Cornac 2.0.0

What's Changed

In this major release, Cornac will support next-item and next-basket recommendation models.

🧱 Models

• Add Deep Neural Network for Temporal Sets Prediction (DNNTSP) Model (#586) @lthoang
• Add Correlation-Sensitive Next-Basket Recommendation (Beacon) Model (#584) @lthoang
• Add Recency Aware Collaborative Filtering for Next Basket Recommendation (UPCF) Model (#587) @lthoang
• Add Session-based Recommendations With Recurrent Neural Networks (GRU4Rec) model (#574) @lthoang
• Fix Session Popular model (#575) @lthoang
• Add RecVAE model (#568) @flywithu
• Add Temporal-Item-Frequency-based User-KNN (TIFUKNN) model (#566) @lthoang

🚀 Features

• Enhance serving evaluation endpoints (#595) @darrylong
• Add column model type into table models (#588) @lthoang
• Add option to evaluate every next item in next-item evaluation (#580) @lthoang
• Add next-item pipeline (#561) @hieuddo
• Make ANNs compatible with Experiment (#563) @tqtg

🐛 Bug Fixes

• [Beacon] max_seq_length is the maximum number of baskets in training sequence (#591) @lthoang

📝 Documentation

• Add missing doc for UPCF model (#589) @lthoang
• Update README.md to use stable branch on readthedocs (#571) @darrylong

✅ Other changes

• Update CircleCI config file (#583) @tqtg
• Add back Circle CI for publishing code coverage (#573) @tqtg

Full Changelog: v1.18.0...v2.0.0

Cornac 1.18.0

What's Changed

Models

• Support ANN search for existing models (#549) @tqtg
• Add Annoy to the list of supported ANN frameworks (#556) @tqtg
• Add Faiss to the list of supported ANN frameworks (#555) @tqtg
• Add ScaNN to the list of supported ANN frameworks (#553) @tqtg
• Add ANN search using HNSWLib (#544) @tqtg
• Add simple model serving (#540) @tqtg
• Add NCF_PyTorch models (#536) @hieuddo

Fixes

• Fix exception check for Cython migration to 3.0 (#558) @tqtg
• PyTorch_NCF fix, normalize running loss (#537) @hieuddo

Other changes

• Rename symbols to ensure consistency (#565) @lthoang
• Add unit test for flask serving (#562) @darrylong
• Make use of verbose setting in Experiment (#564) @tqtg
• Update next-basket evaluation (#559) @lthoang
• Docker version of production serving (#551) @darrylong
• Add clean command (#554) @tqtg
• Update FM supporting message (#550) @tqtg
• Remove exception checking for _dot() (#552) @tqtg
• Add production version of API serving (#547) @darrylong
• Add next-basket recommendation evaluation method (#545) @lthoang
• Add dependency install checks for Cython (#548) @darrylong
• Add PyTorch backend for MF (#546) @hieuddo
• Revamp Documentation (#541) @darrylong
• Add recommend() function for the base class of Recommender (#538) @tqtg
• Clean compiled cpp files (#539) @tqtg

Full Changelog: v1.17.0...v1.18.0

Cornac 1.17.0

What's Changed

New Models

• GCMC Implementation #524
• Add TriRank model #525
• Add LightGCN Model #526
• Add Learn to Rank user Preferences based on Phrase-level sentiment analysis across Multiple categories (LRPPM) model #527
• Implement NGCF #529

Enhancements

• Add transform function for Recommender #522
• Filter positive items for ranking evaluation #523
• Optimize LightGCN Model #531
• Change NGCF Batch Size Param #532

Bug Fixes

• Fix rank function of EFM and ComparERSub models #528
• Fix GCMC Invalid Key Error #533

Cornac 1.16.0

• Add HRDR model (#518)
• Bug fixes and improvements
  • Fix rank function for ComparER_sub (#515, #520)
  • Fix save and load functions for NARRE (#517)
  • Fix typos (#516)
  • Remove unused append function in utils.fast_dict (#521)

Cornac 1.15.4

• Specify numpy version for each Python version for wheel packaging compatibility

Cornac 1.15.3

• Build with Ubuntu 20.04 to be more compatible with Google Colab

Cornac 1.15.2

• Remove CircleCI and AppVeyor
• Support Python [3.8, 3.9, 3.10, 3.11]

Cornac 1.15.1

• Update NCF models to be compatible with TensorFlow 2.0 (#506)
• Remove codecov since it's no longer on PyPI
• Upgrade to ubuntu-20.04 for GH Actions

Cornac 1.15.0

• Add use_bias argument for BPR, WBPR, SBPR models (#473)
• EASEᴿ model implementation (#477)
• Update several models to be compatible with TensorFlow 2.0 (#504)

Cornac 1.14.2

Add mean and std values per metric in CVResult (#467)