Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : CReSTIC – Université de Reims Champagne-Ardenne
Durée : 3 ans
Contact : bart.lamiroy@univ-reims.fr
Date limite de publication : 2023-06-12

Contexte :
Le contexte de cette thèse est le diagnostic des systèmes industriels. Elle contribuera à rendre ces systèmes plus flexibles et résilients par contribution à l’état de l’art du diagnostic et par l’intégration de l’apprentissage automatique. Cette intégration doit d’une part permettre de diagnostiquer des systèmes plus complexes avec moins de ressources de calcul, tout en permettant aux opérateurs humains de valider les diagnostics et d’engager des démarches correctives par une explicabilité suffisante des résultats produits par les outils de diagnostic et d’isolation.
Le travail se placera initialement dans le cadre de systèmes qui possèdent des capteurs et des actionneurs délivrant des signaux binaires (Systèmes à Evènements Discrets – SED) et commandés par des automates programmables industriels. Il pourra être envisagé que le l’étude s’élargisse par la suite à des systèmes hybrides incluant capteurs et actionneurs continus.

Sujet :
Dans cette thèse nous aborderons le diagnostic à travers des approches d’analyse automatique des données opérationnels sans modèle explicite, mais en mettant en œuvre des approches d’apprentissage automatique (notamment les réseaux de neurones) pour détecter et isoler les fautes.
La thèse poursuivra les deux objectifs suivants :
1. Extension des résultats préliminaires de diagnostic présentés dans (Saddem et al., 2022(a,b)) et de pouvoir dimensionner la taille des réseaux de neurones et ses hyper-paramètres en fonction du système surveillé.
2. Faire en sorte que les approches développées dépassent le stade de classifieur ou prédicteur en boite noire, et qu’elles puissent fournir un niveau d’explicabilité utile à un opérateur humain ; soit par formalisation des démarches de vérification (Fawzi et al. 2022) ou autres, soit par l’analyse des réseaux entrainés (Rojat et al. (2022).
L’ensemble de ce travail s’appuiera sur la plateforme Cellflex4.0 de l’Université Reims Champagne-Ardenne (https://crestic.univ-reims.fr/fr/plateformes/cellflex-4-0) et les Jumeaux Numériques associés pour la partie acquisition de données et validation expérimentale. Il bénéficiera également des capacités de calcul de la plateforme HPC Romeo.
L’accompagnement du/de la doctorant.e, assurée par les encadrants lui permettra de prioriser et de choisir les orientations scientifiques pour atteindre les objectifs précités. Les encadrants de thèse feront en sorte qu’elle.il acquière une méthodologie scientifique rigoureuse et pertinente.
Le déroulé de la thèse pourra s’envisager comme suit :
Etape 1 : Appropriation du sujet et développements de pistes scientifiques par l’établissement d’un état de l’art et la réalisation d’une étude bibliographique sur le diagnostic de systèmes industriels d’une part et une étude bibliographique sur les techniques d’apprentissage automatique focalisées sur le diagnostic d’autre part.
Etape 2 : Collecte et préparation des données nécessaires aux approches d’apprentissage identifiées à l’étape précédente. Il s’agit notamment de spécifier les données disponibles ainsi que les techniques qui doivent être utilisées pour leur collecte et ensuite les préparer pour les méthodes de diagnostic en ligne choisies.
Etape 3 : Développement et implémentation d’algorithmes de diagnostic par apprentissage automatique, entraîner l’algorithme pour retourner l’état du système (normal ou défaillant et si défaillant quelle classe de fautes). Valider l’algorithme sur les jumeaux numériques et sur la plateforme Cellflex.
Etape 4 : Explication du diagnostic fourni par un réseau de neurones à travers l’identification des composants défaillants. Cette vérification pourra notamment se faire par recherche explicite dans l’espace en utilisant de l’apprentissage par renforcement profond ou par l’analyse des réseaux entrainés et des approches de XAI plus habituelles.

Profil du candidat :
Ce sujet de thèse s’adresse aux étudiant.e.s ayant un diplôme de Master II ou d’ingénieur, soit issu.e.s d’un parcours ayant donné accès à des connaissances approfondies en systèmes industriels (notamment systèmes à évènements discrets et hybrides) et avec une envie d’investir le champ de l’Intelligence Artificielle, soit des étudiant.e.s de formation informatique avec de très bons acquis théoriques et applicatifs en apprentissage automatique souhaitant les appliquer à des contextes de systèmes industriels.
Au-delà des bases scientifiques citées ci-dessus, les autres compétences principales recherchées sont la curiosité et l’envie d’apprendre, la capacité de travail en équipe, la rigueur scientifique et la capacité de formalisation du raisonnement. Une grande partie du travail étant consacrée à la validation expérimentale des modèles, des compétences solides en programmation seront également nécessaires.

