Pith. sign in

REVIEW 2 cited by

SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex Composite Objectives

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1407.0202 v3 pith:FNGHSSZZ submitted 2014-07-01 cs.LG math.OCstat.ML

SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex Composite Objectives

Aaron Defazio , Francis Bach (INRIA Paris - Rocquencourt , LIENS , MSR - INRIA) , Simon Lacoste-Julien (INRIA Paris - Rocquencourt This is my paper
classification cs.LG math.OCstat.ML
keywords sagamethodcompositeconvergenceconvexfastgradientincremental
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

In this work we introduce a new optimisation method called SAGA in the spirit of SAG, SDCA, MISO and SVRG, a set of recently proposed incremental gradient algorithms with fast linear convergence rates. SAGA improves on the theory behind SAG and SVRG, with better theoretical convergence rates, and has support for composite objectives where a proximal operator is used on the regulariser. Unlike SDCA, SAGA supports non-strongly convex problems directly, and is adaptive to any inherent strong convexity of the problem. We give experimental results showing the effectiveness of our method.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Variance Matters: Improving Domain Adaptation via Stratified Sampling

    cs.LG 2025-12 unverdicted novelty 6.0

    VaRDASS improves unsupervised domain adaptation by using stratified sampling to reduce variance in discrepancy estimation for measures like correlation alignment and MMD, with derived error bounds, an optimality proof...

  2. Intercomparison of Machine Learning Algorithms for Remote Sensing-based In-season Crop Mapping

    cs.LG 2026-06 unverdicted novelty 4.0

    Year-wise cross-validation across ten ML algorithms on Harmonized Landsat-Sentinel imagery shows SVMs achieve mean F1 of 0.74 for almonds in California and 0.59 for corn in Iowa by early June in unseen validation years.