[scikit-learn] implementing regularized random forest
Hello, I am trying to implement my own regularized random forest (RRF) which grows trees in series and selects new features only if they are better than the features used in previous splits. This is for a research project and I will need to ship the code with the publication. So far I have a working proof of concept where I modified the scikit-learn forest, tree, and splitter modules. But this mean that I need to ship my fork version of scikit-learn. Ideally, I am looking for a way to build my own RRF that uses scikit-learn API instead of modifying it. Is it possible? Thanks. Mickael ___ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Re: [scikit-learn] implementing regularized random forest
Mickael, You probably don't need to ship an entire fork, but all the tree internals that you are using (splitter etc.) are part of a private API so yes, you would need to duplicate these into your own implementation. Nicolas On 11/3/20 4:38 PM, Mick Men wrote: Hello, I am trying to implement my own regularized random forest (RRF) which grows trees in series and selects new features only if they are better than the features used in previous splits. This is for a research project and I will need to ship the code with the publication. So far I have a working proof of concept where I modified the scikit-learn forest, tree, and splitter modules. But this mean that I need to ship my fork version of scikit-learn. Ideally, I am looking for a way to build my own RRF that uses scikit-learn API instead of modifying it. Is it possible? Thanks. Mickael ___ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn ___ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
Re: [scikit-learn] Changes in Travis billing
I took this opportunity to migrate from travis-ci.org to travis-ci.com. The project url is now: https://travis-ci.com/github/scikit-learn/scikit-learn The blog post did mention that we need to ask for a number of build credits. Currently we use travis-ci to test the intel c compiler and ARM. Looking forward, we may be doing more ARM on travis because it is the only platform with native ARM support. As a data point, our cron job that runs scipy-dev, icc-build and ARM takes around 70 minutes to run, (ARM takes ~ 12 minutes). This means with a normal allocation of 1000 minutes we can run our cron job ~ 14 times. So we ask for 3000-4000 minutes? Thomas On Mon, Nov 2, 2020 at 10:59 AM Adrin wrote: > > Shall I contact them? Any other volunteers? > > On Mon, Nov 2, 2020 at 11:51 AM Gael Varoquaux > wrote: >> >> Travis is changing it's billing strategy: >> https://blog.travis-ci.com/2020-11-02-travis-ci-new-billing >> >> Open repositories are getting a free initial set of credit. They invite >> open source projects to contact them to benefit from a more liberal >> policy. >> >> I suggest that we do the latter, as I fear that we might run out of >> credits, and I am quite convinced that we could benefit from the liberal >> policy. >> >> Cheers, >> >> Gaël >> >> -- >> Gael Varoquaux >> Research Director, INRIA Visiting professor, McGill >> http://gael-varoquaux.infohttp://twitter.com/GaelVaroquaux >> ___ >> scikit-learn mailing list >> scikit-learn@python.org >> https://mail.python.org/mailman/listinfo/scikit-learn > > ___ > scikit-learn mailing list > scikit-learn@python.org > https://mail.python.org/mailman/listinfo/scikit-learn ___ scikit-learn mailing list scikit-learn@python.org https://mail.python.org/mailman/listinfo/scikit-learn
