Cornac Versions Save

New models

• NARRE: Neural Attentional Rating Regression with Review-level Explanations (#356)
• BiVAECF: Bilateral Variational Autoencoder for Collaborative Filtering (#375)

Improvements

• Update SKMeans to account for max_iter and seed (#377)
• Add SKMeans example (#378)
• Add function to generate gamma variates (#379)

New features

• StratifiedSplit evaluation method (#367)

New changes

• Update link to download the Film Trust dataset (#355)
• Add review-level for review modality (#353)
• Add Amazon Digital Music dataset including ratings and reviews (#357)
• Fix rank() function to work with arbitrary item_indices (#363)

New changes

• Fix ConvMF loss info (#349)

New changes

• Add Review Modality (#344)
• Conditional VAE for Collaborative Filtering (CVAECF) model (#345)
• Update init args and fix dimension mismatch in ConvMF model (#348)

New improvements

• Fix bug of Z dims in SoRec (#340)
• Add a reference to topic model in CTR (#341)

New features

• Add Mean Average Precision (MAP) metric (#338)

New improvements

• Fix bug in graph modality (#333)
• Update models to support clone function (#334)
• Add lambda_reg argument for NMF model (#335)
• Use tqdm.auto for multiple environment compatibility (#336)

New improvements

• Improve the efficiency of KNN methods (#331)
• Use tqdm.auto for compatibility on Jupyter notebook env (#332)

New models

• MTER model is faster with Cython (#320)
• Neighborhood-based methods (UserKNN and ItemKNN) (#329)