REVIEW 3 cited by
Better Theory for SGD in the Nonconvex World
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
abstract
Large-scale nonconvex optimization problems are ubiquitous in modern machine learning, and among practitioners interested in solving them, Stochastic Gradient Descent (SGD) reigns supreme. We revisit the analysis of SGD in the nonconvex setting and propose a new variant of the recently introduced expected smoothness assumption which governs the behaviour of the second moment of the stochastic gradient. We show that our assumption is both more general and more reasonable than assumptions made in all prior work. Moreover, our results yield the optimal $\mathcal{O}(\varepsilon^{-4})$ rate for finding a stationary point of nonconvex smooth functions, and recover the optimal $\mathcal{O}(\varepsilon^{-1})$ rate for finding a global solution if the Polyak-{\L}ojasiewicz condition is satisfied. We compare against convergence rates under convexity and prove a theorem on the convergence of SGD under Quadratic Functional Growth and convexity, which might be of independent interest. Moreover, we perform our analysis in a framework which allows for a detailed study of the effects of a wide array of sampling strategies and minibatch sizes for finite-sum optimization problems. We corroborate our theoretical results with experiments on real and synthetic data.
Forward citations
Cited by 3 Pith papers
-
Convergence of Momentum-Based Optimization Algorithms with Time-Varying Parameters
A unified momentum-based optimization algorithm with time-varying parameters is shown to converge almost surely under generalized Robbins-Monro and Kiefer-Wolfowitz-Blum conditions, even with biased, unbounded-varianc...
-
Topology-Aware Block Coordinate Descent for Qubit Frequency Allocation of Superconducting Quantum Processors
The Snake optimizer is identified with block coordinate descent, and a nearest-neighbor heuristic over a sequence-dependent TSP is proposed to order blocks and reduce calibration cost; the SD-TSP cost is not specified.
-
Strategies for Improving Communication Efficiency in Distributed and Federated Learning: Compression, Local Training, and Personalization
A PhD dissertation showing unified compression theory, personalized accelerated local training, and pruning methods that reduce communication costs in federated learning and maintain accuracy in LLM pruning.
Discussion (0). Continue with ORCID to comment.