About this document

GADMM: Efficient Distributed Learning by shreyas7776 is a document available to read on EtoBox.

The paper presents GADMM, a decentralized framework for distributed machine learning that reduces communication costs by allowing workers to only communicate with two neighboring workers. GADMM is shown to converge to optimal solutions for convex loss functions and outperforms existing algorithms in terms of communication efficiency in linear and logistic regression tasks. Additionally, a variant called Dynamic GADMM (D-GADMM) is introduced to adapt to dynamic network topologies, improving convergence speed

Author

shreyas7776

Language

EN