Mark Fuge - markfuge at gmail - 2012 Implementation of a collaborative filtering algorithm in Python that uses stochastic gradient decent to optimize a linear loss function of movie and user attributes, along with L2 regularization. This implementation was created as part of an assignment for UC Berkeley's CS281B Scalable Machine Learning course. I offer no guarentees on the performance or robustness of the implementation, but it may serve as a useful reference implementation for someone in the future.

Data: This uses the MovieLens Dataset as input. You can read about the format of the input data here and then download the specific zip files containing the data here (specifically, this code used the 10M dataset)