7.1.1 Pl@ntNet

• Keywords:
  Plant identification, Deep learning, Citizen science
• Functional Description:
  Pl@ntNet is a participatory platform and information system dedicated to the production of botanical data through deep learning-based plant identification. It includes 3 main front-ends, an Android app (the most advanced and the most used one), an iOs app (being currently re-developed) and a web version. The main feature of the application is to return the ranked list of the most likely species providing an image or an image set of an individual plant. In addition, Pl@ntNet’s search engine returns the images of the dataset that are the most similar to the queried observation allowing interactive validation by the users. The back-office running on the server side of the platform is based on Snoop visual search engine (a software developed by ZENITH) and on NewSQL technologies for the data management. The application is distributed in more than 200 countries (30M downloads) and allows identifying about 35K plant species at present time.
• Publications:
  hal-01629195, hal-02937618, hal-03343235, hal-01182775
• Contact:
  Alexis Joly
• Participants:
  Antoine Affouard, Jean-Christophe Lombardo, Pierre Bonnet, Hervé Goëau, Mathias Chouet, Hugo Gresse, Julien Champ, Alexis Joly

7.1.2 ThePlantGame

• Keyword:
  Crowd-sourcing
• Functional Description:
  ThePlantGame is a participatory game whose purpose is the production of big taxonomic data to improve our knowledge of biodiversity. One major contribution is the active training of the users based on innovative sub-task creation and assignment processes that are adaptive to the increasing skills of the user. Thousands of players are registered and produce on average about tens new validated plant observations per day. The accuracy of the produced taxonnomic tags is very high (about 95%), which is quite impressive considering the fact that a majority of users are beginners when they start playing.
• Publication:
  hal-01629149
• Contact:
  Alexis Joly
• Participants:
  Maximilien Servajean, Alexis Joly

7.1.3 Savime

• Name:
  Simulation And Visualization IN-Memory
• Keywords:
  Data management., Distributed Data Management
• Functional Description:
  SAVIME is a multi-dimensional array DBMS for scientific applications. It supports a novel data model called TARS (Typed ARray Schema), which extends the basic array data model with typed arrays. In TARS, the support of application dependent data characteristics is provided through the definition of TAR objects, ready to be manipulated by TAR operators. This approach provides much flexibility for capturing internal data layouts through mapping functions, which makes data ingestion independent of how simulation data has been produced, thus minimizing ingestion time.
• Publication:
  lirmm-01620376
• Contact:
  Patrick Valduriez
• Participants:
  Hermano Lustosa, Fabio Porto, Patrick Valduriez
• Partner:
  LNCC - Laboratório Nacional de Computação Científica

7.1.4 OpenAlea

• Keywords:
  Bioinformatics, Biology
• Functional Description:
  OpenAlea is an open source project primarily aimed at the plant research community. It is a distributed collaborative effort to develop Python libraries and tools that address the needs of current and future works in Plant Architecture modeling. It includes modules to analyze, visualize and model the functioning and growth of plant architecture. It was formally developed in the Inria VirtualPlants team.
• Release Contributions:
  OpenAlea 2.0 adds to OpenAlea 1.0 a high-level formalism dedicated to the modeling of morphogenesis that makes it possible to use several modeling paradigms (Blackboard, L-systems, Agents, Branching processes, Cellular Automata) expressed with different languages (Python, L-Py, R, Visual Programming, ...) to analyse and simulate shapes and their development.
• Publications:
  hal-01166298, hal-00831811
• Contact:
  Christophe Pradal
• Participants:
  Christian Fournier, Christophe Godin, Christophe Pradal, Frédéric Boudon, Patrick Valduriez, Esther Pacitti, Yann Guédon
• Partners:
  CIRAD, INRAE

7.1.5 Imitates

• Name:
  Indexing and mining Massive Time Series
• Keywords:
  Time Series, Indexing, Nearest Neighbors
• Functional Description:
  Time series indexing is at the center of many scientific works or business needs. The number and size of the series may well explode depending on the concerned domain. These data are still very difficult to handle and, often, a necessary step to handling them is in their indexing. Imitates is a Spark Library that implements two algorithms developed by Zenith. Both algorithms allow indexing massive amounts of time series (billions of series, several terabytes of data).
• Publication:
  lirmm-01886794
• Contact:
  Florent Masseglia
• Partners:
  New York University, Université Paris-Descartes

7.1.6 UMX

• Name:
  open-unmix
• Keywords:
  Source Separation, Audio
• Scientific Description:
  UMX implements state of the art audio/music source separation with deep neural networks (DNNs). It is intended to serve as a reference in the domain. It has been presented in two major scientific communications: An Overview of Lead and Accompaniment Separation in Music (https://hal-lirmm.ccsd.cnrs.fr/lirmm-01766781) and Music Separation with DNNs (Making it work (ISMIR 2018 Tutorial) https://sigsep.github.io/ismir2018_tutorial/index.html#/cover).
• Functional Description:
  UMX implements audio source separation with deep learning, using the Pytorch and Tensorflow frameworks. It comprises the code for both training and testing the separation networks, in a flexible manner. Pre- and post-processing around the actual deep neural nets include sophisticated specific multichannel filtering operations.
• Publication:
  lirmm-01766781
• Authors:
  Antoine Liutkus, Fabian Robert Stoter, Emmanuel Vincent
• Contact:
  Antoine Liutkus

7.1.7 TDB

• Keywords:
  Data assimilation, Big data, Data extraction
• Scientific Description:
  TDB comes as a building block for audio machine learning pipelines. It is a scraping tool that allows large scale data augmentation. Its different components allow building a large dataset of samples composed of related audio tracks, as well as the associated metadata. Each sample comprises a dynamic number of entries.
• Functional Description:

  TDB is composed of two core submodules. First, a data extraction pipeline allows scraping a provider url so as to extract large amounts of audio data. The provider is assumed to offer audio content in a freely-accessible way through a hardcoded specific structure. The software automatically downloads the data locally under a raw data format. To aggregate the raw data set, a list of item ids is used. The item ids will be requested from the provider given a url in parallel fashion. Second, a data transformation pipeline allows transforming the raw data into a dataset that is compatible with machine learning purposes. Each produced subfolder contains a set of audio files corresponding to a predefined set of sources, along with the associated metadata. A working example is provided.

