installation: no module spherical kmeans

[lena2813]

I'm using anaconda, installing in the environment where i have python3.4 version.
I tried downloading via pip and via python setup.py install in this environment and also in a global, python2.7 environment, each time uninstalling everything and trying again. No luck. Everywhere I get a no-module "spherical_kmeans" message. What can be done here? Thanks a lot!

[jasonlaska]

Ah bummer! I haven't tried it in anaconda, only pip (also with virtualenv) in python2.7. One note is that it requires numpy/scipy to be installed ahead of time.

It's possible that you might need to change this line: https://github.com/clara-labs/spherecluster/blob/develop/spherecluster/von_mises_fisher_mixture.py#L26

to be from spherecluster import spherical_kmeans instead (and then try running setup) and possible add spherical_kmeans to the __init__.py in the package.

Let me know if you figure out so I can patch it.

[jasonlaska]

@el2813 I've updated the package to use relative paths; try installing 0.1.2 and see if that helps. Cheers.

[lena2813]

Wow! Thanks a lot for replying! And... I'm amazed by your speed of reply. Awesome^.^

I've installed numpy/scipy way before the spherecluster installation using anaconda's package manager - conda. But what I've noticed (interest thing, by the way), when i pip install spherecluster, it simultaneously pip installs numpy &scipy like doesn`t recognize conda's installs. That's why in the screenshots above you can see that numpy/scipy appears twice.

Also thanks so much!!! for the link. I gonna try it out tonight and let you know if it works.

[lena2813]

yes!!!! it works! Soooo happy. Tnx! I saw awesome mit documentation about your spherical kmeans and von mises-fisher mixture, and I was like I have to try these!!! I gonna do word2vec and then feed that to one of your algorithms. For my current clustering with classical kmeans, preceded by sklearn tfidfvectorizer, I get mediocre results described with metrics below. So I'm super excited to try your package. have a nice day!

[jasonlaska]

Good luck! Results will vary depending how your data is distributed; just another tool to try (among several other clustering algs in sklearn).