Search
Josephson Junctions by Electrostatic tuning of Dichalcogenides – JJEDi
The objective of this project is to develop gate tunable Josephson Junctions (JJ) based on transition metal dichalcogenides (TMDs). Monolayers will be used as the unique material forming the junction where the electronic phase will be tuned by local gates. The proposed junction will be fully controlled in-situ.This approach will be pushed down to the limit of few nanometer wide junctions. This will provide a perfect platform to test the physics and control of topological JJ.
Coherent High Q Micro-resonator Blue Light Sources – COMBO
COMBO aims to unveil the opportunity to develop coherent sources in the blue domain by use of whispering-gallery mode resonators. The ambition of the COMBO project is to develop a WGM resonator platform in the blue range (405-480 nm) for the demonstration of compact single frequency laser diodes and blue Kerr frequency combs.
Light-Bullets in Semiconductor Lasers – BLASON
The aim of this project is to conceive, to realize and to operate semiconductor laser devices for the generation and control of spatiotemporal solitons, also called “light bullets” (LB). LB will be implemented in Vertical External Cavity Surface Emitting semiconductor devices (VECSEL) with an external cavity closed by a semiconductor saturable absorber mirror (SESAM). Samples will be developed in the project in order to match the parameters requirements for LB existence.
Bidimensional RF optoelectronic devices based on PtSe2 – BIRDS
The intrinsic characteristics (doping, gap, mobility) of 2D materials can be largely modified by controlling their thickness and their environment (electric field, dielectric environment). These control possibilities have no equivalent in classical massive 3D semiconductors and provide opportunities to develop new electronic, optical and optoelectronic devices.
MULTI-variate, -temporal, -resolution and -SourCe remote sensing image Analysis and LEarning – MULTISCALE
MULTI-variate, -temporal, -resolution and -SourCe remote sensing image Analysis and LEarning
End-to-End Neural Approaches for Speech Translation – ON-TRAC
The ON-TRAC project proposes to radically change the architectures currently used in speech translation by exploring end-to-end neural approaches.<br />By performing this task with a single deep neural network, it is possible to better optimize its performance compared to a cascade system which requires first to transcribe automatically and then translate this transcription.<br />With ON-TRAC, it becomes possible to translate without transcribing the source language.
Effective Inference of Cleaning Programs from Data Annotations – InfClean
Besides reliable models for decision making, we need data that has been processed from its original, raw state into a curated form, a process referred to as «data cleaning«. In this process, engineers and domain experts collect specifications, such as business rules. Specifications are then encoded in programs to be executed over the raw data to identify and fix errors. This process is expensive and does not provide any formal guarantee on the ultimate quality of the data.
Distributed, Personalized, Privacy-Preserving Learning for Speech Processing – DEEP-PRIVACY
The project concerns the development of distributed, personalized and privacy preserving approaches for speech recognition. We propose a hybrid approach in which the terminal of the user performs private computations locally and does not share the raw speech data, while some inter-user computations (such as model optimization) are performed on a server or a peer-to-peer network, with speech data that are shared after anonymization.
Data to Knowledge in Agriculture and Biodiversity – D2KAB
D2KAB implements processes to extract and formalize knowledge – semantically rich, interoperable, open – from agronomy/agriculture and biodiversity/ecology data (data to knowledge). The project also studies scientific methods and tools to exploit and disseminate this knowledge in different scenarios in agriculture or biodiversity.
A PAC-Bayesian Representation Learning Perspective – APRIORI
The tools offered by the PAC-Bayesian theory allow to bring an original point of view on representation learning methods while bringing strong theoretical guarantees and new directions to develop new algorithms.