Hi everyone,
I have been reading a few papers about using (penalized) linear regression
to recover networks from noisy biological data, and I thought they would
make a very useful addition to sklearn. In particular, there's a few
really interesting techniques described in this paper:
http://www.sciencedirect.com/science/article/pii/S0005109811001075
1) The ability to specify ahead of time the expected sign of the
coefficients.
2) The ability to tweak the coefficients recovered to obtain a stable
matrix
As far as I can tell, in sklearn, it's possible to specify if all
coefficients are positive, but it's not possible to specify the sign of
individual coefficients. Although the algorithms are described in detail,
the actual implementation is a bit beyond me, since I'm not very familiar
with a few of the results from linear algebra that they use. I wrote to
one of the authors of the article to ask a few questions, and he was
helpful, and shared the MATLAB code used for the article, although he
specified that he only wanted it used for academic purposes.
Assuming I can get permission from the author to use his code, would anyone
else be interested in working together on a PR on these topics? I should
add, even if the author gives permission to share his code, the code is
very low-level matrix manipulation, and would have to be nearly completely
re-written, and have tests/docs added, so it would be a fair bit of work.
Alternatively, I think it's possible to just follow the description in the
paper, and just write it completely from scratch without worrying about any
licensing issues.
Federico
------------------------------------------------------------------------------
Monitor your physical, virtual and cloud infrastructure from a single
web console. Get in-depth insight into apps, servers, databases, vmware,
SAP, cloud infrastructure, etc. Download 30-day Free Trial.
Pricing starts from $795 for 25 servers or applications!
http://p.sf.net/sfu/zoho_dev2dev_nov
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
