1 / 22

Non-Isometric Manifold Learning Analysis and an Algorithm

Non-Isometric Manifold Learning Analysis and an Algorithm. Piotr Doll á r, Vincent Rabaud, Serge Belongie. University of California, San Diego. I. Motivation. Extend manifold learning to applications other than embedding

marv
Download Presentation

Non-Isometric Manifold Learning Analysis and an Algorithm

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Non-Isometric Manifold Learning Analysis and an Algorithm Piotr Dollár, Vincent Rabaud, Serge Belongie University of California, San Diego

  2. I. Motivation • Extend manifold learning to applications other than embedding • Establish notion of test error and generalization for manifold learning

  3. Linear Manifolds (subspaces) Typical operations: • project onto subspace • distance to subspace • distance between points • generalize to unseen regions

  4. Non-Linear Manifolds Desired operations: • project onto manifold • distance to manifold • geodesic distance • generalize to unseen regions

  5. II. Locally Smooth Manifold Learning (LSML) Represent manifold by its tangent space Non-local Manifold Tangent Learning [Bengio et al. NIPS05] Learning to Traverse Image Manifolds [Dollar et al. NIPS06]

  6. Learning the tangent space Data on d dim. manifold in D dim. space

  7. Learning the tangent space Learn function from point to tangent basis

  8. Loss function

  9. Optimization procedure

  10. III. Analyzing Manifold Learning Methods • Need evaluation methodology to: • objectively compare methods • control model complexity • extend to non-isometric manifolds

  11. Evaluation metric Applicable for manifolds that can be densely sampled

  12. Finite Sample Performance Performance: LSML > ISOMAP > MVU >> LLE Applicability: LSML >> LLE > MVU > ISOMAP

  13. Model Complexity All methods have at least one parameter: k Bias-Variance tradeoff

  14. LSML test error

  15. LSML test error

  16. IV. Using the Tangent Space • projection • manifold de-noising • geodesic distance • generalization

  17. Projection x' is the projection of x onto a manifold if it satisfies: gradient descent is performed after initializing the projection x' to some point on the manifold: tangents at projection matrix

  18. Manifold de-noising Goal: recover points on manifold ( ) from points corrupted by noise ( ) - original - de-noised

  19. Geodesic distance Shortest path: On manifold: gradient descent

  20. Geodesic distance  embedding of sparse and structured data can apply to non-isometric manifolds 

  21. Generalization Reconstruction within the support of training data Generalization beyond support of training data 

  22. Thank you!

More Related