Non-smooth setting of stochastic decentralized convex optimization problem over time-varying Graphs

Lobanov, Aleksandr; Veprikov, Andrew; Konin, Georgiy; Beznosikov, Aleksandr; Gasnikov, Alexander; Kovalev, Dmitry

doi:10.1007/s10287-023-00479-7

Search citation statements

Order By: Relevance

Paper Sections

Select...

Citation Types

Supporting

Mentioning

Contrasting

Year Published

2023

2024

Publication Types

Select...

Article2

Book1

Relationship

Self Cite1

Independent2

Authors

Journals

Cited by 3 publications

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

Accelerated Zero-Order SGD Method for Solving the Black Box Optimization Problem Under “Overparametrization” Condition

Lobanov,

Gasnikov

2023

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Frequently, the burgeoning field of black-box optimization encounters challenges due to a limited understanding of the mechanisms of the objective function. In this paper, we provide the new concept of the "Order Oracle" as a novel approach to solving such problems. The Order Oracle offers a unique perspective, using only access to the order between function values (possibly with some bounded noise), but without assuming access to their values. As theoretical results, we provide estimates of the convergence rates of the algorithms (obtained by integrating the Order Oracle into existing optimization "tools") in the non-convex, convex, and strongly convex settings. Our theoretical results demonstrate the effectiveness of the Order Oracle through numerical experiments. Finally, we show the possibility of accelerating the convergence of such algorithms.

show abstract