Search
AT2TA - Analogies: from Theory to Tools and Applications – AT2TA
Analogical reasoning is a remarkable capability of human reasoning. Analogical proportions are statements of the form “A is to B as C is to D”. They are the basis of analogical inference that has been used in machine learning (ML) tasks such as classification, decision making, and automatic translat
Machine Translation for Open Science – MaTOS
The MaTOS (Machine Translation for Open Science) project aims to develop new methods for the machine translation (MT) of complete scientific documents, as well as automatic metrics to evaluate the quality of these translations. Our main application target is the translation of scientific articles be
Similarity Measure Learning for Analogical Transfer – SMeLT
The aim of SMeLT is to provide a methodology to learn a similarity measure that is optimized for a given analogical transfer task. Among the different tasks that computational analogy systems implement, the transfer task matches a predictive and hypothetical inference in which some knowledge i
Epistemic Reinforcement Learning – epiRL
EpiRL project aims at investigating the combination of epistemic planning and reinforcement learning (RL), by proposing new algorithms that are efficient, adaptive, and capable of computing decisions relying on theory of knowledge and belief. We expect from this approach an efficiency in the generat
Adaptive Co-Construction of Ethics for LifElong TRrustworthy AI – ACCELER-AI
As many applications involving Artificial Intelligence (AI) may have a beneficial or harmful impact on humans, there is a societal and a research debate about the way to incorporate ethical capabilities into them. This urges AI researchers to develop more ethically-capable agents shifting from ethic
Audio Quality Analysis for Representing, Indexing and Unifying Signals – AQUA-RIUS
Audio quality is an important characteristic which conveys intrinsic information about the audio creation process from recording to the studio post-mastering effects. In a recent study, we developed a pioneered method to objectively extract signal features enabling us to predict the applied audio ef
Bandits improve patients follow-up – BIP-UP
Machine learning is accompanied by many hopes in health. In project BIP-UP, we will explore the follow-up of patients who underwent bariatric surgery. Currently, this follow-up is the same for all patients though practicians are well aware that a personalized follow-up would better suit. Patient fol
NEXT GENERATION INFORMATION PROCESSING OF MICROSCOPY VECTOR-VALUED IMAGES: APPLICATION IN CELL POLARIZED IMAGING – POLARISCOPIA
Unlike conventional fluorescence microscopy, the new generation of polarized light-based microscopy instruments allow one to probe the orientation of fluorescently tagged biomolecules in cells. As the generated data are now 3D+time vector-valued signals encompassing density and orientation of molecu
Robust and Efficient Deep Learning based Audiovisual Speech Enhancement – REAVISE
Speech enhancement is a fundamental problem in signal processing that aims to improve the quality and intelligibility of a speech signal recorded in a noisy environment. This is of paramount practical importance, e.g. for automatic speech recognition systems and hearing assistive devices. While huma
Learning with limited annotations for medical image classification – MIMIC
Nowadays, Deep Learning has achieved a breakthrough in the field of Artificial Intelligence. However, huge, labeled datasets are needed to train on. Collecting such extensive annotated data is time and resources consuming and it is not feasible in the medical domain. Indeed, unlike the case of n
Advancing Federated LearnIng while Reducing tHe Carbon Footprint – DELIGHT
AI technologies today are too energy-intensive to be compatible with our sustainable development objectives. While recent work has assessed the carbon footprint of traditional learning methods, the carbon footprint of an emerging approach such as federated learning is not sufficiently studied. The D
Search the Web with Things – MeKaNo
In MeKaNo, we aim to search the web with things, in order to get more accurate results over a wide diversity of sources. Traditional web search engines search the web with strings. However, keyword search often returns many irrelevant documents, pushing users to refine their keyword list following a
ExtraCtion of LAtent knowledge in Documents by conjointly Analyzing Texts and TAbles – ECLADATTA
Identifying, extracting, structuring, and storing knowledge are major knowledge management tasks. They constitute important challenges for organizations, partly because knowledge is scattered across different types of sources (e.g. databases, spreadsheets, textual documents) and heterogeneously repr
Automatic Simplification of Scientific Texts – SimpleText
Information access systems provide users with key information from reliable sources such as scientific literature; however, non-experts tend to avoid these sources due to its complex language or their lack of background knowledge. Text simplification removes some of these barriers. SimpleText will b
Normative Artificial Intelligence for regulating MANufacturing – NAIMAN
The digital transformation of manufacturing industries provides a nurturing environment for the adoption of more autonomous and (self-)adaptive technologies that can quickly and flexibly respond to endogenous and exogenous changes, while being transparent and complying with sustainable regulations.
