prior and forward models for inverse problems on manifolds with boundaries
- Spectral analysis of weighted Laplacians arising in data clustering
- Gaussian Process Landmarking for Three-Dimensional Geometric Morphometrics
- Diffusion maps embedding and transition matrix analysis of the large-scale flow structure in turbulent Rayleigh-Bénard convection
- Fractional diffusion maps
- Koopman spectra in reproducing kernel Hilbert spaces
- Forecasting turbulent modes with nonparametric diffusion models: learning from noisy data
- Error bounds of the invariant statistics in machine learning of ergodic Itô diffusions
- Kernel-based methods for solving time-dependent advection-diffusion equations on manifolds
- What is variable bandwidth?
- Conditional expectation using compactification operators
- Solving PDEs on unknown manifolds with machine learning
- Computing committors in collective variables via Mahalanobis diffusion maps
- Analysis of the feedback particle filter with diffusion map based approximation of the gain
- Dimensionality reduction of complex metastable systems via kernel embeddings of transition manifolds
- Koopman Operator Family Spectrum for Nonautonomous Systems
- Ghost Point Diffusion Maps for Solving Elliptic PDEs on Manifolds with Classical Boundary Conditions
- Balancing geometry and density: path distances on high-dimensional data
- Model free data assimilation with Takens embedding
- Manifold learning for organizing unstructured sets of process observations
- Efficient Bayesian estimation of the generalized Langevin equation from data
- Nonparametric uncertainty quantification for stochastic gradient flows
This page was built for publication: Variable bandwidth diffusion kernels
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q900774)
