Boundary detection algorithm inspired by locally linear embedding
boundary detectionlocal covariance matrixlocally linear embeddingmanifold learningmanifold with boundarynearest neighborhood search scheme
Applications of differential geometry to data and computer science (53Z50) Spectral theory; eigenvalue problems on manifolds (58C40) Nonparametric estimation (62G05) Classification and discrimination; cluster analysis (statistical aspects) (62H30) Learning and adaptive systems in artificial intelligence (68T05)
Locally linear embedding (LLE) is an unsupervised learning algorithm that computes low-dimensional, neighborhood-preserving embeddings of high-dimensional inputs. LLE maps its inputs into a single global coordinate system of lower dimensionality. In the second section of the article, some basics on the (LLE) technique are provided. For details, see [\textit{S. T. Rowels} and \textit{I. K. Saul}, Science 290, No. 5500, 2323--2326 (2000; \url{doi:10.1126/science.290.5500.2323}).\N\NLLE has been widely applied in data science. The present paper is devoted to the problem of identification of boundary points from samples on an embedded compact manifold with boundary. Inspired by the LLE technique, a new algorithm is developed, the Boundary Detection algorithm (BD-LLE) which uses barycentric coordinates and can be implemented using either an \( \epsilon\)-radius ball scheme or a KNN scheme.\N\NThe architecture of the Boundary Detection algorithm is described in the third section. The model of a manifold with boundary is introduced and the steps of the BD-LLE algorithm are shown. These include the construction of a boundary indicator (BI), its relation to the local covariance matrix, the selection of the regularizer and selection of the scale parameters. For details see the SM1- and SM2-parts of the Supplementary Material. The theoretical analysis of the local covariance matrix in the KNN scheme and of the boundary indicator in the \(\epsilon\)-radius ball scheme as well as in the KNN scheme are done in the fourth section of the article. One reports that the proofs of all results are to be found in the SM3 and SM4 -- parts of the Supplementary Material. Numerical results on performances of BD-LLE algorithm are summarized in the fifth section. One reports that details and discussions are summarized in the SM5 section of the Supplementary Material. Boundary detection on noisy data is developed in the sixth section of the article; one mentions that more details and a review of the diffusion maps (DM) algorithm are to be found in the SM6-part of the Supplementary Material. For general discussions and future work, section seven of the article is reserved.
- A computational geometry approach for determination of boundary of workspaces of planar manipulators with arbitrary topology.
- Boundary estimation from point clouds: algorithms, guarantees and applications
- Convergence of graph Laplacian with kNN self-tuned kernels
- Density estimation on manifolds with boundary
- Diffusion maps
- Diffusion maps for embedded manifolds with boundary with applications to PDEs
- Ghost Point Diffusion Maps for Solving Elliptic PDEs on Manifolds with Classical Boundary Conditions
- Graph Based Gaussian Processes on Restricted Domains
- Graph connection Laplacian methods can be made robust to noise
- Impact of Signal-to-Noise Ratio and Bandwidth on Graph Laplacian Spectrum From High-Dimensional Noisy Point Cloud
- Improved spectral convergence rates for graph Laplacians on \(\varepsilon \)-graphs and \(k\)-NN graphs
- Local eigenvalue decomposition for embedded Riemannian manifolds
- Local linear regression on manifolds and its geometric interpretation
- Minimax boundary estimation and estimation with boundary
- Multiscale geometric methods for data sets. I: Multiscale SVD, noise and curvature.
- Non-asymptotic analysis of tangent space perturbation
- On the shape of a set of points in the plane
- Scalability and robustness of spectral embedding: landmark diffusion is all you need
- Set Estimation from Reflected Brownian Motion
- Tangent space estimation for smooth embeddings of Riemannian manifolds
- Think globally, fit locally under the manifold setup: asymptotic analysis of locally linear embedding
- Three-dimensional alpha shapes
- Vector diffusion maps and the connection Laplacian
This page was built for publication: Boundary detection algorithm inspired by locally linear embedding
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6876747)
