Binary Principal Component Analysis in the Netflix Collaborative Filtering Task
Laszlo Kozma, Alexander Ilin and Tapani Raiko
In: 2009 IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2009), 2-4 Sep 2009, Grenoble, France.
We propose an algorithm for binary principal component
analysis (PCA) that scales well to very high dimensional
and very sparse data. Binary PCA finds components from
data assuming Bernoulli distributions for the observations.
The probabilistic approach allows for straightforward treatment
of missing values. An example application is collaborative
filtering using the Netflix data. The results are comparable
with those reported for single methods in the literature
and through blending we are able to improve our previously
obtained best result with PCA.