Distributed Mini-Batch SDCA

Martin Takáč; Peter Richtárik; Nathan Srebro

Distributed Mini-Batch SDCA

Martin Takáč, Peter Richtárik, Nathan Srebro

School of Mathematics

Research output: Working paper

Abstract

We present an improved analysis of mini-batched stochastic dual coordinate ascent for regularized empirical loss minimization (i.e. SVM and SVM-type objectives). Our analysis allows for flexible sampling schemes, including where data is distribute across machines, and combines a dependence on the smoothness of the loss and/or the data spread (measured through the spectral norm).

Original language	English
Publisher	ArXiv
Publication status	Published - 29 Jul 2015

Keywords / Materials (for Non-textual outputs)

cs.LG
math.OC

Access to Document

1507.08322v1

Cite this

@techreport{04df786e99be4605b085063c8e2b183d,

title = "Distributed Mini-Batch SDCA",

abstract = "We present an improved analysis of mini-batched stochastic dual coordinate ascent for regularized empirical loss minimization (i.e. SVM and SVM-type objectives). Our analysis allows for flexible sampling schemes, including where data is distribute across machines, and combines a dependence on the smoothness of the loss and/or the data spread (measured through the spectral norm).",

keywords = "cs.LG, math.OC",

author = "Martin Tak{\'a}{\v c} and Peter Richt{\'a}rik and Nathan Srebro",

year = "2015",

month = jul,

day = "29",

language = "English",

publisher = "ArXiv",

type = "WorkingPaper",

institution = "ArXiv",

}

TY - UNPB

T1 - Distributed Mini-Batch SDCA

AU - Takáč, Martin

AU - Richtárik, Peter

AU - Srebro, Nathan

PY - 2015/7/29

Y1 - 2015/7/29

N2 - We present an improved analysis of mini-batched stochastic dual coordinate ascent for regularized empirical loss minimization (i.e. SVM and SVM-type objectives). Our analysis allows for flexible sampling schemes, including where data is distribute across machines, and combines a dependence on the smoothness of the loss and/or the data spread (measured through the spectral norm).

AB - We present an improved analysis of mini-batched stochastic dual coordinate ascent for regularized empirical loss minimization (i.e. SVM and SVM-type objectives). Our analysis allows for flexible sampling schemes, including where data is distribute across machines, and combines a dependence on the smoothness of the loss and/or the data spread (measured through the spectral norm).

KW - cs.LG

KW - math.OC

M3 - Working paper

BT - Distributed Mini-Batch SDCA

PB - ArXiv

ER -

Distributed Mini-Batch SDCA

Abstract

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Cite this