Understanding the limits of reliable information transmission over noisy communication channels and designing efficient codes for such channels are major cornerstones of information and coding theory. Most techniques developed in this area in the last 75 years have been targeted at discrete memoryless channels, which have the property that the i-th received symbol is an independent noisy function of the i-th transmitted symbol only. As a result, sender and receiver are synchronized, greatly simplifying their analysis. The study of such channels has given rise to a rich theory with many important applications beyond their original motivation. The project's goal is to tackle fundamental problems in the theory of information and efficient coding for channels which cause a loss of synchronization between sender and receiver. Besides theoretical interest, these channels capture important properties of modern data storage systems, such as DNA-based data storage. Almost all techniques designed for discrete memoryless channels break down when applied to channels with loss of synchronization. Therefore, studying even the simplest such channels (like the Binary Deletion Channel, which independently deletes each input bit with some probability) requires developing conceptually new techniques. I expect these techniques to have groundbreaking influence in other areas, like the techniques developed for discrete memoryless channels did. In this project, I aim to characterize the capacity of channels with synchronization errors and to design reliable codes with nearly-optimal rate and efficient encoding and error-correction procedures for these channels. This includes multi-trace channels with synchronization errors (that produce multiple corrupted outputs on a given input) and channels with correlated synchronization errors, both motivated by applications to DNA-based data storage.

YAHYA-6G aims to propose new signal processing solutions doped with machine-learning. We will focus on the detection and compensation of RF imperfections in mMIMO (massive Multiple input Multiple output) based NOMA (Non-orthogonal multiple access) pair . In other hand, YAHYA-6G target is to minimize the long-term power consumption based on the stochastic optimization theory for mMIMO-NOMA IoT networks with EH (Energy Harvesting) in presence of RF imperfections. Thus the objectives of the YAHYA-6G project are: 1- Identify major RF imperfections that may occur in a multi-access / multi-antenna broadband system. 2- Propose new solutions to optimize the energy efficiency at the RF transmitters. This solution will focus on the power amplifier that represents 60 at 70% of the energy consumed in an RF transmitter. 3- Analyze the impact of these RF imperfections on mobile radio systems exploiting NOMA technologies. 4- Propose a Deep Learning online learning process to detect the NOMA channel characteristics and compensate the effect of HPA nonlinearity. A joint detection of the NOMA interference and HPA (High Power Amplifier) nonlinearity will be studied in mMIMO-NOMA system. 5- Resolve a non convex based problem coping with the expected 6G requirements, with a particular focus on optimal resource scheduling and computation capacity allocation and reducing energy consumption of wireless devices, through a set of new algorithms . 6- Realize a demonstrator based on the SDR (Software Defined Radio) USRP cards on which some algorithms developed in the project will be implemented.

YAHYA-6G aims to propose new signal processing solutions doped with machine-learning. We will focus on the detection and compensation of RF imperfections in mMIMO (massive Multiple input Multiple output) based NOMA (Non-orthogonal multiple access) pair . In other hand, YAHYA-6G target is to minimize the long-term power consumption based on the stochastic optimization theory for mMIMO-NOMA IoT networks with EH (Energy Harvesting) in presence of RF imperfections. Thus the objectives of the YAHYA-6G project are: 1- Identify major RF imperfections that may occur in a multi-access / multi-antenna broadband system. 2- Propose new solutions to optimize the energy efficiency at the RF transmitters. This solution will focus on the power amplifier that represents 60 at 70% of the energy consumed in an RF transmitter. 3- Analyze the impact of these RF imperfections on mobile radio systems exploiting NOMA technologies. 4- Propose a Deep Learning online learning process to detect the NOMA channel characteristics and compensate the effect of HPA nonlinearity. A joint detection of the NOMA interference and HPA (High Power Amplifier) nonlinearity will be studied in mMIMO-NOMA system. 5- Resolve a non convex based problem coping with the expected 6G requirements, with a particular focus on optimal resource scheduling and computation capacity allocation and reducing energy consumption of wireless devices, through a set of new algorithms . 6- Realize a demonstrator based on the SDR (Software Defined Radio) USRP cards on which some algorithms developed in the project will be implemented.

In recent years, transformer-based deep learning models such as BERT or GPT-3 have led to impressive results in many natural language processing (NLP) tasks, exhibiting transfer and few-shot learning capabilities. However, despite faring well in benchmarks, current deep learning models for NLP often fail badly in the wild: they are bad at out-of-domain generalization, they do not exploit contextual information, they are poorly calibrated, and their memory is not traceable. These limitations stem from their monolithic architectures, which are good for perception, but unsuitable for tasks requiring higher-level cognition. In this project, I attack these fundamental problems by bringing together tools and ideas from machine learning, sparse modeling, information theory, and cognitive science, in an interdisciplinary approach. First, I will use uncertainty and quality estimates for utility-guided controlled generation, combining this control mechanism with the efficient encoding of contextual information and integration of multiple modalities. Second, I will develop sparse and structured memory models, together with attention descriptive representations towards conscious processing. Third, I will build mathematical models for sparse communication (reconciling discrete and continuous domains), supporting end-to-end differentiability and enabling a shared workspace where multiple modules and agents can communicate. I will apply the innovations above to highly challenging language generation tasks, including machine translation, open dialogue, and story generation. To reinforce interdisciplinarity and maximize technological impact, collaborations are planned with cognitive scientists and with a scale-up company in the crowd-sourcing translation industry.

The population in Europe is living longer and healthier, and that is a great achievement. On the other hand, an ageing population raises major financial and social challenges. One in four (25%) persons living in Europe could be aged 65+ by 2050. The greater expectancy of life in Europe is posing serious challenges to healthcare, namely through the associated increasing incidence of various diseases, as well as health conditions, which the elderly are mostly prone. One of the latter conditions is bone fractures, which can typically occur as a consequence of osteoporosis. Furthermore, the consequences of associated complications in fracture recovering include further costs, not only for the patient but also for the European society in general. To address such growing issue, the multidisciplinary consortium of ROBUST takes this as a challenging use case for demonstrating the relevance and proficiency of smart mobile eHealth systems as innovative solutions to address surging issues in our ageing society. ROBUST targets developing a new concept and platform for remote monitoring of patients’ healing process, in the eHealth domain. An integrated mobile eHealth system will be devised exploiting recent advances in RF-based sensing technologies, which are being investigated in this consortium. The system will be able to respond promptly to dynamic and complex situations, while preserving control, safety and privacy, in a reliable and energy efficiency manner. The ROBUST system will include a fast feedback loop that dynamically processes sensing information to generate, accordingly, instructions to the patient, encompassing cognitive and learning capabilities as well. ROBUST is committed to create an exceptional network, which is multidisciplinary and intersectoral in nature, for staff exchange in the mobile eHealth field, namely targeting structure training and knowledge sharing towards enhancing the European innovation capacity in relevant eHealth systems and applications.