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: Incremental Learning in Diagonal Linear Networks

Abstract: Diagonal linear networks (DLNs) are a toy simplification of artificial neural networks; they consist in a quadratic reparametrization of linear regression inducing a sparse implicit regularization. In this paper, we describe the trajectory of the gradient flow of DLNs in the limit of small initialization. We show that incremental learning is effectively performed in the limit: coordinates are successively activated, while the iterate is the minimizer of the loss constrained to have support on the active coordinates only. This shows that the sparse implicit regularization of DLNs decreases with time. This work is restricted to the underparametrized regime with anti-correlated features for technical reasons.
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC)
Cite as: arXiv:2208.14673 [cs.LG]
  (or arXiv:2208.14673v1 [cs.LG] for this version)

Submission history

From: Raphael Berthier [view email]
[v1] Wed, 31 Aug 2022 08:00:04 GMT (50kb,D)

Link back to: arXiv, form interface, contact.