Dataset: x1, x2, x3, ... xm
Is X-test anomalous ?
anomaly detection vs supervised learning
choosing what features to use
non-gaussian features, maybe log(x),sqrt(x),x^2 to gaussian-like features
multivariate G distributon
Don't model p(x1) p(x2) separately.
Modle p(x) all in one go.
collaborative filtering algorithm
finding related movies
smallest|| xi - xj ||
users who have not rated any movies