  Each component has several submodules, in particular, network handling and audio transcoding. Thus, TDB can be viewed as an extract-transform-load (ETL) pipeline that enables applications such as deep learning on large amounts of audio data, assuming that an adequate data provider url is fed into the software.

• Contact:
  Antoine Liutkus
• Participants:
  Antoine Liutkus, Fabian Robert Stoter

7.1.8 UMX-PRO

• Name:
  Unmixing Platform - PRO
• Keywords:
  Audio signal processing, Source Separation, Deep learning
• Scientific Description:

  UMX-PRO is written in Python using the TensorFlow 2 framework and provides an off-the-shelf solution for music source separation (MSS). MSS consists in extracting different instrumental sounds from a mixture signal. In the scenario considered by UMX-PRO, a mixture signal is decomposed into a pre-defined set of so called targets, such as: (scenario 1) {"vocals", "bass", "drums", "guitar", "other"} or (scenario 2) {"vocals", "accompaniment"}.

  The following key design choices were made for UMX-PRO. The software revolves around the training and inference of a deep neural network (DNN), building upon the TensorFlow v2 framework. The DNN implemented in UMX-PRO is based on a BLSTM recurrent network. However, the software has been designed to be easily extended to other kinds of network architectures to allow for research and easy extensions. Given an appropriately formatted database (not part of UMX-PRO), the software trains the network. The database has to be split into train and valid subsets, each one being composed of folders called samples. All samples must contain the same set of audio files, having the same duration: one for each desired target. For instance: {vocals.wav, accompaniment.wav}. The software can handle any number of targets, provided they are all present in all samples. Since the model is trained jointly, a larger number of targets increases the GPU memory usage during training. Once the models have been trained, they can be used for separation of new mixtures through a dedicated end-to-end separation network. Interestingly, this end-to-end network comprises an optional refining step called expectation-maximization that usually improves separation quality.

• Functional Description:
  UMX-PRO implements a full audio separation deep learning pipeline in Tensorflow v2. It provides everything needed to train and use a deep learning model for separating music signals, including network architecture, data pipeline, training code, inference code as well as pre-trained weights. The software comes with full documentation, detailed comments and unit tests.
• Authors:
  Antoine Liutkus, Fabian Robert Stoter
• Contact:
  Antoine Liutkus

8.1.1 Scientific Workflows for Plant Phenotyping and Modeling

Participants: Christophe Pradal, Esther Pacitti, Patrick Valduriez.

With the development of new plant phenotyping platforms, biologists are faced with a deluge of data to be computationally analyzed. The OpenAlea software addresses this challenge, with the support of data-intensive scientific workflows 10 and cache-aware techniques to process huge datasets in the cloud 5.

In 18, we propose an automatic phenotyping method called PhenoTrack3D to extract to extract a $3D+t$ reconstruction of maize from images. It allows the study of plant architecture and individual organ development over time during the entire growth cycle. The method tracks the development of each organ from a time-series of plants whose organs have already been segmented in 3D using existing methods. In 22, we propose another automatic phenotyping method for root systems. This method combines an imaging device and an automatic analysis workflow based on registration and topological tracking, capable of accurately describing the topology and geometry of observed root systems in $2D+t$. In complement to these phenotyping methods, available in OpenAlea, we designed the mathematical model HydroRoot to study the role of root architecture and anatomy in water uptake. In 17, we investigate how the interplay between conductivities along radial (e.g. aquaporins) and axial (e.g. xylem vessels) pathways determines water transport properties of highly branched root system architectures. In 14, we model the concomitant transport of water and solutes in roots under water deficit, by considering hydrostatic and osmotic forces on maize roots. In 28, we assess to what extent the phenomics and modelling communities can address the issues of interoperability and data exchange, using a science mapping approach.

8.1.2 Elastic Scalable Transaction Processing

Participants: Patrick Valduriez.

Scaling ACID transactions in a cloud database is hard, and providing elastic scalability even harder. In 24, we present a solution for elastic scalable transaction processing. Unlike previous solutions, it does not require any hardware assistance. Yet, it does scales linearly to 100s of servers. We show the correctness of our solution. Finally, we provide a thorough performance evaluation of our solution using LeanXcale, an industrial-strength NewSQL database system on Amazon Web Services (AWS). The results show linear scalability, e.g., 5 million TPC-C NewOrder transactions per minute (TPM) with 200 nodes, which is greater than the TPC-C throughput obtained by the 9th highest result in all history using dedicated hardware used exclusively (not shared like in our evaluation) for the benchmark. Furthermore, the efficiency in terms of TPM per core is double that of the two top TPC-C results.

8.1.3 Distributed Intelligence on the Edge-to-Cloud Continuum

Participants: Daniel Rosendo, Patrick Valduriez.

The large scale and optimized deployment of learning-based workflows across the Edge-to-Cloud Continuum requires extensive and reproducible experimental analysis of application execution on representative testbeds. A thorough experimental analysis requires the assessment of the impact of multiple factors, such as: model accuracy, training time, network overhead, energy consumption, processing latency, among others. In 27, we propose a comprehensive review of the state-of-the-art in learning-based analytics for the Edge-to-Cloud Continuum. The main simulation, emulation, deployment systems, and testbeds for experimental research on the Edge-to-Cloud Continuum available today are also surveyed, with special attention on experiment reproducibility.

8.1.4 Data and Model Management with Gypscie

Participants: Patrick Valduriez.

