An Empirical Study on the Performance of Spectral Manifold Learning Techniques
- 4.6k Downloads
In recent years, there has been a surge of interest in spectral manifold learning techniques. Despite the interest, only little work has focused on the empirical behavior of these techniques. We construct synthetic data of variable complexity and observe the performance of the techniques as they are subjected to increasingly difficult problems. We evaluate performance in terms of both a classification and a regression task. Our study includes Isomap, LLE, Laplacian eigenmaps, and diffusion maps. Among others, our results indicate that the techniques are highly dependent on data density, sensitive to scaling, and greatly influenced by intrinsic dimensionality.
KeywordsNear Neighbour Regression Task Regression Measure Nonlinear Dimensionality Reduction Intrinsic Dimensionality
Unable to display preview. Download preview PDF.
- 4.Brand, M.: Charting a Manifold. In: NIPS, vol. 15, pp. 977–984. IEEE Press, Los Alamitos (2003)Google Scholar
- 5.Donoho, D.L., Grimes, C.: Hessian Eigenmaps: Locally Linear Embedding Techniques for High-dimensional Data. In: PNAS, vol. 100, pp. 5591–5596. National Academy Sciences, Washington (2003)Google Scholar
- 8.Yeh, M.C., Lee, I.H., Wu, G., Wu, Y., Chang, E.Y.: Manifold Learning, a Promised Land or Work in Progress. In: Proc. of IEEE Intl. Conf. on Multimedia and Expo, pp. 1154–1157. IEEE Press, New York (2005)Google Scholar
- 9.Niskanen, M., Silven, O.: Comparison of Dimensionality Reduction Methods for Wood Surface Inspection. In: Proc. of the 6th Intl. Conference on Quality Control by Artificial Vision, pp. 179–188 (2003)Google Scholar
- 10.van der Maaten, L., Postma, E.O., van den Herik, H.J.: Dimensionality Reduction: A Comparative Review. Technical report, Tilburg Uni (2009)Google Scholar