We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML
< prev | next >
new | recent | 1612

Change to browse by:

cs
cs.DC
cs.IT
cs.LG
math
stat
stat.CO

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Statistics > Machine Learning

Title: Gradient Coding

Authors: Rashish Tandon, Qi Lei, Alexandros G. Dimakis, Nikos Karampatziakis
(Submitted on 10 Dec 2016 (v1), last revised 8 Mar 2017 (this version, v2))
Abstract: We propose a novel coding theoretic framework for mitigating stragglers in distributed learning. We show how carefully replicating data blocks and coding across gradients can provide tolerance to failures and stragglers for Synchronous Gradient Descent. We implement our schemes in python (using MPI) to run on Amazon EC2, and show how we compare against baseline approaches in running time and generalization error.
Subjects: Machine Learning (stat.ML); Distributed, Parallel, and Cluster Computing (cs.DC); Information Theory (cs.IT); Machine Learning (cs.LG); Computation (stat.CO)
Cite as: arXiv:1612.03301 [stat.ML]
  (or arXiv:1612.03301v2 [stat.ML] for this version)

Submission history

From: Qi Lei [view email]
[v1] Sat, 10 Dec 2016 14:25:00 GMT (77kb,D)
[v2] Wed, 8 Mar 2017 01:00:33 GMT (111kb,D)
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Link back to: arXiv, form interface, contact.