Introduction

hopwise

hopwise is an advanced extension of the RecBole library, designed to enhance recommendation systems with the power of knowledge graphs. By integrating knowledge embedding models, path-based reasoning methods, and path language modeling approaches, hopwise supports both recommendation and link prediction tasks with a focus on explainability.

It aims to help the researchers to reproduce and develop recommendation models.

In the lastest release, our library includes all the algorithms already present in Recbole [Model List], along with two 🆕 new categories of models and numerous other improvements:

• General Recommendation

• Sequential Recommendation

• Context-aware Recommendation

• Knowledge-based Recommendation

• Path Reasoning based Recommendation

• Knowledge Graph Embeddings for Recommendation and Link prediction

We have also added 4 new datasets in addition to the 44 datasets already available in Recbole [Collected Datasets].

New Features:

• We added 7 Path-Based (some of them from scratch)

• We added 14 knowledge graph embedding methods

• We added 4 new datasets

• We added 12 new metrics covering Beyond-Accuracy and Path-Quality Metrics

• We added the feature to evaluate from a checkpoint

• We added the support for Link Prediction along the recommendation task on KGE

• We added the support for optuna in hyperparameters hyper_tuning

• We added a new data sample feature to sample paths from a knowledge graph

  We also covered the saving of different dataloaders so you don’t need to sample each time new paths

• We added support for uv

• We added support for embeddings visualization through tSNE

  We also prepared a case study to show how to use it inside run_example folder

Get Started

• Install Hopwise
• Distributed Training

User Guide

• Usage
• Config Introduction
• Data Module Introduction
• Tasks & Models
• Training & Evaluation Introduction
• Hyperparameters Tuning Introduction

Architecture

• U Configuration
• Create Datasets
• Create Samplers
• Create DataLoaders
• Create Trainers
• Add Models to hopwise
• Create Metrics
• API Reference

The Team

hopwise is developed and maintained by the `Trustworthy Artificial Intelligence Laboratory @ University of Cagliari`.

License

hopwise uses MIT License <https://github.com/tail-unica/hopwise/blob/main/LICENSE>.