We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Deep Autoencoders: From Understanding to Generalization Guarantees

Abstract: A big mystery in deep learning continues to be the ability of methods to generalize when the number of model parameters is larger than the number of training examples. In this work, we take a step towards a better understanding of the underlying phenomena of Deep Autoencoders (AEs), a mainstream deep learning solution for learning compressed, interpretable, and structured data representations. In particular, we interpret how AEs approximate the data manifold by exploiting their continuous piecewise affine structure. Our reformulation of AEs provides new insights into their mapping, reconstruction guarantees, as well as an interpretation of commonly used regularization techniques. We leverage these findings to derive two new regularizations that enable AEs to capture the inherent symmetry in the data. Our regularizations leverage recent advances in the group of transformation learning to enable AEs to better approximate the data manifold without explicitly defining the group underlying the manifold. Under the assumption that the symmetry of the data can be explained by a Lie group, we prove that the regularizations ensure the generalization of the corresponding AEs. A range of experimental evaluations demonstrate that our methods outperform other state-of-the-art regularization techniques.
Subjects: Machine Learning (cs.LG); Group Theory (math.GR); Machine Learning (stat.ML)
Journal reference: R. Cosentino, R. Balestriero, R. Baraniuk, B. Aazhang, 2nd Annual Conference on Mathematical and Scientific Machine Learning (2021)
Cite as: arXiv:2009.09525 [cs.LG]
  (or arXiv:2009.09525v3 [cs.LG] for this version)

Submission history

From: Romain Cosentino Mr [view email]
[v1] Sun, 20 Sep 2020 21:01:18 GMT (1505kb,D)
[v2] Fri, 29 Jan 2021 02:15:43 GMT (2675kb,D)
[v3] Wed, 24 Nov 2021 18:43:46 GMT (2850kb,D)

Link back to: arXiv, form interface, contact.