Recommender System

IPython Notebooks

• Recommender System: online viwer, source

Survive in The Practical Data Analytics World

Two critical abilities:

• Build algorithms using fundamental routines. [Our LSH, K-means, and PCA tutorials]
  • Basic math skill. Can parse research works and understand the algorithms.
  • Familiar with building blocks, e.g. numpy, scipy, networkx, pandas, ...
• Use existing package. [this tutorial]
  • Survey and quickly evaluate a package.
  • Find the documentation and find the community.
  • Learn from examples.
  • Tap into the internals and be able to upgrade when necessary.

Different Ways to Install Python Packages (optional)

• System distributions
  • apt-get install for Debian/Ubuntu
  • yum for RedHat/CentOS
  • brew for MacOS (non-official but de facto package manager)
• From PyPI
  • pip install (recommended)
  • easy_install (traditional and less used now)
• From source repository: [this tutorial]
  • Download the package
  • python setup.py install. --user is a good option.
  • Useful to install latest work-in-progress packages.
• virtualenv
  • Useful when you need multiple Python versions
  • Useful when you need multiple package versions for different projects

Outcome of This Tutorial

• Have a clear idea of the setup and objective of a recommender system.
• Try neighbourhood based methods and model-based methods.
• Learn to use existing packages.
• Learn to modify existing packages when in need.