Yes, I’m using it. Here’s an example. I also posted my own code yesterday; you
can find it in the archives.
http://scikit-learn.org/dev/auto_examples/ensemble/plot_ensemble_oob.html
It works well with n_jobs > 0; ignore the comment in the example. Examining the
code in forest.py, using warm_start=Trure means the incremental number of
additional trees are fit.
For large datasets (400,000 cases, 3,500 features), I’m experiencing
bottlenecks before and after the trees are actually built.
https://github.com/scikit-learn/scikit-learn/issues/4898
The OOB error versus the number of trees as an evaluation tool is mentioned in
the paper
http://projecteuclid.org/euclid.ssu/1257431567
Free for download.
Dale Smith, Ph.D.
Data Scientist
[http://host.msgapp.com/Extranet/96621/Signature%20Images/sig%20logo.png]<http://nexidia.com/>
d. 404.495.7220 x 4008 f. 404.795.7221
Nexidia Corporate | 3565 Piedmont Road, Building Two, Suite 400 | Atlanta, GA
30305
[http://host.msgapp.com/Extranet/96621/Signature%20Images/sig%20Blog.jpeg]<http://blog.nexidia.com/>
[http://host.msgapp.com/Extranet/96621/Signature%20Images/sig%20LinkedIn.jpeg]
<https://www.linkedin.com/company/nexidia>
[http://host.msgapp.com/Extranet/96621/Signature%20Images/sig%20Google.jpeg]
<https://plus.google.com/u/0/107921893643164441840/posts>
[http://host.msgapp.com/Extranet/96621/Signature%20Images/sig%20twitter.jpeg]
<https://twitter.com/Nexidia>
[http://host.msgapp.com/Extranet/96621/Signature%20Images/sig%20Youtube.jpeg]
<https://www.youtube.com/user/NexidiaTV>
From: Rafael Calsaverini [mailto:rafael.calsaver...@gmail.com]
Sent: Thursday, June 25, 2015 11:59 AM
To: scikit-learn-general
Subject: [Scikit-learn-general] Warm_start on Random Forest Classifiers
I saw the new parameter warm_start on the RandomForestClassifier class and was
curious about what is its most common use. I can see two uses for it: (1)
instead of fitting a huge forest in one go, fit it many times in the same data
set and check the improvement in cross validation, (2) fit the data in a whole
new data set and so doing some kind of "online mini-batch Random Forest", which
might be useful for some of my use cases.
The first use is very straightforward and it seems obvious that a forest
trained two times in the same dataset with warm_start=True will have similar
performance with a forest trained only "once" with double n_estimators.
The second case though seem to break a few assumptions of the statistical ideas
behind Random Forests. Have anyone tried it successfully?
Thanks,
Rafael Calsaverini
------------------------------------------------------------------------------
Monitor 25 network devices or servers for free with OpManager!
OpManager is web-based network management software that monitors
network devices and physical & virtual servers, alerts via email & sms
for fault. Monitor 25 devices for free with no restriction. Download now
http://ad.doubleclick.net/ddm/clk/292181274;119417398;o
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
