milk users Google Group

Re: Saving a copy of the model

l...@luispedro.org (Luis Pedro Coelho) — Sun, 05 May 2013 07:20:34 UT

You should be able to pickle it.

HTH
--
Luis Pedro Coelho | EMBL | [link]

Saving a copy of the model

jump4...@gmail.com (Joy M) — Sun, 05 May 2013 02:45:14 UT

Is there a way to keep a copy of a learned model and only apply it (as the
learning takes some time)?
Alternatively, is there a way to save the parameters of a model and
(quickly) create another model using those parameters?
Thank you,
Joy

Milk Unsupervised KMeans Clustering tuning

nagarajansan...@gmail.com (Sankar) — Tue, 23 Apr 2013 19:30:27 UT

Dear Sir,
I have applied the subject algorithm with Mahalanobis distance for a Kaggle
challenge that typically requires applying unsupervised methods.
I have used max iteration up to 1000 to start with.
Are there any specific parameters that could yield a better performance ?
for instance consideration on the max iteration based on the size of data?

ANN: Milk 0.5.2

l...@luispedro.org (Luis Pedro Coelho) — Wed, 10 Apr 2013 11:34:38 UT

Hello everyone,

I have now regained access to my PyPI account and released milk 0.5.2 there.

This is purely a bug fix release for the way that eigen was packaged. It should solve many compilation problems, though.

LINKS
~~~~~~

Documentation: [link]
Mailing List: [link]

Re: Milk Installation issue on Ubuntu 11.10 and 12.04

nagarajansan...@gmail.com (Sankar) — Fri, 22 Mar 2013 14:47:11 UT

Thank you. Resolved.
Sankar

Re: Milk Installation issue on Ubuntu 11.10 and 12.04

l...@luispedro.org (Luis Pedro Coelho) — Fri, 22 Mar 2013 10:43:18 UT

Hi,

On Ubuntu, you should run

sudo apt-get install python-numpy python-scipy libeigen3-dev

to get all dependencies.

HTH
--
Luis Pedro Coelho

LxMLS 2013: [link]

Milk Installation issue on Ubuntu 11.10 and 12.04

nagarajansan...@gmail.com (Sankar) — Fri, 22 Mar 2013 07:31:06 UT

Sir,
I have [compile] problems when installing Milk on a Ubuntu 11.10 and then
on a 12.04 server
Here is the output I get while installing (I tried both using pip and then
downloading the milk archive and using setup.py)
Appreciate your inputs
Thanks
Sankar
......
building extension "milk.supervised._svm" sources

Re: Milk -Random Forest - Applying test data to predict

l...@luispedro.org (Luis Pedro Coelho) — Thu, 21 Mar 2013 17:30:12 UT

Hi Sankar,

learner = …
model = learner.train(training_feature s, labels)
predictions = learner.apply_many(testing_fea tures)

works with any milk learner.

--
Luis Pedro Coelho

LxMLS 2013: [link]

Milk -Random Forest - Applying test data to predict

nagarajansan...@gmail.com (Reji) — Thu, 21 Mar 2013 17:27:13 UT

Sir,
'
Given a training data file and a test data (typically supplied in Kaggle
competitions), Would like to know how to pass the given ' test data' to
predict after training a model on RF using Milk-RF . I believe
Cross-Validation APIs with the folding use part of the training data as
test data ?

Re: Can't compile on OS X Lion

l...@luispedro.org (Luis Pedro Coelho) — Thu, 07 Mar 2013 18:21:40 UT

I have fixed an important issue in the git version.

Unfortunately, I lost my access to PyPI so I cannot upload it :( Once this gets sorted out, I will upload it immediately.

In the meanwhile, you can try downloading from github:

[link]

Re: Can't compile on OS X Lion

i.mulv...@elifesciences.org (Ian Mulvany) — Wed, 06 Mar 2013 17:15:41 UT

In addition I have tried installing on ubuntu via Vagrant as a VM and I get
the following error:
error: Command "g++ -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O2
-Wall -fPIC -I/usr/lib/python2.6/dist-pack ages/numpy/core/include
-I/usr/include/python2.6 -c milk/supervised/_perceptron.cp p -o

Re: Can't compile on OS X Lion

i.mulv...@elifesciences.org (Ian Mulvany) — Wed, 06 Mar 2013 17:13:37 UT

I've just tried the suggestion of switching the compiler, and I've tried
against the bare git repo, as well as the modified version of milk linked
to in this thread, and I had no success. The fatal error is:
^
1 error generated.
milk/supervised/_lasso.cpp:10: 10: fatal error: 'eigen3/Eigen/Dense' file

Lasso compile error

g.h.ba...@gmail.com (Gökhan Hasan BakIr) — Sun, 03 Mar 2013 09:29:17 UT

There seems to be some error in the compile setup.
compile options:
'-I/Library/Python/2.7/site-pa ckages/numpy-1.8.0.dev_9597b1f _20120920-py2.7-macosx-10.8-x8 6_64.egg/numpy/core/include
-I/System/Library/Frameworks/P ython.framework/Versions/2.7/i nclude/python2.7
-c'
extra options: '-std=c++0x -stdlib=libc++'

Re: Can't compile on OS X Lion

m...@michaelchughes.com (Mike Hughes) — Thu, 17 Jan 2013 06:42:34 UT

I'm having similar trouble as Eric. Lots of trouble building and installing
on OS X Lion, but I finally found a partial workaround.
When I try to compile milk out-of-the-box, I get this error:
unrecognized command line option '-stdlib=libc++'
This extra compiler argument doesn't seem to be supported by my version of

Re: Can't compile on OS X Lion

n...@cape.com (Norman Vine) — Wed, 09 Jan 2013 02:27:07 UT

My previous post were the results from running inside the milk build directory

running after 'setup.py install' things are much better

eg imports don't fail

[link]

In [1]: from milk import tests

In [2]: tests.run()

…….

============================== ============================== ==========