[Scikit-learn-general] Stochastic dual coordinate aescent solver for linear models

[Ahmed SaadAliden]

Hi,

I am thinking about adding SDCA
<http://www.jmlr.org/papers/volume14/shalev-shwartz13a/shalev-shwartz13a.pdf>
"Stochastic dual coordinate ascent" solver to linear_model,
The results shows a very nice convergence proprieties comparing to SGD. I
am wondering if anybody is already working in adding it to scikit ?

Thanks,
~Ahmed

[Alexandre Gramfort]

hi,

sdca is used internally in liblinear so offered by our logreg estimator.

otherwise it's implemented in lightning :
https://github.com/scikit-learn-contrib/lightning

A


On Sat, Apr 9, 2016 at 8:14 PM, Ahmed SaadAliden

Hi,
I am thinking about adding SDCA "Stochastic dual coordinate ascent" solver
to linear_model,
The results shows a very nice convergence proprieties comparing to SGD. I am
wondering if anybody is already working in adding it to scikit ?
Thanks,
~Ahmed
------------------------------------------------------------------------------
Find and fix application performance issues faster with Applications Manager
Applications Manager provides deep performance insights into multiple tiers
of
your business applications. It resolves application problems quickly and
reduces your MTTR. Get your free trial! http://pubads.g.doubleclick.net/
gampad/clk?id=1444514301&iu=/ca-pub-7940484522588532
_______________________________________________
Scikit-learn-general mailing list
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

[Mathieu Blondel]

And also in LinearSVC with dual=True. The only difference is that the
choice of dual variable is cyclic (with prior permutation) instead of
random.

See this 2008 paper:
http://www.csie.ntu.edu.tw/~cjlin/papers/cddual.pdf

Mathieu

On Sun, Apr 10, 2016 at 9:53 PM, Alexandre Gramfort <

Post by Alexandre Gramfort
hi,
sdca is used internally in liblinear so offered by our logreg estimator.
https://github.com/scikit-learn-contrib/lightning
A
On Sat, Apr 9, 2016 at 8:14 PM, Ahmed SaadAliden

Hi,
I am thinking about adding SDCA "Stochastic dual coordinate ascent"

solver

to linear_model,
The results shows a very nice convergence proprieties comparing to SGD.

I am

wondering if anybody is already working in adding it to scikit ?
Thanks,
~Ahmed

------------------------------------------------------------------------------

Find and fix application performance issues faster with Applications

Manager

Applications Manager provides deep performance insights into multiple

tiers

of
your business applications. It resolves application problems quickly and
reduces your MTTR. Get your free trial! http://pubads.g.doubleclick.net/
gampad/clk?id=1444514301&iu=/ca-pub-7940484522588532
_______________________________________________
Scikit-learn-general mailing list
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

------------------------------------------------------------------------------
Find and fix application performance issues faster with Applications Manager
Applications Manager provides deep performance insights into multiple
tiers of
your business applications. It resolves application problems quickly and
reduces your MTTR. Get your free trial! http://pubads.g.doubleclick.net/
gampad/clk?id=1444514301&iu=/ca-pub-7940484522588532
_______________________________________________
Scikit-learn-general mailing list
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general