This page offers an overview of recent collaborative research projects that the S²A team has contributed to, as well as a glance at our academic and industrial partners, with whom we have produced joint publications.
Ongoing projects
| 2022 - 2025 | 5 partners | 1.9 M€
The AUDIBLE project aims to revolutionise hearable technologies (especially TWS/earbuds) by developing a platform which will enable unprecedented use cases through artificial intelligence (AI) innovations, a highly energy-efficient and powerful DSP and AI processor, and the integration of a miniaturized biometric sensor. Read More »
Industrial Chair to promote research and training in the field of AI, bringing together academic and industrial partners: Télécom Paris, Airbus, Idemia, Renault-Ampère. Read More »
PEPR IA - Foundry | 2024 - 2029 | partners |
The core vision of FOUNDRY is that robustness in AI – a desideratum which has eluded the field since its inception – cannot be achieved by blindly throwing more data and computing power to larger and larger models with exponentially growing energy requirements. Instead, we intend to rethink and develop the core theoretical and methodological foundations of robustness and reliability that are needed to build and instill trust in ML-powered technologies and systems from the ground up. Read More »
ANR Far-See | 2025 - 2029 | partners |
ANR PRCE project, bringing together Télécom Paris and IDEMIA.
ANR CFTextAD | 2024 - 2028 | partners |
The CFTextAD aims at creating a unified framework for textual anomaly detection. We propose combine existing models of natural language representation with anomaly detection methods. We propose to study transfer learning on pre-trained models, and to characterized the detected anomalies, by diversifying our evaluation methods and focusing on interpretable approaches. Our goal is to obtain a tool that will improve the state of the art for various language processing tasks (Mathieu Labeau). Read More »
ANR APDO | 2021 - 2025 | partners |
The project aims at developing primal-dual optimization algorithms with optimal worst-case complexity. More generally, we look at improving existing algorithms for solving saddle point problems, whether they are convex-concave or not (Olivier Fercoq). Read More »
DEESSE | 2024 - 2028 | 3 partners | 157 K€
The scientific ambition of the DEESSE project is to develop multichannel source separation methods capable of generalizing and adapting to real signals recorded by Ambisonics arrays. To achieve this objective, we assume that it is necessary to go beyond the separation paradigm most widely used in the literature, which consists on the one hand of extracting a source signal from a mixture by spatial filtering and/or masking in a transformed domain (fixed or learned), and on the other hand of learning fully-supervised models on synthetic data alone.
AQUARIUS | 2023 - 2026 | 3 partners | 157 K€
Audio quality is an important characteristic that conveys intrinsic information about the process of creating a musical work from recording to studio mastering. Recently, we proposed an innovative method to detect the list of effects applied to a signal and the decade of creation of a work. Thus, the AQUA-RIUS project aims at an exhaustive study of audio quality with a deep learning methodology through 3 scientific issues: 1) audio quality analysis and modelling, 2) audio quality simulation for data augmentation in a machine learning framework to improve the robustness of the trained models and 3) reverse engineering to enable signal restoration and audio quality control. The expertise in signal processing and deep learning of the 3 project partners (IBISC, IRCAM, Telecom Paris) is the main asset of this project. Read More »
BHAI | 2021 - 2025 | partners |
Hybrid Artifical Intelligence for Byzantine Sigillography (Laurence Likforman, in collaboration with Sorbonne University). Read More »
REFAIR ANR/JCJC | 2023 - 2027 | partners |
REFAIR (REvisiting the Foundations of Algorithmic FAIrness for Graphs) is an ANR JCJC project (PI: Charlotte Laclau) that aims to adopt an interdisciplinary to bridging existing works in sociology with concepts from machine learning, and more precisely of representation learning, to better formalize and understand the mechanisms responsible for bias in graphs (Charlotte Laclau, with Gent University).
SAROUMANE ANR/JCJC | 2023 - 2026 | 1 partners | 270 k€
SAROUMANE is an ANR JCJC (Young researcher project) project that aims to tackle speaker diarization problem using well-known explicable Deep Bayesian networks such as heavy-tailed variational autoencoder (Principal investigator: Mathieu FONTAINE). The speaker diarization aims to answer to the question: "who speak and when ?" and is still a major research direction in speech processing with various applications such as automatic speech transcription. The project also includes speaker diarization using, spatial audio information, multiple sensors or linguistic aspects (prosody).
ELIAS aims at establishing Europe as a leader in Artificial Intelligence (AI) research that drives sustainable innovation and economic development. Read More »
Relying on European common standards, the EU-funded OMEGA-X project aims to implement an energy data space. This will include federated infrastructure, data marketplace and service marketplace, involving data sharing between different stakeholders and demonstrating its value for concrete energy use cases while guaranteeing scalability and interoperability with other data space initiatives (Florence d'Alché-Buc and Mathieu Fontaine). Read More »
HI-Audio ERC | 2022 - 2027 | 1 partners | 2,5 M€
Hi-Audio: Hybrid and Interpretable Deep neural audio machines (Principal Investigator: Gaël Richard) is a European Research Council Advanced Grant (AdG) project supported by the European Union’s Horizon 2020 research and innovation program under Grant Agreement-101052978. HI-Audio aims to build controllable and frugal machine listening models based on expressive generative modelling and Hybrid deep learning models with application to audio scene analysis, music information retrieval and sound transformation and synthesis.
LISTEN | 2022 - | 3 partners |
LISTEN is a joint laboratory launched by Télécom Paris with Valeo, Bruitparif and Music World Media, to develop cutting edge machine listening methodologies and systems. The joint laboratory will focus its research efforts on five fundamental issues, frugal learning based on scarce data, multi-view, multi-task & distributed learning, model-based deep learning, self-supervised learning, and finally (deep) generative models. Read More »
CIFRE Thalès | 2022 - 2025 | partners |
Scheduling using Transformers for adf hoc networks (Philippe Ciblat).
PEPR-NF-JEN | 2025 - | partners |
Modelling the carbon footprint of transceivers (Philippe Ciblat).
CIFRE Total | 2022 - 2025 | partners |
Reinforcement learning for wind turbine field yaw optimization: application to sea plants.
CIFRE Vinci | 2024 - 2027 | partners |
Optimization of energy consumptions in airports HVAC systems using machine learning and data-driven optimal control (Olivier Fercoq and Radu Dragomir).
Past projects
EU H2020 ETN | 2018 - 2022 | 17 partners | 1,05 M€
MIP-Frontiers is a multidisciplinary, transnational and cross-sectoral European training network for MIR researchers that aims to train a new generation of researchers. We bring together leading MIR groups and a wide range of industrial and cultural stakeholders in order to contribute to Europe’s leading role in this field of scientific innovation and accelerate the impact of innovation on European products and industry. Read More »
DSAIDIS | 2019 - 2024 | 7 partners |
Data Science & Artificial Intelligence for Digitalized Industry & Services is Télécom Paris’s 15th Chair. Established for a five-year period with the help of the Fondation Mines-Télécom and the support of Airbus Defence & Space, Engie, Idemia, Safran and Valeo, it is held by Florence d’Alché-Buc, Full Professor at Télécom Paris in the fields of Computer Science and Applied Mathematics. It is concerned with: 1) developing predictive analytics on time series and data streams; 2) exploiting large scale, heterogeneous, partially labeled data; 3) machine learning for trusted and robust decision; 4) learning through interactions with the environment. Read More »
Industrial and Institutional Partners