Artificial Intelligence (AI) algorithms, and in particular Deep Neural Networks (DNNs), typically run in the cloud on clusters of CPUs and GPUs. To be able to run DNNs algorithms out of the cloud and onto distributed Internet-of-Things (IoT) devices, customized HardWare platforms for Artificial Intelligence (HW-AI) are required. However, similar to traditional computing hardware, HW-AI is subject to hardware faults, occurring due to manufacturing faults, component aging and environmental perturbations. Although HW-AI comes with some inherent fault resilience, faults can still occur after the training phase and can seriously affect DNN inference running on the HW-AI. As a result, DNN prediction failures can appear, seriously affecting the application execution. Furthermore, if the hardware is compromised, then any attempt to explain AI decisions risks to be inconclusive or misleading. One of the overlooked aspects in the state-of-the-art is the impact that hardware faults can have in the application execution and the decisions of HW-AI. This impact is of significant importance, especially when HW-AI is deployed in safety-critical and mission-critical applications, such as robotics, aerospace, smart healthcare, and autonomous driving. RE-TRUSTING is the first project to include the impact of HW-AI reliability on the safety, trust, and explainability of AI decisions. Typical reliability approaches, such as on-line testing and hardware redundancy, or even retraining, are less appropriate for HW-AI due to prohibited area and power overheads. Indeed, DNNs are large architectures with important memory requirements, coming along with an immense training set. RE-TRUSTING will address these limitations by exploiting the particularities of HW-AI architectures to develop low-cost and efficient reliability strategies. To achieve that, RE-TRUSTING will develop fault models and perform a failure analysis of HW-AI with the aim to study their vulnerability towards explaining the HW-AI. Explainable HW-AI signifies reassuring that the HW-AI is fault-free, thus neither compromising nor biasing the AI decision-making. In this regard, RE-TRUSTING aims at bringing confidence into the AI decision-making by explaining the hardware on which AI algorithms are being executed.

Understanding, modeling, forecasting and reconstructing fine-scale and large-scale processes and their interactions are among the key scientific challenges in ocean-atmosphere science. Artificial Intelligence (AI) technologies and models open new paradigms to address poorly-resolved or poorly-observed processes in ocean-atmosphere science from the in-depth exploration of available observation and simulation big data. In this context, this proposal aims to bridge the physical model-driven paradigm underlying ocean & atmosphere science and AI paradigms with a view to developing geophysically-sound learning-based and data-driven representations of geophysical flows accounting for their key features (e.g., chaos, extremes, high-dimensionality). We specifically address three key methodological questions: (i) How to learn physically-sound representations of geophysical flows? (ii) Which learning paradigms for the representation of geophysical extremes? (iii) how to learn computationally-efficient representations and algorithms for data assimilation?. Upper ocean dynamics will provide the scientifically-sound sandbox for evaluating and demonstrating the relevance of these learning-based paradigms to address model-to-observation and/or sampling gaps for the modeling, forecasting and reconstruction of imperfectly or unobserved geophysical random flows. To implement these objectives, we gather a transdisciplinary expertise in Numerical Methods (INRIA GRA & Rennes), Applied Statistics (IMT, LSCE), Artificial Intelligence (IMT, LIP6) and Ocean and Atmosphere Science (IGE, INRIA GRA, LOPS), complemented by the participation of two SMEs (Ocean Data Lab and Ocean Next) to anticipate the added value of AI technologies in future earth observation missions and coupled observation-simulation systems.

Imagine you have to answer the following questions: how to build a computer-aided diagnosis tool for neurological disorders from images acquired from different medical imaging devices? that could identify which emotion is feeling a person from her face and her voice? How could these tools be still operational even when some data of a type is missing and/or poor quality? These questions are at the core of some problems addressed by the Institut de Neurosciences de la Timone (INT), where people have expertise in brain imaging based medical diagnosis, and Picxel, a SME centered on affective computing. The Laboratoire d'Informatique de Paris 6 (LIP6), the Laboration Hubert Curien (LaHC), and the Laboratoire d'Informatique Fondamentale de Marseille (LIF, head of the PI) are the other partners that are teaming up with INT and Picxel: in this project, they provide their renowned knowledge in machine learning, wherein they have developed, theoretical, algorithmic, and practical contributions. The five partners will closely work together to propose original and innovative advances in machine learning with a constant concern to articulate theoretical and applicative findings. The above questions pose the problem of (a) building a classifier capable of predicting the class (i.e. a diagnosis, or an emotion) of some object, (b) that of taking advantage of the few modalities or *views* used to depict the objects to classify and, possibly (c) that of building relevant representations that take advantages of these views. This is precisely what the present project aims at: the development of a well-founded machine learning framework for learning in the presence of what we have dubbed *interacting views*, and which is *the* notion we will take time to uncover and formalize. To address the issues of multiview learning, we propose to structure as follows. On the one hand, we will devote time to establish when and how classical (i.e. monoview-based) learnability results carry over to the multiview setting (WP1); this may require us to brush up on our understanding of different notions and accompanying measures of interacting views. On the other hand, possibly building upon the results just mentioned, we will build new dedicated multiview learning algorithms, according to the following lines of research: a) we will investigate the problem of learning (compact) multiview representations (WP2), then b) we will create new algorithms by leveraging some recent works on transfer learning -- multitasks and domain adaptation -- to the multiview setting (WP3), and, c) we will address the scalability of our algorithms to real-life conditions, such as large-dimension datasets and missing views (WP4). Finally, the performances of our learning algorithms will be assessed on benchmark datasets, both synthetic and real, that we will collect and make available for the machine learning community (WP5). Beyond the mere evaluation of our algorithms, these dataset will be disseminated to promote reproducible research, to identify the most suitable algorithms in a multiview setting, and to make the machine learning community aware of the exciting problems of multiview learning for affective computing and brain-image analysis.