As predictive analytics using ML models (or models for short) become prevalent in different stages of scientific exploration, a new set of artifacts are produced during the models’ life-cycle that need to be managed. In addition to the models’ versions, ML life-cycle artifacts include the collected training data and pre-processing workflows, data labels and selected features, model training, tuning and monitoring statistics and provenance information. However, to realize the full potential of data science, these artifacts must be built and combined, which can be very complex as there can be many. Furthermore, they should be shared and reused, in particular, in different execution environments such as HPC or Spark clusters. In order to support the complete ML life-cycle process and produced artifacts, we propose the Gypscie framework 45, to develop, share, improve and publish ML artifacts. Gypscie is integrated with the SAVIME database system for querying tensor data. In particular, we proposed a data-driven approach for selecting pre-trained temporal models to be applied at each query point 46, which avoids training a different model for each domain point, thus saving model training time. We implemented this approach into the SAVIME database system 44.

8.2.1 Time Series Prediction and Anomaly Detection

Participants: Esther Pacitti.

TSPred is a prediction process that seamlessly integrates nonstationary time series transformations with state-of-the-art statistical and machine learning methods. In 29, we describe a novel implementation of TSPred for time series prediction in association with data preprocessing. It is made available as an R-package, which provides functions for defining and conducting time series prediction, including data pre(post)processing, decomposition, modeling, prediction, and accuracy assessment. Furthermore, TSPred enables user-defined methods, which significantly expands the applicability of the framework.

Time series event detection is related to studying methods for detecting observations in a series with special meaning. These observations differ from the expected behavior of the data set. In 41, we propose a novel method for detecting events in nonstationary time series The method, entitled Forward and Backward Inertial Anomaly Detector (FBIAD), analyzes inconsistencies in observations concerning surrounding temporal inertia (forward and backward). FBIAD is shown to outperform other methods both in accuracy and elapsed time.

8.2.2 Entropy-Based Segmentation of Time Series

Participants: Lamia Djebour, Reza Akbarinia, Florent Masseglia.

Many applications today generate data at increasing rates. The data may concern personal activities (e.g., electricity or water consumption using smart-meters or smart plugs) or professional activities (e.g., heart activity through monitoring or farming using plants sensors). This results in the production of large and complex data, usually in the form of time series. Data mining techniques on massive sets of time series have drawn a lot of interest since their application may lead to improvements in a large number of these activities, relying on fast and accurate similarity search in time series for performing tasks like , e.g., classification, clustering, and motif discovery. Because of the big data volumes, these tasks can be slow on raw data. This is why approximate representation of time series is needed as a means to allow fast computation of similarity search. In 34, 20, we propose a new representation technique, called ASAX (Adaptive SAX), which yields a variable-size segmentation of time series with high precision in retrieval tasks thanks to its lower information loss. Our representation is based on entropy measurement for detecting the time intervals that should be split. We also propose a lower bounding method that allows approximating the distance between the original time series based on their representations in ASAX. The experimental results show that the more the data distribution in the time domain is unbalanced, the greater the precision gain of ASAX.

8.2.3 Parallel Techniques for Variable Size Segmentation of Time Series Datasets

Participants: Lamia Djebour, Reza Akbarinia, Florent Masseglia.

Given the high data volumes in time series applications, or simply the need for fast response times, it is usually necessary to rely on alternative, shorter representations of the series, usually with some precision loss. This incurs approximate comparisons of time series where precision is a major issue. In 33, we propose efficient parallel techniques for improving the execution time of our segmentation algorithm. In this work, we first propose a new representation technique, called ASAX-SSE, which allows obtaining a variable-size segmentation of time series based on SSE (sum of squared error). It can reduce significantly the error incurred by possible splittings at different steps of the representation calculation. Then, we propose efficient parallel algorithms for improving the execution time of our segmentation approach using GPUs. We implemented our approach and conducted empirical experiments using more than 120 real world datasets. The results show significant performance gains in terms of precision for similarity search compared to SAX. They show the effectiveness of our parallel algorithms, e.g., up to ×45 faster than the sequential algorithm for 1M time series.

8.3.1 New Methods for Species Distribution Modeling at Large Scale

Participants: Benjamin Deneu, Christophe Botella, Alexis Joly, François Munoz.

In 21, we rather ask whether the temporal dimension of remote sensing images can also be exploited by deep-SDMs. We therefore built a substantial and original dataset (called DeepOrchidSeries) aimed at modelling the distribution of orchids on a global scale based on Sentinel-2 Image Time-series. Thanks to this ambitious dataset, we trained several deep-SDMs based on Convolutional Neural Networks (CNNs) whose input was extended to include the temporal dimension. To quantify the contribution of the temporal dimension, we designed a novel interpretability methodology based on temporal permutation tests, temporal sampling and temporal averaging.

In 16, we introduce a new Bayesian dynamic species distribution model to explore the roles of long-distance dispersal and age-structured fecundity in the transient invasion dynamics of Plectranthus barbatus, a woody plant invader in South Africa.

8.3.2 AI-based Herbarium Specimens Analysis

Participants: Hervé Goëau, Alexis Joly.

8.3.3 Evaluation of Species Identification and Prediction Algorithms

Participants: Alexis Joly, Herve Goeau, Benjamin Deneu, Titouan Lorieul, Camille Garcin.

8.3.4 New features in the Pl@ntNet platform

Participants: Alexis Joly, Benjamin Deneu, Jean-Christophe Lombardo, Antoine Affouard.

First, we developed an offline mode of the Pl@ntNet mobile application, which makes it possible to identify plant images without any connexion by downloading a compressed version of the full Pl@ntNet model. This embedded version can be automatically updated when a wifi connexion is available and the newly observed plants are uploaded automatically to Pl@ntNet's server. Data such as images and common names can also be downloaded to improve user's experience (at the price of more storage on the smartphone). About 10K users already have been using this new feature since its release in October 2022.

