Thursday 14 November 2013

Software I use for recommender systems

I use python, haskell, postgresql, emacs, gensim, vowpal wabbit, and debian (I don't much about other tools like R/matlab that would be good for recommender)

Potentially useful -
python libraries are - numpy, scipy, scikit-learn, pandas  gensim
unix tools are - awk, sed, vowpal wabbit, waffles machine learning
sql tools are - postgres, sqlite

In open source databases there is pytables, postgres, sqlite, index db's like tokyo, leveldb. Out of these I just use postgres (it's fast enough - index db's and pytables probably aren't worth the difference).

Of machine learning libraries I use gensim and vowpal wabbit. Both are on-line (use little memory) and fast. The others are good and I've tested them a bit (but my desktop only has 500mb).

In languages I use python, awk, sql, and haskell (depending on my mood - they're all really fun to work in). I use awk/wc/grep for simple searches/counts of files, python or haskell to do more complex extractions. SQL is best for the most complex querying and I'm trying to build up a whole bunch of tables for auto-tagging text.

No comments:

Post a Comment