In this section you will find EURA NOVA’s latest news and activities. In this section you will find EURA NOVA's scientific publications and reports.
Recommendation systems (RS)s play an important role in the online lives of people as they personalize the user experiment and help to find relevant items. The challenges exhibited by RSs are the data noise and the data sparsity. We propose a multi-modal GAN-based recommendation engine, called RecoGAN, that learns user representation according to multi-modal item representation. Experiments on two multi-modal datasets demonstrate the effectiveness of RecoGAN and its competitiveness with state-of-the-art information retrieval frameworks. RecoGAN brings an improvement of up to 25% in all metrics compared to the closest state-of-the-art frameworks.
Hamza Frigui, Amine Ghrab, Ichraf Lahouli, RecoGAN: Multimodal Recommendation Engine based on AutoEncoders and Generative Adversarial Networks, Proc. of IntelliSys 2021.
The final paper will be published after the conference.