Modern scientific methods heavily rely on large-scale 3D simulations. However, current data production speeds are much higher than storage speeds (5 orders of magnitude on typical supercomputers). This imbalance constitutes a major bottleneck in the scientific computing pipeline, such that most of the data generated by a simulation is not saved to disk, and thus remains unvisualized, unexplored and unanalyzed. TORI addresses this data bottleneck by introducing the next generation data reduction tools for large-scale scientific 3D data. TORI’s angle of attack is based on original and important advances in Topological Data Analysis (TDA), a class of techniques popularized in scientific visualization. TORI addresses data reduction at two levels: (i) at the data level, by deriving an analysis framework for ensembles of topological objects that is inspired by optimal transport, and (ii) at the computation level, by entirely revisiting TDA to adapt it to the context of high-performance in-situ data analysis. To identify informative datasets (i), TORI will introduce efficient methods for distance computations, barycenter evaluations and trajectory analysis. To perform this analysis on-the-fly (ii), TORI will revisit TDA with task parallel algorithms, coarse-to-fine computations and TDA-aware lossy compressors. TORI will be implemented in open-source in the Topology ToolKit, a leading TDA package, and interfaced with standard scientific computing packages (VTK, ParaView). It will be integrated in simulation codes with Catalyst and evaluated on real-life use cases in climatology, geophysics and astrophysics. TORI will have a far reaching impact on all fields of science using large-scale 3D simulations. By bringing together optimal transport and TDA in an innovative coarse-to-fine model, TORI will establish TDA as a standard tool for the analysis of large-scale ensemble datasets and it will initiate a new line of research in high-performance in-situ data analysis.

Today, information systems are one of the world's main resources. As accentuated by the Covid-19 crisis, our society relies on an ever-increasing need to process and communicate data, with significant repercussions on politics, defense, health, innovation, daily-life and the economy. The level of security remains a major issue for many use-cases, where secret key encryption are provably secure and can be implemented in the real world via quantum solutions. Quantum-safe communication, the first commercially available quantum technology, provides a unique means to establish, between distant locations, random strings of identical secret bits, with a level of security unattainable using conventional approaches. The implementation of actual quantum systems has become crucial, given the strong military, societal and economic impacts. This path, considered as one of the most promising for IT innovation, benefits from largely endowed R&D programs, such as the EU Flagship and other national initiatives (UK, Germany, China, USA, France). With the development of quantum computers and sensors, it becomes of prime necessity to connect them. Consequently, tasks such as distributed quantum computing and sensing will lead to a large-scale quantum Internet. The major obstacle to the adoption of such networks lies in the limited distance (~100 km) over which they can be deployed, due to losses in optical fibers and the curvature of the Earth. In the absence of reliable quantum repeaters, the space segment represents the only potential way to circumvent this limitation. To date, the only real demonstrations have been made in China (Micius satellite), but many projects are underway at the international scale. SoLuQS aims at effectively answering this demand by building industrial "entanglement source" prototypes that meet the constraints of spatialization, without compromising their performance. The key words of our achievements will be compactness and integrability, allowing satellite exploitation for both civil and military domains. These devices will eventually allow the connection of 2 metropolitan quantum networks (Paris and Nice). SoLuQS will therefore follow the promising path of new telecom-compatible laser optical communication systems in free space, and is thus part of the ASTRID AAP's thematic axis 3, "Cryptography - Communication", with a focus on "network security", their "operational implementation" based on "multimodal entanglement", as well as "space solutions". We will develop, at the French scale, the necessary tools for spatialization, in view of establishing a secure space/ground communication link, in order to anticipate future satellite realizations. SoLuQS brings together the best international teams in quantum communication (INPHYNI and LIP6) as well as a major French space industrial group (Thales Alenia Space) which will promote both integration and spatialization of the achievements. The consortium will pursue an active knowledge dissemination strategy. IP and the attraction of industrialists have a directly exploitable economic value, both in terms of patents, market reach, and creation of start-ups. We will ensure the training of staff and students as well as the promotion of partners in both the academic and industrial communities. These activities will be complemented by dissemination actions (international conferences, scientific and general public publications, etc.) in order to maximize the project impact. Taken as a whole, our actions will ensure France to play a leading role on the international level, in terms of disruptive quantum technologies for space quantum communication.