There's a good breakdown of the algo used at Amazon here. Useful for anyone looking at decent sized data sets:

http://www.cs.umd.edu/~samir/498/Amazon-Recommendations.pdf