Bio

I am a Research Scientist at Deezer, a French music streaming service operating in 180 countries. Based in Paris, I am a member of the Research team where, since 2018, I have been conducting fundamental and applied research on deep learning, graph mining, music information retrieval, and music recommendation. Since 2023, I have also been coordinating the scientific projects carried out with Deezer engineers responsible for deploying music recommender systems on the service.

From 2018 to 2022, I simultaneously completed a Ph.D. at École Polytechnique, under the supervision of Romain Hennequin and Michalis Vazirgiannis. My doctoral thesis and scientific publications from this period primarily focused on representation learning with graph neural networks, and various applications to music recommendation.

Prior to my Ph.D., I worked as a Data Scientist at Deezer for nearly two years. Closely collaborating with product, tech, and design teams, I conducted product-oriented projects to improve UX and conversion on the service. Even before, I also handled other exciting challenges in Canada and China. Feel free to visit my LinkedIn profile for more details about my experience.

I graduated in 2016 from École Normale Supérieure Paris-Saclay, completing a Master's degree in Mathematics, Computer Vision, and Machine Learning (MVA), and from ENSAE Paris, completing a Master of Engineering in Data Science, both with first-class honours. I am always looking for professionals with common interests to learn and share experiences, so feel free to reach out!

---

News

• October 2024: We are thrilled to be among the “Best Full Paper” candidates at RecSys 2024!
• August 2024: We will present three papers on music recommendation at RecSys 2024 in Bari.
• July 2024: I was fortunate to be invited by VinAI to discuss graph autoencoders and music recommendation in Hanoi. The seminar video is now accessible online.
• June 2024: We presented our workshop paper on vMF-exp at ICML 2024 ARLET in Vienna.
• March 2024: We have won the “Best Industry Paper” award at ECIR 2024 in Glasgow!
• February 2024: I was honored to present my Ph.D. research at Fudan University in Shanghai.

---

Publications

See me also on Google Scholar and ORCID.org.

2024

Note: publications from 2023 and 2024 reflect my Research Coordinator role at Deezer, where I was responsible for supervising projects. I am often listed as the second or last author in these articles.

• Transformers Meet ACT-R: Repeat-Aware and Sequential Listening Session Recommendation
  V.A. Tran, G. Salha-Galvan, B. Sguerra, R. Hennequin
  18th ACM Conference on Recommender Systems (RecSys 2024)
  [PDF] [Code] [Data] Best full paper candidate

• Do Recommender Systems Promote Local Music? A Reproducibility Study Using Music Streaming Data
  K. Matrosova, L. Marey, G. Salha-Galvan, T. Louail, O. Bodini, M. Moussallam
  18th ACM Conference on Recommender Systems (RecSys 2024)
  [PDF] [Code] [Data]

• Let’s Get It Started (Again!): Fostering the Discoverability of New Releases on Deezer
  L. Briand, T. Bontempelli, W. Bendada, M. Morlon, F. Rigaud, B. Chapus, G. Salha-Galvan
  Workshop on Music Recommender Systems
  18th ACM Conference on Recommender Systems (RecSys 2024)
  (Workshop presentation showcasing our ECIR 2024 work)
  [PDF] [Slides]

• vMF-exp: von Mises-Fisher Exploration of Large Action Sets with Hyperspherical Embeddings
  W. Bendada, G. Salha-Galvan, R. Hennequin, T. Bontempelli, T. Bouabça, T. Cazenave
  Workshop on Aligning Reinforcement Learning Experimentalists and Theorists
  41st International Conference on Machine Learning (ICML 2024)
  [PDF] [Code (soon)]

• Let’s Get It Started: Fostering the Discoverability of New Releases on Deezer
  L. Briand, T. Bontempelli, W. Bendada, M. Morlon, F. Rigaud, B. Chapus, G. Salha-Galvan
  46th European Conference on Information Retrieval (ECIR 2024)
  [PDF] [Slides] Best industry track paper award

2023

• Track Mix Generation on Music Streaming Services using Transformers
  W. Bendada, T. Bontempelli, M. Morlon, B. Chapus, T. Cador, T. Bouabça, G. Salha-Galvan
  17th ACM Conference on Recommender Systems (RecSys 2023)
  [PDF] [Blog Post]

• On the Consistency of Average Embeddings for Item Recommendation
  W. Bendada, G. Salha-Galvan, T. Bouabça, T. Cazenave
  17th ACM Conference on Recommender Systems (RecSys 2023)
  [PDF] [Code] [Poster]

• Attention Mixtures for Time-Aware Sequential Recommendation
  V.A. Tran, G. Salha-Galvan, B. Sguerra, R. Hennequin
  46th Int. ACM SIGIR Conf. on Research and Development in Information Retrieval (SIGIR 2023)
  [PDF] [Code]

• A Scalable Framework for Automatic Playlist Continuation on Music Streaming Services
  W. Bendada, G. Salha-Galvan, T. Bouabça, T. Cazenave
  46th Int. ACM SIGIR Conf. on Research and Development in Information Retrieval (SIGIR 2023)
  [PDF] [Code]

