Ramp - Rapid Machine Learning prototyping in Python

jlipps · on Nov 28, 2012

I'm doing a machine-learning project for my company (Sauce Labs) and decided to give Ramp a try. Despite some effort put into understanding the API and digging into code to ensure I was passing in all the right kinds of parameters when I hit random errors, I think it was worth it. I was able to test and compare 20 different machine learning algorithms and even more feature sets really easily. I'd definitely recommend checking it out!

benmanns · on Nov 28, 2012

I don't know if it's this package or just Python in general, but I'm having a really hard time getting things up and running with the Kaggle insults example. To get going I ran

  pip install numpy
  pip install pandas
  pip install scikit-learn
  sudo apt-get install libyaml-dev
  pip install nltk
  pip install gensim
  pip install numexpr
  pip install Cython
  sudo apt-get install libhdf5-serial-dev
  pip install h5py
  pip install tables
  pip install ramp

But I'm still running into some numpy issue with the cross-validation loop (`for config in factory:`).

synparb · on Nov 29, 2012

Most of those packages, except gensim are available as part of the free community edition of Anaconda:

http://docs.continuum.io/anaconda/1.2/pkgs.html

https://store.continuum.io/cshop/anaconda

Getting a large chunk of the python scientific stack in one fell swoop is why I often use Anaconda or EPD.

achompas · on Nov 29, 2012

If you're on Mac, the Scipy Superpack is really the way to go for scientific Python:

http://fonnesbeck.github.com/ScipySuperpack/