2012/5/11 Mathieu Blondel <[email protected]>:
> All algorithms which supports a warm_start constructor option should also be
> usable similarly to partial_fit. For example:
>
> from sklearn.linear_model import Lasso
>
> clf = Lasso(warm_start=True)
> clf.fit(X_subset1, y_subset1)
> clf.fit(X_subset2, y_subset2)
> ...
>
> Another idea is to learn a different classifier on each subset and use a
> mixture of the classifiers. As a mixture weight, a simple choice is 1 /
> n_mixtures.
>
> clf = LinearSVC()
> clf.fit(X_subset1, y_subset1)
> clf2 = LinearSVC()
> clf2.fit(X_subset2, y_subset2)
> clf.coef_ += clf.coef_
> ...
> clf.coef_ /= n_mixtures

Unfortunately I don't think you can assign coef_ on liblinear wrapper
models due to internal memory layout constraints.

-- 
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to