Another development is GeoPl@ntNet, an original web application that enables to find out which plant species have already been observed or are potentially present in this area by selecting a point on a map and drawing a rectangle around it. In order to predict species in places where there is little or no data, the application uses Deep-SDM 19, a deep learning model trained to predict the species present from high-resolution satellite data (from the IGN) and environmental data (temperature, precipitation, soil type, etc.). This technology is based on the outcomes of the PhD thesis of Benjamin Deneu and has been implemented within the framework of the European project Cos4Cloud, which develops open technologies for citizen science. The service is currently adapted for predictions in Metropolitan France.

8.3.5 Migratory Data Modeling

Participants: Ondrej Cifka, Antoine Liutkus.

We developed new deep neural networks that are able to successfully model and predict migration of animals, in the context of a collaboration with ecologists from the CEFE CNRS laboratory in Montpellier. The training data for such models has been gathered from the international MoveBank dataset and consists of time series of GPS data of various length, one for each animal in an acquisition campaign. Such data offers many challenges on its own, such as irregular sampling frequency, different species being represented, missing data, etc.

Our work involved designing Transformer models that could account for past positions as well as complementary geographical context data that was extracted from thirdparty sources like temperature, altitude, soil type, etc. This involved a significant engineering effort to allow training on GPU at effective speed.

8.3.6 Large Language Models for Protein Design

Participants: Baldwin Dumortier, Antoine Liutkus.

AlphaFold is a recent breakthrough in computational biology that successfully predicts the folding, i.e. the spatial configuration, of a protein given its defining sequence of amino acids. This achievement unlocked many new research efforts from a large community of researchers in computational biology. In the context of an ongoing collaboration with INRAE and CNRS, we focused on the original idea of designing a model that could successfully achieve inverse folding, i.e., outputting a sequence of amino acids when its input is a sequence of spatial positions. Such a model would actually be very useful for protein design. Leveraging our recent work on positional encoding and the large AlphaFold database that was released publicly on 2022, we were able to train such a network that achieved a very good accuracy 51. This involved a significant engineering effort to develop an efficient data pipeline.

Since the release of our technical report, several other teams have been interested in the same area and our perspectives on this topic are numerous. We notably intend to leverage our unique collaboration between a computer science and biology labs to organize the first large scale biological validation campaign for the several protein generative models that were proposed by the community in 2022, yet evaluated only through simulations.

8.3.7 Blackbox Explanatory Methods for Large Language Models

Participants: Ondrej Cifka, Antoine Liutkus.

While working on large language models for migratory modeling, we realized that an increasingly interesting research topic is the design of explanatory methods for such models. Indeed, although sheer performance is often fundamental, it is actually not always the case. For instance, ecologists are not interested so much in the capacity of a model in accutely predicting the future position of animals, but rather in successfully identifying which geographical variables are relevant in doing these predictions, or to which extent past context is exploited to make such predictions. The question of context length hence soon became central, because is can directly be related to the notion of memory for animal migrations.

This collaboration led us to develop context length probing50, which is a new original way to study the behavior of a large language model by studying how it reacts to input contexts of different lengths. Doing so, its simplicity is remarkable, departing from several other explanatory methods that often require accessing the actual internals of a model or further specific training.

9.1.1 Associate Teams in the framework of an Inria International Lab or in the framework of an Inria International Program

9.2.1 Visits of international scientists

9.3.1 Horizon Europe

9.3.2 H2020 projects

Institut de Convergence Agriculture numérique #DigitAg, (2017-2023), 275 Keuro.

Participants: Alexis Joly, Florent Masseglia, Esther Pacitti, Christophe Pradal, Patrick Valduriez.

#DigitAg brings together in a partnership of seventeen actors (public research and teaching organizations, transfer actors and companies) with the objective of accelerating and supporting the development of agriculture companies in France and in southern countries based on new tools, services and uses. Based in Montpellier with an office in Toulouse and Rennes and led by Irstea, #DigitAg's ambition is to become a world reference for digital agriculture. In this project, Zenith is involved in the analysis of big data from agronomy, in particular, plant phenotyping and biodiversity data sharing.

ANR PerfAnalytics (2021-2024), 100 Keuro.

Participants: Reza Akbarinia, Florent Masseglia.

PPR "Antibiorésistance": structuring tool "PROMISE" (2021-2024), 240 Keuro.

Participants: Reza Akbarinia, Florent Masseglia.

The objective of the PROMISE (PROfessional coMmunIty network on antimicrobial reSistancE) project is to build a large data warehouse for managing and anlyzing antimicrobial resitance (AMR) data. It gathers 21 existing professional networks and 42 academic partners from three sectors, human, animal, and environment. The project is based on the following transdisciplinary and cross-sectoral pillars: i) fostering synergies to improve the one-health surveillance of antibiotic consumption and AMR, ii) data sharing for improving the knowledge of professionals, iii) improving clinical research by analyzing the shared data.

CASDAR CARPESO (2020-2022), 87 Keuro.

Participants: Julien Champ, Hervé Goëau, Alexis Joly.

9.4.1 Others

Ministry of Culture (2019-2021), 130 Keuro

Participants: Alexis Joly, Jean-Christophe Lombardo.

DINUM, 80 Keuro

Participants: Reza Akbarinia, Florent Masseglia.

CACTUS Inria exploratory action (2020-2022), 200 Keuro

Participants: Alexis Joly, Joaquim Estopinan.

MUSE AI3P (2021-2024), 80 Keuro

Participants: Baldwin Dumortier, Antoine Liutkus.

MUSE MULTINODE (2021-2024), 100 Keuro

Participants: Antoine Liutkus.