AD-Lib: An Aggregation-Disaggregation LIBrary for sequential decision models – AD-LIB
The ubiquity of SAT, Constraint Programming (CP) and Mixed Integer Programming (MIP) solvers in the optimization community has demonstrated the importance of generic algorithms able to solve any problem that can be expressed in a specific paradigm. In particular, any optimization methods have ac
Improving psychiatric screening with artificial intelligence – PORTRAIT
The aim of PORTRAIT project is to develop an adaptive test method that makes it possible to adapt the questions of a test according to the previous answers of the subject. For this, the project aims to extend recent advances in recommender systems and reinforcement learning methods to adapt tests. T
An argumentation-based platform for e-democracy – AGGREEY
E-democracy is a form of government that allows everybody to participate in the development of laws. It has numerous benefits since it strengthens the integration of citizens in the political debate. Several on-line platforms exist; most of them propose to represent a debate in the form of a graph,
Wasserstein Gradient Flows for Optimization and Sampling: non asymptotic properties and the non log-concave setting – WOS
An important problem in machine learning and computational statistics is to sample from an intractable target distribution. In Bayesian inference for instance, the latter corresponds to the posterior distribution of the parameters, which is known only up to an intractable normalisation constant, and
Explainable and parsimonious Preference models to get the most out of Inconsistent Databases – EXPIDA
The EXPIDA project aims to provide database consumers with a rich family of explainable methods to handle imperfect (i.e., inconsistent and/or uncertain) data with the aim to assist the analysis of aviation incidents data, and therefore to improve the civil air flights safety. More explicitly, our a
Frugal and Adaptive Testing – FATE
Testing is the process of gathering observations about an unknown system (for example a new drug and a placebo) in order to answer a question (e.g. which treatment is more efficient). Good testing protocols are such that the test can be stopped after few observations, while obtaining a correct answe
GRaphs and Algorithms for 3D proteIn structurE and dyNamics classificaTion – GRADIENT
Shape classification is one of the most important tasks in computer vision as demonstrated by the large body of work dealing with 3D shape analysis. Recent advances in 3D data acquisition and the availability of large 3D repositories have been instrumental in the design of new and more efficient alg
AdaPting and exPLaining fairnEss for Preference-based assIgnmEnt – APPLE-PIE
Many real-life applications deal with preference-based assignments. In such multi-agent problems, agents have preferences over items (activities, resources, or even other agents), and these preferences must be aggregated into a collective decision which is an assignment of agents to these items. One
Communication-efficient decentralized, adaptive and reliable optimization over multitask graphs – CEDRO
CEDRO falls into the broad theme of performing decentralized inference (stochastic optimization, estimation, and learning) over graphs. It notably recognizes the increasing ability of many emerging technologies to collect data in a decentralized and streamed manner. Therefore, the focus is on design
Efficient self-supervised learning for inclusive and innovative speech technologies. – E-SSL
Following previous major advances, self-supervised learning (SSL) has recently emerged as one of the most promising artificial intelligence (AI) methods. With this technique, it becomes feasible to take advantage of the colossal amounts of existing unlabeled data to significantly improve the results
Deciphering plant genotype-phenotype Interactions using knowledge Graphs and AI – DIG-AI
The demand for food is expected to grow substantially in the future. To meet this challenge in a context of climate change, a better understanding of genotype-phenotype relationships is crucial to improve crop production capacities. Agronomic research is witnessing an unprecedented revolution in the
Speaker diarization with a unified robust multimodal and spatial audio model – SAROUMANE
The speaker diarization (SD) aims to answer the question: “who speaks and when ?”. It still remains a challenging problem due to its various complex real scenario configurations (propagation environment, large number and moving speaker ...). In presence of at least two speakers (meeting, phone conve
Mapping Aerial imagery by learning on Game Engine-based simulations – MAGE
International satellite constellations such as Sentinel-2 and national high resolution imaging programs such as SPOT or BDORTHO in France have made Earth Observation (EO) data abundant. However, this data is unlabeled, i.e. misses semantic information useful to train machine learning (ML) models for
REPUBLIC: A Quest for {R}obustn{E}ss, {P}rivacy, and {U}n{B}iasedness in AI with Sequential {L}earn{I}ng under {C}onstraints – REPUBLIC
Developing responsible AI asks for effectively incorporating three fundamental aspects: robustness, privacy, and unbiasedness (fairness). In the REPUBLIC project, we propose to investigate these aspects for Reinforcement Learning (RL) with the framework of RL under constraints. The research roadma
The Why behind scenes – WhyBehindScenes
The project WhyBehindScenes aims at developing new methods for automatically understanding the storyline in videos, and in particular the why behind the scenes in edited videos (films and TV shows). This will be investigated in two directions: first, by automatically understanding the storyline by f
Scale-Space for Machine Learning on 3d point clouds – SSLAM
Some decades ago, the advent of digital photography led to the formation and development of dedicated research fields (e.g., image processing, pattern recognition, computer vision), which have revolutionized our daily lives. We believe that, in a near future, 3d acquisition will have a similar impac
Event-Centric Reasoning for Interpreting Everyday Narratives – ERIANA
Making sense of everyday narratives is a highly challenging task, which requires a deep understanding of the language, and abundant background knowledge about the world. While human readers can rely on high-level reasoning to draw conclusions, current NLP models largely lack this ability. Commonsen
Multimodel Streaming Data Management – POLYFLOW
The need for addressing data variety paved the road to the advent of multidatabases, which the Turing Award winner M. Stonebraker endorsed. Multi-databases include multistores, which expose a unified declarative query interface over heterogeneous data, and polystores which combine the benefits of mu