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: On the Convergence of SGD with Biased Gradients

Abstract: We analyze the complexity of biased stochastic gradient methods (SGD), where individual updates are corrupted by deterministic, i.e. biased error terms. We derive convergence results for smooth (non-convex) functions and give improved rates under the Polyak-Lojasiewicz condition. We quantify how the magnitude of the bias impacts the attainable accuracy and the convergence rates (sometimes leading to divergence).
Our framework covers many applications where either only biased gradient updates are available, or preferred, over unbiased ones for performance reasons. For instance, in the domain of distributed learning, biased gradient compression techniques such as top-k compression have been proposed as a tool to alleviate the communication bottleneck and in derivative-free optimization, only biased gradient estimators can be queried. We discuss a few guiding examples that show the broad applicability of our analysis.
Comments: Accepted to ICML 2020 Workshop "Beyond First Order Methods in ML Systems", updated 2021
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
Cite as: arXiv:2008.00051 [cs.LG]
  (or arXiv:2008.00051v2 [cs.LG] for this version)

Submission history

From: Sebastian U. Stich [view email]
[v1] Fri, 31 Jul 2020 19:37:59 GMT (36kb)
[v2] Sun, 9 May 2021 19:49:46 GMT (4761kb,D)

Link back to: arXiv, form interface, contact.