Regularization paremeter in Megam

-2

Hi,

I would like to use Megam (Mega Model Optimization Package) for my text classification task as a classifier. I have been told that it implements a regularization parameter. Which regularization parameter does it implement?

asked May 12 '11 at 07:24

Dexter
416●24●34●38

One Answer:

oldest newest most voted

-1

The Gaussian prior is equivalent to a squared L2 regularizer. Hence the following two command line parameters will do:

-lambda <float> specify the precision of the Gaussian prior (default: 1)

-tune tune lambda using repeated optimizations (starts with specified -lambda value and drops by half each time until optimal dev error rate is achieved)

answered May 12 '11 at 08:13

ogrisel
4989●9●55●91

I am calling megam via NLTK. Is there any way to specify these parameters OR to chuck megam and go for Logistic Regression in scikits.learn with a l2 penalty?

(May 12 '11 at 13:06) Dexter

Read the source, Luke :)

(May 13 '11 at 05:07) ogrisel

Subscription:

Once you sign in you will be able to subscribe for any updates here

Tags:

text-classification ×66
regularization ×36
megam ×2

Asked: May 12 '11 at 07:24

Seen: 1,845 times

Last updated: May 13 '11 at 05:07

Regularization paremeter in Megam

Subscription:

Related questions