Stochastic Relational Models (SRM)
· Short
Introduction:
SRM is a Gaussian process model to predict the links between entities, based on
entity attributes (if given) and some observed links. In its simplest form, the model can be seen as
learning from an partially observed relational matrix T, where T(i, j) means a
relationship between entity Vi and entity Uj. After training, the model can
predict the missing entries in this matrix T.
· Stochastic
Relational Models for Discriminative Link Prediction
Kai Yu, Wei
Advances in Neural
Information Processing Systems 19 (NIPS), MIT Press, 2007
· Data
used in the experiments (in Matlab 7 format):
synthetic data (3KB), movie data
(1,435KB, including all the 10 training/testing splits )