10.1.1 Scientific events: organisation

10.1.2 Scientific events: selection

10.1.3 Journal

10.1.4 Invited talks

• P. Valduriez: "Innovation : startup strategies", CEFET, Rio de Janeiro, 5 August 2022.
• A. Joly: "Plant disease characterization based on deep learning", CAPTE conference, Avignon, 2022.
• A. Joly: "Environmental ethics and fairness issues in the design of AI algorithms for biodiversity", PIT colloquium, online, 2022.
• A. Joly: "Pl@ntNet overview", German federal workshop on Geotagged photos in the context of the CAP, online 2022.
• A. Joly: "Tutorial on Vision Transformers", Imaginecology, Villeurbane, 2022.
• C. Pradal: "Plant & Crop Modeling", One Planet Fellowship Seminar, Montpellier, May 2022
• C. Pradal: "Towards Digital Twins for Horticulture", keynote talk, IHC 2022, Angers, August 2022
• C. Pradal: "Multiscale plant modeling and phenotyping in OpenAlea", keynote talk, OMICS Symposium 2022, Cali, Colombia, November 2022.

10.1.5 Leadership within the scientific community

• E. Pacitti: Member of the Steering Committee of the BDA conference.
• A. Joly: Foundator & scientific coordinator of LifeCLEF virtual lab: computer-assisted identification of living organisms (19 collaborators for the organization, 100s of registrants/participants, 100s publications, 1000s citations)
• A. Joly: Scientific and technical director of the Pl@ntNet platform: AI-based citizen science for plant biodiversity (3 permanent researchers, 4 engineers, 3 PhD students, 1 postdoc, tens of national and international partners, thousands of developers with an account on Pl@ntNet API, millions of end-users).
• A. Joly: Steering board of Cos4Cloud, H2020 research project (6M euros) aimed at integrating citizen science in the European Open Science Cloud (EOSC) through the co-design of innovative services.
• A. Liutkus: elected member of the Acoustic, Speech and Music signal Processing Technical Area Committee of the EURASIP (European Signal Processing Association).
• C. Pradal: Scientific and technical director of the OpenAlea platform: Multiscale platform for plant modelling (2 permanent researchers, 3 PhD students, 2 postdoc, tens of national partners, thousands of end-users).

10.1.6 Scientific expertise

• R. Akbarinia: Expert for STIC AmSud international program.
• A. Joly: expert for the French National HPC grand equipment (GENCI)
• A. Joly: member of the jury of the participatory research award of INRAE
• A. Joly: member of the jury of Inria researchers recruitment (CRCN and ISFP)
• A. Liutkus: expert for ANR.
• C. Pradal: member of the INRAE evaluation comitee CSS (Scientific Specialist Commission) in Plant Integrated Biology

10.1.7 Research administration

• E. Pacitti: manager of Polytech' Montpellier's International Relationships for the computer science department (100 students).
• P. Valduriez: scientific manager for the Latin America zone at Inria's Direction of Foreign Relationships (DRI) and scientific director of the Inria-Brasil strategic partnership.
• F. Masseglia: deputy scientific director of Inria for the domain "Perception, Cognition And Interaction".
• R. Akbarinia: Scientific referent for research data at Inria Antenna of Montpellier; Member of Inria national commission for research data.
• C. Pradal: Team leader with C. Granier of the PhenoMEn team of the AGAP Institut

10.2.1 Teaching

Most permanent members of Zenith teach at the Licence and Master degree levels at UM2.

Esther Pacitti responsibilities on teaching (theoretical, home works, practical courses,exams) and supervision at Polytech' Montpellier UM, for engineer students:

• IG3: Database design, physical organization, 54h, level, L3, 50 students.
• IG4: Distributed Databases and NoSQL, 80h , level M1, 50 students.
• Large Scale Information Management (Iot, Recommendation Systems, Graph Databases), 27h, level M2, 20 students.
• Supervision of industrial projects
• Supervision of master internships.
• Supervision of computer science discovery projects.

Patrick Valduriez:

• Professional: Big Data Architectures, 24h, level M2, Capgemini Institut.

Alexis Joly:

• Polytech' Montpellier: Content-Based Image Retrieval, 3h, level M2.

Christophe Pradal

• Univ. Montpellier: Root System Modelling, 15h, level M2.

10.2.2 Supervision

PhD & HDR:

• Defended PhD: Benjamin Deneu, Large-scale and High-resolution Species Distribution Modelling using Deep Learning, Univ. Montpellier. Advisors: Alexis Joly, François Munoz, Maximilien Servajean, Pierre Bonnet.
• PhD in progress: Daniel Rosendo, Enabling HPC-Big Data Convergence for Intelligent Extreme-Scale Analytics, started Oct 2019, Univ. Rennes. Advisors: Gabriel Antoniu, Alexandru Costan, Patrick Valduriez.
• PhD in progress: Camille Garcin, Multi-class classification with high label ambiguity and a long-tailed distribution. Advisors: Joseph Salmon, Maximilien Servajean, Alexis Joly.
• PhD in progress: Joaquim Estopinan, Species Conservation Status Prediction. Advisors: Alexis Joly, François Munoz, Maximilien Servajean, Pierre Bonnet.
• PhD in progress: Cesar Leblanc, Predicting biodiversity future trajectories through deep learning. Advisors: Alexis Joly, Maximilien Servajean, Pierre Bonnet.
• PhD in progress: Tanguy Lefort, Ambiguity of classification labels and expert feedback. Advisors: Joseph Salmon, Benjamin Charlier, Alexis Joly.
• Defended PhD: Lamia Djebour, Adaptive Segmentation Techniques for Efficient Representation of Time Series Datasets. Defended September 2022. Advisors: Reza Akbarinia, Florent Masseglia.
• HDR defended: A. Liutkus on principled methods for signal processing 49.

10.2.3 Juries

Members of the team participated to the following PhD or HDR committees:

• R. Akbarinia: Robin CARPENTIER, PhD, University of Paris-Saclay.
• R. Akbarinia: Hussein EL KHANSA, PhD, University of Montpellier.
• A. Joly: Paul Guelorget, PhD, Institut Polytechnique de Paris.
• A. Joly: Yang Loong, PhD, University of Malaya.
• A. Joly: Alexis Delaforge, University of Montpellier.
• A. Joly: Benjamin Deneu, PhD, University of Montpellier.
• A. Liutkus: Pritish Chandna (UPF Barcelona), Stylianos Mimilakis (Fraunhofer IDMT Ilmenau).
• F. Masseglia: Hubert Naacke, HDR defense, Sorbonne Université, reviewer.
• F. Masseglia: Walid Zeghdaoui, PhD, Université Lumière Lyon 2.

