3PC: Three point compressors for communication-efficient distributed training and a better theory for lazy aggregation
We propose and study a new class of gradient communication mechanisms for communication-efficient training -- three point compressors (3PC) -- as well as efficient distributed nonconvex optimization algorithms that can take advantage of them. Unlike most established approaches, which rely on a stati...
Saved in:
Main Authors: | RICHTARIK, Peter, SOKOLOV, Igor, FATKHULLIN, Ilyas, GASANOV, Elnur, LI, Zhize, GORBUNOV, Eduard |
---|---|
Format: | text |
Language: | English |
Published: |
Institutional Knowledge at Singapore Management University
2022
|
Subjects: | |
Online Access: | https://ink.library.smu.edu.sg/sis_research/8685 https://ink.library.smu.edu.sg/context/sis_research/article/9688/viewcontent/ICML22_full_3pc.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Singapore Management University |
Language: | English |
Similar Items
-
MARINA: Faster non-convex distributed learning with compression
by: GORBUNOV, Eduard, et al.
Published: (2021) -
Lazy XML updates: Laziness as a virtue of update and structural join efficiency
by: Catania, B., et al.
Published: (2013) -
Lazy rectangular hybrid automata
by: Agrawal, M., et al.
Published: (2013) -
CANITA: Faster rates for distributed convex optimization with communication compression
by: LI, Zhize, et al.
Published: (2021) -
The discrete time behavior of lazy linear hybrid automata
by: Agrawal, M., et al.
Published: (2013)