2022

• New Frontiers in Graph Autoencoders: Joint Community Detection and Link Prediction
  G. Salha-Galvan, J.F. Lutzeyer, G. Dasoulas, R. Hennequin, M. Vazirgiannis
  Workshop on New Frontiers in Graph Learning
  36th Conference on Neural Information Processing Systems (NeurIPS 2022)
  [PDF] [Code]

• Flow Moods: Recommending Music by Moods on Deezer
  T. Bontempelli, B. Chapus, F. Rigaud, M. Morlon, M. Lorant, G. Salha-Galvan
  16th ACM Conference on Recommender Systems (RecSys 2022)
  [PDF] [Blog Post] [Poster] [Video]

• Contributions to Representation Learning with Graph Autoencoders and Applications to Music Recommendation
  G. Salha-Galvan
  Ph.D. thesis, École Polytechnique, Institut Polytechnique de Paris
  [PDF] [Slides from the defense]

2021

• Modularity-Aware Graph Autoencoders for Joint Community Detection and Link Prediction
  G. Salha-Galvan, J.F. Lutzeyer, G. Dasoulas, R. Hennequin, M. Vazirgiannis
  Neural Networks 153, 474-495, Elsevier (2021 impact factor: 9.657)
  [PDF] [Code]

• Cold Start Similar Artists Ranking with Gravity-Inspired Graph Autoencoders
  G. Salha-Galvan, R. Hennequin, B. Chapus, V.A. Tran, M. Vazirgiannis
  15th ACM Conference on Recommender Systems (RecSys 2021)
  [PDF] [Code] [Data] [Video] Best student paper honorable mention

• Hierarchical Latent Relation Modeling for Collaborative Metric Learning
  V.A. Tran, G. Salha-Galvan, R. Hennequin, M. Moussallam
  15th ACM Conference on Recommender Systems (RecSys 2021)
  [PDF] [Code] [Video]

• A Semi-Personalized System for User Cold Start Recommendation on Music Streaming Apps
  L. Briand, G. Salha-Galvan, W. Bendada, M. Morlon, V.A. Tran
  27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2021)
  [PDF] [Blog Post] [Code] [Data] [Video]

• Modéliser la Perception des Genres Musicaux à travers Différentes Cultures à partir de Ressources Linguistiques
  E.V. Epure, G. Salha-Galvan, M.Moussallam, R. Hennequin
  28ème Conférence sur le Traitement Automatique des Langues Naturelles (TALN 2021)
  (Summary paper presenting our EMNLP 2020 work)
  [PDF] [Code]

2020

Note: I published under the name G. Salha until 2020.

• FastGAE: Scalable Graph Autoencoders with Stochastic Subgraph Decoding
  G. Salha, R. Hennequin, J.B. Remy, M. Moussallam, M. Vazirgiannis
  Neural Networks 142, 1-19, Elsevier (2020 impact factor: 8.05)
  [PDF] [Code]

• Modeling the Music Genre Perception across Language-Bound Cultures
  E.V. Epure, G. Salha, M. Moussallam, R. Hennequin
  2020 Conference on Empirical Methods in Natural Language Processing (EMNLP 2020)
  [PDF] [Code] [Data] [Video]

• Carousel Personalization in Music Streaming Apps with Contextual Bandits
  W. Bendada, G. Salha, T. Bontempelli
  14th ACM Conference on Recommender Systems (RecSys 2020)
  [PDF] [Code] [Data] [Video] Best short paper honorable mention

• Simple and Effective Graph Autoencoders with One-Hop Linear Models
  G. Salha, R. Hennequin, M. Vazirgiannis
  2020 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2020)
  [PDF] [Code]

• Multilingual Music Genre Embeddings for Effective Cross-Lingual Music Item Annotation
  E.V. Epure, G. Salha, R. Hennequin
  21st International Society for Music Information Retrieval Conference (ISMIR 2020)
  [PDF] [Code] [Data]

• Muzeeglot : Annotation Multilingue et Multi-Sources d’Entités Musicales à partir de Représentations de Genres Musicaux
  E.V. Epure, G. Salha, F. Voituret, M. Baranes, R. Hennequin
  27ème Conférence sur le Traitement Automatique des Langues Naturelles (TALN 2020)
  [PDF (in French)] [Slides (in English)] [Code]

2019

• Keep It Simple: Graph Autoencoders Without Graph Convolutional Networks
  G. Salha, R. Hennequin, M. Vazirgiannis
  Workshop on Graph Representation Learning
  33rd Conference on Neural Information Processing Systems (NeurIPS 2019)
  [PDF] [Code]

• Gravity-Inspired Graph Autoencoders for Directed Link Prediction
  G. Salha, S. Limnios, R. Hennequin, V.A. Tran, M. Vazirgiannis
  28th ACM International Conference on Information and Knowledge Management (CIKM 2019)
  [PDF] [Code]

• A Degeneracy Framework for Scalable Graph Autoencoders
  G. Salha, R. Hennequin, V.A. Tran, M. Vazirgiannis
  28th International Joint Conference on Artificial Intelligence (IJCAI 2019)
  [PDF] [Code]

2018

• Adaptive Submodular Influence Maximization with Myopic Feedback
  G. Salha, N. Tziortziotis, M. Vazirgiannis
  2018 IEEE/ACM Int. Conf. on Adv. in Social Networks Analysis and Mining (ASONAM 2018)
  [PDF]