10.3.1 Internal or external Inria responsibilities

• F. Masseglia is local contact for Montpellier of the national project "1 Scientifique - 1 Classe, Chiche!".
• F. Masseglia is Member of the strategic committee of Fondation Blaise Pascal.
• Pl@ntNet Community management: A. Joly and H. Gresse spend several hours a week animating Pl@ntNet's user community. This includes: animating the community of developers using Pl@ntNet API (thousands of users, animating Pl@ntNet's social networks (twitter account, facebook account), managing the mailbox (contact@plantnet-project.org) and writing articles in the news section of the Pl@ntNet web site.

10.3.2 Articles and contents

• A. Joly is co-author of a popularization article in "The Conversation" magazine entitled "PlantNet, eBird, Spipoll, iNaturalist… ces applis au service de l’i-écologie".

International journals

• 14 articleF.Fabrice Bauget, V.Virginia Protto, C.Christophe Pradal, Y.Yann Boursiac and C.Christophe Maurel. A root functional-structural model allows to assess effects of water deficit on water and solute transport parameters.Journal of Experimental BotanyDecember 2022
  HALDOIback to text
• 15 articleQ.Quentin Bertrand, Q.Quentin Klopfenstein, M.Mathurin Massias, M.Mathieu Blondel, S.Samuel Vaiter, A.Alexandre Gramfort and J.Joseph Salmon. Implicit differentiation for fast hyperparameter selection in non-smooth convex learning.Journal of Machine Learning Research23149April 2022, 1-43
  HAL
• 16 articleC.Christophe Botella, P.Pierre Bonnet, C.Cang Hui, A.Alexis Joly and D. M.David M. Richardson. Dynamic Species Distribution Modeling Reveals the Pivotal Role of Human-Mediated Long-Distance Dispersal in Plant Invasion.Biology1192022, 1293
  HALDOIback to text
• 17 articleY.Yann Boursiac, C.Christophe Pradal, F.Fabrice Bauget, M.Mikaël Lucas, S.Stathis Delivorias, C.Christophe Godin and C.Christophe Maurel. Phenotyping and modeling of root hydraulic architecture reveal critical determinants of axial water transport.Plant Physiology1902June 2022, 1289-1306
  HALDOIback to text
• 18 articleB.Benoît Daviet, R.Romain Fernandez, L.Llorenç Cabrera-Bosquet, C.Christophe Pradal and C.Christian Fournier. PhenoTrack3D: an automatic high-throughput phenotyping pipeline to track maize organs over time.Plant Methods18130December 2022
  HALDOIback to text
• 19 article B.Benjamin Deneu, A.Alexis Joly, P.Pierre Bonnet, M.Maximilien Servajean and F.François Munoz. Very High Resolution Species Distribution Modeling Based on Remote Sensing Imagery: How to Capture Fine-Grained and Large-Scale Vegetation Ecology With Convolutional Neural Networks? Frontiers in Plant Science 13 May 2022
  HAL DOI back to text back to text
• 20 articleL.Lamia Djebour, R.Reza Akbarinia and F.Florent Masseglia. Variable-Size Segmentation for Time Series Representation.Transactions on Large-Scale Data- and Knowledge-Centered Systems2022
  HALback to text
• 21 articleJ.Joaquim Estopinan, M.Maximilien Servajean, P.Pierre Bonnet, F.François Munoz and A.Alexis Joly. Deep Species Distribution Modeling From Sentinel-2 Image Time-Series: A Global Scale Analysis on the Orchid Family.Frontiers in Plant Science13April 2022, 839327
  HALDOIback to text
• 22 articleR.Romain Fernandez, A.Amandine Crabos, M.Morgan Maillard, P.Philippe Nacry and C.Christophe Pradal. High-throughput and automatic structural and developmental root phenotyping on Arabidopsis seedlings.Plant MethodsDecember 2022
  HALDOIback to text
• 23 articleH.Hervé Goëau, T.Titouan Lorieul, P.Patrick Heuret, A.Alexis Joly and P.Pierre Bonnet. Can Artificial Intelligence Help in the Study of Vegetative Growth Patterns from Herbarium Collections? An Evaluation of the Tropical Flora of the French Guiana Forest.Plants114February 2022, 530-552
  HALDOIback to text
• 24 articleR.Ricardo Jimenez-Peris, D.Diego Burgos-Sancho, F.Francisco Ballesteros, M.Marta Patiño-Martinez and P.Patrick Valduriez. Elastic scalable transaction processing in LeanXcale.Information Systems108September 2022, 102043
  HALDOIback to text
• 25 articleJ.Ji Liu, C.Carlyna Bondiombouy, L.Lei Mo and P.Patrick Valduriez. Two-Phase Scheduling for Efficient Vehicle Sharing.IEEE Transactions on Intelligent Transportation Systems2312022, 457-470
  HALDOI
• 26 articleJ.Ji Liu, D.Daxiang Dong, X.Xi Wang, A.An Qin, X.Xingjian Li, P.Patrick Valduriez, D.Dejing Dou and D.Dianhai Yu. Large-scale Knowledge Distillation with Elastic Heterogeneous Computing Resources.Concurrency and Computation: Practice and ExperienceSpecial issue2022, e7272
  HALDOI
• 27 articleD.Daniel Rosendo, A.Alexandru Costan, P.Patrick Valduriez and G.Gabriel Antoniu. Distributed intelligence on the Edge-to-Cloud Continuum: A systematic literature review.Journal of Parallel and Distributed Computing166August 2022, 71-94
  HALDOIback to text
• 28 articleC.Clément Saint Cast, G.Guillaume Lobet, L.Llorenç Cabrera-Bosquet, V.Valentin Couvreur, C.Christophe Pradal, F.Francois Tardieu and X.Xavier Draye. Connecting plant phenotyping and modelling communities: lessons from science mapping and operational perspectives.in silico Plants41January 2022, 1-13
  HALDOIback to text
• 29 articleR.Rebecca Salles, E.Esther Pacitti, E.Eduardo Bezerra, F.Fabio Porto and E.Eduardo Ogasawara. TSPred: A framework for nonstationary time series prediction.Neurocomputing467January 2022, 197-202
  HALDOIback to text
• 30 articleE.Elisa Senger, S.Sonia Osorio, K.Klaus Olbricht, P.Paul Shaw, B.Béatrice Denoyes, J.Jahn Davik, S.Stefano Predieri, S.Saila Karhu, S.Sebastian Raubach, N.Nico Lippi, M.Monika Höfer, H.Helen Cockerton, C.Christophe Pradal, E.Ebru Kafkas, S.Suzanne Litthauer, B.Björn Usadel and B.Bruno Mezzetti. Towards smart and sustainable development of modern berry cultivars in Europe.Plant Journal11152022, 1238-1251
  HALDOI
• 31 articleR.Renan Souza, L. G.Leonardo G Azevedo, V.Vítor Lourenço, E.Elton Soares, R.Raphael Thiago, R.Rafael Brandão, D.Daniel Civitarese, E. V.Emilio Vital Brazil, M.Marcio Moreno, P.Patrick Valduriez, M.Marta Mattoso, R.Renato Cerqueira and M. a.Marco a S Netto. Workflow Provenance in the Lifecycle of Scientific Machine Learning.Concurrency and Computation: Practice and Experience3414June 2022, e6544
  HALDOI
• 32 articleS.Sasha Woods, M.Maria Daskolia, A.Alexis Joly, P.Pierre Bonnet, K.Karen Soacha, S.Sonia Liñan, T.Tim Woods, J.Jaume Piera and L.Luigi Ceccaroni. How Networks of Citizen Observatories Can Increase the Quality and Quantity of Citizen-Science-Generated Data Used to Monitor SDG Indicators.Sustainability147March 2022
  HALDOI

