Tuning OnlineLogisticRegression Algo

[Salman Mahmood <sa...@influestor.com>]

I have developed few binary models for classification purposes. They work pretty well except in some cases.
For this reason I decided to tune the learning algorithm. I see there are configuration options like 
.alpha()
.stepOffset()
.decayExponent()
.learningRate()

I know one needs to run the classifier hundred or thousands of time to adjust and see what values gives you the best result, but for starter I need to get some idea about what these methods do. Can somebody point me in a right direction?
Thanks.

Re: Tuning OnlineLogisticRegression Algo

[Ted Dunning <te...@gmail.com>]

See this page:  http://leon.bottou.org/research/stochastic

Google is your friend.

This API is, however, not particularly friendly.  Therefore, you will have
to read about the basics and be able to figure these things out from first
principles.  There is some documentation in the code.  You can also read
about the basic principles here:

http://alias-i.com/lingpipe/demos/tutorial/logistic-regression/read-me.html

On Tue, Oct 9, 2012 at 3:27 PM, Salman Mahmood <sa...@influestor.com>wrote:

> I have developed few binary models for classification purposes. They work
> pretty well except in some cases.
> For this reason I decided to tune the learning algorithm. I see there are
> configuration options like
> .alpha()
> .stepOffset()
> .decayExponent()
> .learningRate()
>
> I know one needs to run the classifier hundred or thousands of time to
> adjust and see what values gives you the best result, but for starter I
> need to get some idea about what these methods do. Can somebody point me in
> a right direction?
> Thanks.