Large-Scale Manifold Learning Export

@misc{citeulike:5919451, abstract = {This paper examines the problem of extracting lowdimensional manifold structure given millions of highdimensional face images. Specifically, we address the computational challenges of nonlinear dimensionality reduction via Isomap and Laplacian Eigenmaps, using a graph containing about 18 million nodes and 65 million edges. Since most manifold learning techniques rely on spectral decomposition, we first analyze two approximate spectral decomposition techniques for large dense matrices (Nystr\"{o}m and Column-sampling), providing the first direct theoretical and empirical comparison between these techniques. We next show extensive experiments on learning low-dimensional embeddings for two large face datasets: CMU-PIE (35 thousand faces) and a web dataset (18 million faces). Our comparisons show that the Nystr\"{o}m approximation is superior to the Column-sampling method. Furthermore, approximate Isomap tends to perform better than Laplacian Eigenmaps on both clustering and classification with the labeled CMU-PIE dataset. 1.}, author = {Talwalkar, Ameet and Rowley, Henry}, citeulike-article-id = {5919451}, citeulike-linkout-0 = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.142.1673}, keywords = {learning, manifold}, posted-at = {2009-10-10 14:46:19}, priority = {2}, title = {Large-Scale Manifold Learning}, url = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.142.1673} }

TY - ELEC ID - citeulike:5919451 L3 - citeulike-article-id:5919451 N2 - This paper examines the problem of extracting lowdimensional manifold structure given millions of highdimensional face images. Specifically, we address the computational challenges of nonlinear dimensionality reduction via Isomap and Laplacian Eigenmaps, using a graph containing about 18 million nodes and 65 million edges. Since most manifold learning techniques rely on spectral decomposition, we first analyze two approximate spectral decomposition techniques for large dense matrices (Nyström and Column-sampling), providing the first direct theoretical and empirical comparison between these techniques. We next show extensive experiments on learning low-dimensional embeddings for two large face datasets: CMU-PIE (35 thousand faces) and a web dataset (18 million faces). Our comparisons show that the Nyström approximation is superior to the Column-sampling method. Furthermore, approximate Isomap tends to perform better than Laplacian Eigenmaps on both clustering and classification with the labeled CMU-PIE dataset. 1. TI - Large-Scale Manifold Learning KW - learning KW - manifold AU - Talwalkar, Ameet AU - Rowley, Henry UR - http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.142.1673 ER -