International peer-reviewed conferences

• 33 inproceedingsL.Lamia Djebour, R.Reza Akbarinia and F.Florent Masseglia. Parallel Techniques for Variable Size Segmentation of Time Series Datasets.ADBIS 2022 - 26th European Conference on Advances in Databases and Information Systems13389Lecture Notes in Computer ScienceTurin, Italy2022, 148-162
  HALDOIback to text
• 34 inproceedingsL.Lamia Djebour, R.Reza Akbarinia and F.Florent Masseglia. Variable size segmentation for efficient representation and querying of non-uniform time series datasets.SAC 2022 - 37th ACM/SIGAPP Symposium on Applied ComputingVirtual Event, United StatesApril 2022, 395-402
  HALDOIback to text
• 35 inproceedingsC.Camille Garcin, M.Maximilien Servajean, A.Alexis Joly and J.Joseph Salmon. Stochastic smoothing of the top-K calibrated hinge loss for deep imbalanced classification.ICML 2022 - 39th International Conference on Machine Learning162Baltimore, United StatesPMLR2022, 7208-7222
  HAL
• 36 inproceedingsH.Hervé Goëau, P.Pierre Bonnet and A.Alexis Joly. Overview of PlantCLEF 2022: Image-based plant identification at global scale.CLEF 2022 - Conference and Labs of the Evaluation Forum3180CEUR Workshop Proceedings153Bologne, Italy2022, 1916-1928
  HAL
• 37 inproceedingsA.Alexis Joly, H.Hervé Goëau, S.Stefan Kahl, L.Lukáš Picek, T.Titouan Lorieul, E.Elijah Cole, B.Benjamin Deneu, M.Maximilien Servajean, A.Andrew Durso, I.Isabelle Bolon, H.Hervé Glotin, R.Robert Planqué, W.-P.Willem-Pier Vellinga, H.Holger Klinck, T.Tom Denton, I.Ivan Eggel, P.Pierre Bonnet, H.Henning Müller and M.Milan Šulc. LifeCLEF 2022 Teaser: An Evaluation of Machine-Learning Based Species Identification and Species Distribution Prediction.Lecture Notes in Computer ScienceECIR 2022 - 44th European Conference on Information Retrieval ResearchLNCS-13186Advances in Information Retrieval : proceedings, part IIStavanger, NorwaySpringer International PublishingApril 2022, 390-399
  HALDOI
• 38 inproceedingsA.Alexis Joly, H.Hervé Goëau, S.Stefan Kahl, L.Lukáš Picek, T.Titouan Lorieul, E.Elijah Cole, B.Benjamin Deneu, M.Maximilien Servajean, A.Andrew Durso, H.Hervé Glotin, R.Robert Planqué, W.-P.Willem-Pier Vellinga, A.Amanda Navine, H.Holger Klinck, T.Tom Denton, I.Ivan Eggel, P.Pierre Bonnet, M.Milan Šulc and M.Marek Hrúz. Overview of LifeCLEF 2022: An Evaluation of Machine-Learning Based Species Identification and Species Distribution Prediction.Lecture Notes in Computer ScienceCLEF 2022 - 13th International Conference of the CLEF AssociationLNCS-13390Experimental IR Meets Multilinguality, Multimodality, and InteractionBologna, ItalySpringer International PublishingAugust 2022, 257-285
  HALDOIback to text
• 39 inproceedingsS.Stefan Kahl, A.Amanda Navine, T.Tom Denton, H.Holger Klinck, P.Patrick Hart, H.Hervé Glotin, H.Hervé Goëau, W.-P.Willem-Pier Vellinga, R.Robert Planqué and A.Alexis Joly. Overview of BirdCLEF 2022: Endangered bird species recognition in soundscape recordings.CLEF 2022 - Conference and Labs of the Evaluation Forum3180CEUR Workshop Proceedings154Bologne, Italy2022, 1929-1939
  HAL
• 40 inproceedingsC.César Leblanc, A.Alexis Joly, T.Titouan Lorieul, M.Maximilien Servajean and P.Pierre Bonnet. Species Distribution Modeling based on aerial images and environmental features with Convolutional Neural Networks.CEUR Workshop Proceedings 3180CLEF 2022 - Conference and Labs of the Evaluation ForumBologne, ItalySeptember 2022, 2123-2150
  HAL
• 41 inproceedingsJ.Janio Lima, P.Pedro Alpis, R.Rebecca Salles, L.Luciana Escobar, F.Fabio Porto, E.Esther Pacitti, R.Rafaelli Coutinho and E.Eduardo Ogasawara. Forward and Backward Inertial Anomaly Detector: A Novel Time Series Event Detection Method.IJCNN 2022 - IEEE International Joint Conference on Neural NetworksPadova, ItalyIEEE2022, 1-8
  HALDOIback to text
• 42 inproceedingsT.Titouan Lorieul, E.Elijah Cole, B.Benjamin Deneu, M.Maximilien Servajean, P.Pierre Bonnet and A.Alexis Joly. Overview of GeoLifeCLEF 2022: Predicting species presence from multi-modal remote sensing, bioclimatic and pedologic data.CLEF 2022 - Conference and Labs of the Evaluation Forum3180CEUR Workshop Proceedings155Bologne, Italy2022, 1940-1956
  HAL
• 43 inproceedingsP.Paul Mangold, A.Aurélien Bellet, J.Joseph Salmon and M.Marc Tommasi. Differentially Private Coordinate Descent for Composite Empirical Risk Minimization.ICML 2022 - 39th International Conference on Machine Learning162Baltimore, United StatesPMLR2022, 14948-14978
  HAL
• 44 inproceedingsA. C.Anderson Chaves Silva, P.Patrick Valduriez and F. A.Fábio André Machado Porto. Integrating Machine Learning Model Ensembles to the SAVIME Database System.SBBD 2022 - Brazilian Symposium on DatabasesBuzios, BrazilSeptember 2022, 1-8
  HALback to text
• 45 inproceedingsP.Patrick Valduriez and F.Fabio Porto. Data and Machine Learning Model Management with Gypscie.CARLA 2022 - Workshop on HPC and Data Sciences meet Scientific ComputingPorto Alegre, BrazilSeptember 2022, 1-2
  HALback to text
• 46 inproceedingsR.Rocío Zorrilla, E.Eduardo Ogasawara, P.Patrick Valduriez and F.Fabio Porto. A Data-Driven Model Selection Approach to Spatio-Temporal Prediction.SBBD 2022 - Brazilian Symposium on DatabasesBuzios, BrazilSeptember 2022, 1-12
  HALback to textback to text

