Pith. sign in

REVIEW 4 cited by

A simple uniformly optimal method without line search for convex optimization

Not yet reviewed by Pith; the record is open.

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 2310.10082 v3 pith:6USCNYDQ submitted 2023-10-16 math.OC cs.LG

classification math.OCcs.LG
keywords convexoptimizationlineoptimalsearchac-fgmconvergenceproblem
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

Line search (or backtracking) procedures have been widely employed into first-order methods for solving convex optimization problems, especially those with unknown problem parameters (e.g., Lipschitz constant). In this paper, we show that line search is superfluous in attaining the optimal rate of convergence for solving a convex optimization problem whose parameters are not given a priori. In particular, we present a novel accelerated gradient descent type algorithm called auto-conditioned fast gradient method (AC-FGM) that can achieve an optimal $\mathcal{O}(1/k^2)$ rate of convergence for smooth convex optimization without requiring the estimate of a global Lipschitz constant or the employment of line search procedures. We then extend AC-FGM to solve convex optimization problems with H\"{o}lder continuous gradients and show that it automatically achieves the optimal rates of convergence uniformly for all problem classes with the desired accuracy of the solution as the only input. Finally, we report some encouraging numerical results that demonstrate the advantages of AC-FGM over the previously developed parameter-free methods for convex optimization.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

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

  1. Nesterov Finds GRAAL: Optimal and Adaptive Gradient Method for Convex Optimization

    math.OC 2025-07 conditional novelty 7.0 of 10

    Accelerated GRAAL is the first adaptive first-order method that proves near-optimal accelerated complexity for convex L-smooth and (L0,L1)-smooth functions with geometric stepsize growth.

  2. Gradient Methods with Online Scaling Part I. Theoretical Foundations

    math.OC 2025-05 conditional novelty 7.0 of 10

    Online scaled gradient methods adapt matrix step sizes via online learning, match the best fixed step size asymptotically, and achieve non-asymptotic superlinear convergence on smooth strongly convex problems.

  3. Online Learning-guided Learning Rate Adaptation via Gradient Alignment

    cs.LG 2025-06 conditional novelty 6.0 of 10

    GALA adapts the learning rate through online learning on gradient alignment, achieves a data-adaptive convergence rate for normalized SGD, and shows robust empirical performance across initial learning rates.

  4. Doubly Smoothed Optimistic Gradients: A Universal Approach for Smooth Minimax Problems

    math.OC 2025-06 reject novelty 6.0 of 10

    DS-OGDA is a single-loop first-order method claimed to handle a broad class of smooth minimax problems with a universal step-size schedule and best-known iteration complexity in each subclass.

Pith tools