Stable release

Main changes:

• Support for Julia v0.4 and precompilation
• Saving latent vectors and beta to disk
• Included movielens data
• Computation of standard deviation of the predictions

Parallel CG

Major changes:

• Parallel sparse matrix with Hilbert curve sorting
• Parallel CG
• own NormalWishart code

Parallel sampling

Main new features:

• parallel sampling of latent variables
• general speed improvements
• parallel sparse feature matrix
• direct solve for F'F if medium dimensional (6500 or lower)
• full matrix(tensor) prediction

lambda_beta sampling

Main changes

• lambda_beta sampling for entities
• improved print layout
• support for Julia v0.4
• ability to predict whole matrix

minor improvement

• improved tests for alpha sampling
• support for relation features

modified beta sampling

Changes:

• support for different beta sampling
• fixed bugs

tensor support

Main changes:

• support multi-way relationships (tensors)
• support multiple relationships for an entity

alpha sampling

Main changes

• moved alpha to relation model
• alpha can be sampled from data

One relation factorization

Initial release:

1. support for one relation factorization,
2. input can be sparse matrix or dataframe