Edition (books, proceedings, special issue of a journal)

• 47 periodicalTransactions on Large-Scale Data- and Knowledge-Centered Systems LI: Special Issue on Data Management - Principles, Technologies and Applications.13410Lecture Notes in Computer Science (LNCS)2022, 137
  HALDOI

Doctoral dissertations and habilitation theses

• 48 thesisL.Lamia Djebour. Adaptive Segmentation Techniques for Efficient Representation of Time Series Datasets.Université de MontpellierSeptember 2022
  HAL
• 49 thesisA.Antoine Liutkus. Principled methods for mixtures processing.Université de MontpellierFebruary 2022
  HALback to text

Reports & preprints

• 50 miscO.Ondřej Cífka and A.Antoine Liutkus. Black-box language model explanation by context length probing.December 2022
  HALback to text
• 51 miscB.Baldwin Dumortier, A.Antoine Liutkus, C.Clément Carré and G.Gabriel Krouk. PeTriBERT : Augmenting BERT with tridimensional encoding for inverse protein folding and design.August 2022
  HALDOIback to text
• 52 miscQ.Quentin Groom, M.Mathias Dillen, W.Wouter Addink, A.Arturo Ariño, C.Christian Bölling, P.Pierre Bonnet, L.Lorenzo Cecchi, E.Elizabeth Ellwood, R.Rui Figueira, P.-Y.Pierre-Yves Gagnier, O.Olwen Grace, A.Anton Güntsch, H.Helen Hardy, P.Pieter Huybrechts, R.Roger Hyam, A.Alexis Joly, I.Isabel Larridon, V. K.Vamsi Krishna Kommineni, L.Laurence Livermore, R. J.Ricardo Jorge Lopes, J.Jeremy Miller, S.Sofie Meeus, K.Kenzo Milleville, M.Marc Pignal, R.Renato Panda, J.Jorrit Poelen, B.Blagoj Ristevski, T.Tim Robertson, C.Cristina Rufino, J.Joaquim Santos, M.Maarten Schermer, K.Katja Seltmann, B.Ben Scott, H.Heliana Teixeira, M.Maarten Trekels and J.Jitendra Gaikwad. Envisaging a global infrastructure to exploit the potential of digitised collections.2022
  HALDOI
• 53 miscT.Tanguy Lefort, B.Benjamin Charlier, A.Alexis Joly and J.Joseph Salmon. Improve learning combining crowdsourced labels by weighting Areas Under the Margin.October 2022
  HAL
• 54 misc Q.Quentin Leroy, O.Olivier Buisson and A.Alexis Joly. How does the degree of novelty impacts semi-supervised representation learning for novel class retrieval? November 2022
  HAL
• 55 miscP.Paul Mangold, A.Aurélien Bellet, J.Joseph Salmon and M.Marc Tommasi. High-Dimensional Private Empirical Risk Minimization by Greedy Coordinate Descent.2022
  HAL