Distributed algorithms

KAUST Workshop on Distributed Training in the Era of Large Models

-

Auditorium between B4 & 5, L0, R0215

Join leading researchers and innovators to explore how distributed training is reshaping the next generation of large-scale AI models.

BiCoLoR: Communication-Efficient Optimization with Bidirectional Compression and Local Training

-

B9 L2 R2325

We introduce BiCoLoR, the first algorithm to combine local training with bidirectional compression using arbitrary unbiased compressors, achieving accelerated complexity and demonstrating superior empirical performance.

Distributed Algorithms for Computing a Common Fixed Point of a Group of Nonexpansive Operators

Dr. Xiuxian Li, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore

-

B9 L2 H1

This talk is concerned with the problem of seeking a common fixed point for a finite collection of nonexpansive operators over time-varying multi-agent networks in real Hilbert spaces. Each operator is assumed to be only privately and approximately known to each individual agent, and all agents need to cooperate to solve this problem by local communications over time-varying networks. To handle this problem, inspired by the centralized inexact Krasnosel’ski˘ı-Mann (IKM) iteration, two distributed algorithms, called distributed inexact Krasnosel’ski˘ı-Mann (D-IKM) iteration and distributed inexact block-coordinate Krasnosel’ski˘ı-Mann (D-IBKM) iteration, are proposed. It is shown that the two algorithms can converge weakly to a common fixed point of the family of nonexpansive operators.

Oussama Dhifallah

MS Degree, Computer, Electrical and Mathematical Sciences and Engineering , joined Harvard University as a PhD student.

Laurent Condat

Senior Research Scientist, Computer Science
I am a senior researcher working in optimization, in the group of Prof. Peter Richtárik.