8 expansion sub packages! Recbole launches 2.0!

Easy to use and powerful recommendation algorithm framework Bole (RecBole) Once again has released a new version ！

  In Bole 1.0 On the basis of , In this update, we launch RecBole2.0, Include  8  Latest expansion packs ！ Covering the latest topics and directions of multiple recommendation systems from data to models ！ It provides an easy-to-use and powerful tool library for the research of multiple fields of recommendation system ！

As an owner from data processing 、 Model development 、 A one-stop whole process hosting framework from algorithm training to scientific evaluation , Bole has been officially released since the first day , As of today , It has been officially released for operation for about 19 Months . in the meantime , Community activity has been rising steadily , The number of users is growing , our github repro We've got it 1.9k individual star and 364 individual fork. On this basis , Our team continues to expand and update it , Expand from the perspective of data and model , For different research directions , today RecBole Launch eight extension kits ：

X

     Data oriented , We focus on three important research topics ： Data sparsity 、 Data deviation and data distribution offset , For these three data problems , We have developed five benchmark toolkits , They correspond to each other Meta learning (RecBole-MetaRec), Data to enhance (RecBole-DA), Depolarization (RecBole-Debias), Fairness (RecBole-FairRec) and Cross domain recommendation (RecBole-CDR).

     Model oriented , We consider providing more support for recommendation algorithms based on emerging model architectures , Two benchmark toolkits have been developed , Respectively be based on transformer Model of (RecBole-TRM) and Model based on graph neural network (RecBole-GNN), besides , We aim at Person post match Developed an application toolkit (RecBole-PJF).

    We have developed a total of 65 A recommended system model , And all the corresponding data are provided 、 Model and evaluation interfaces , It is a great expansion for Bole . After this expansion , Bole has become a more comprehensive algorithm library of recommendation system , Total coverage 130+ Recommend system models and 11 Different tasks .

• Address of thesis ：

• https://arxiv.org/pdf/2206.07351

• Project home address ：

• https://recbole.io

• The first phase of github( master station )：

• https://github.com/RUCAIBox/RecBole

• Phase two github( This release )：

• https://github.com/RUCAIBox/RecBole2.0