Formation et compétences requises :
Master II ou d’ingénieur, soit en Automatique, soit en Informatique

Adresse d’emploi :
Reims

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : LERIA – Université d’Angers
Durée : 3 years
Contact : nicolas.gutowski@univ-angers.fr
Date limite de publication : 2023-06-12

Contexte :
In many areas of chemistry, the discovery of new molecules often involves building upon an existing effective compound through chemical reactions (addition, substitution, etc.) to enhance its properties. The emergence of truly new molecules is a rarer phenomenon but can pave the way for intensification and profound transformations in the field. It is precisely with this goal in mind that research on de novo generation of molecules with desired properties has been developed, particularly for drug and material discovery. Challenges in this research domain include the size of the search space and the difficulty of generating molecules that can be synthesized.

Molecules can be represented as graphs, where the vertices are labeled according to atom types and the edges represent bond types. This is a problem of generating graph structures, where the objective is the combination of one or more functions to optimize and constraints to satisfy. To address our application in chemistry, we have recently proposed an evolutionary algorithm for molecule generation called EvoMol [1], which can freely explore the chemical space and tackle diverse problems. This generator has achieved benchmark results in multi-property optimization and applied problems. It is capable of incorporating synthesizability constraints [2] and promoting diversity in the generated molecules [3]. However, significant challenges still need to be addressed, and two of these challenges are at the core of the proposed topic.

The first area of improvement is the selection of actions on the graph, which is currently random in EvoMol. One would expect an intelligent method to apply a policy for choosing actions that have been successful in the past, similar to a chemist adding a known chemical function to enhance a target property. Preliminary work by N. Gutowski and B. Da Mota has shown the potential of reinforcement learning using bandit algorithms and Q-Learning for certain problems, highlighting the need for further methods. The second area of improvement relates to synthesizability, which is a crucial objective for real-world applications. We have proposed synthesizability constraints [2] that make the generated molecules likely to be synthesizable. Other works propose heuristic scores [4,5], sometimes based on retrosynthesis [6]. A method capable of constructing a molecule with the desired properties, along with the steps for its synthesis, would have many advantages. Although some work has emerged in this field [7], they are limited to simple problems.

Artificial learning involving graphs is a promising theme with numerous applications that has benefited from major advances in deep learning, such as Graph Neural Networks (GNN [8]), Graph Convolutional Networks (GCN [9]), and more recently, Graph Transformer Networks (GTN [10]). These approaches have been quickly adapted and applied to various applications, including molecular generation [11,12]. However, these powerful architectures have been relatively less explored in the context of constrained exploration and diverse optimization objectives, such as those of interest in our project (e.g., organic solar cells). For these complex objectives of sequentially constructing useful and realistic molecules, reinforcement learning appears to be a promising alternative to the meta-heuristics and latent space manipulation approaches [13] (such as variational autoencoders for molecules [14]) that have been employed thus far. The latter approaches face well-known challenges of posterior collapse and are limited to optimizing simple (i.e., differentiable) properties. While reinforcement learning is primarily used in the domain of games and robotics to learn optimal sequences of actions, its application for controlled generation of complex data has recently shown diverse promising developments, particularly in the field of natural language generation [15,16]. In protein chemistry, it has achieved widespread success through tools like AlphaFold [17]. However, although there are some works on molecular optimization problems [11,18,19], the issue of synthesizability is almost always minimized or neglected, leaving this task to retrosynthesis tools. However, even if such methods were used on molecules generated by current generative models, there is no guarantee that the necessary reagents to produce these molecules themselves can be obtained. This top-down method, performed a posteriori, is computationally expensive and weak. It does not allow setting objectives for the construction cost of the structure or other criteria that would help narrow down the search space. In chemistry, for example, one might want to minimize synthesis costs, the number of steps, minimize hazardous or difficult-to-recycle waste, etc. The joint integration of the bottom-up graph construction process and the optimization of these properties would be an elegant, effective, and original approach.

The application of deep reinforcement learning techniques for the discovery of molecules that are both stable, synthesizable, and exhibit properties of interest in the target domains is not without various scientific challenges that need to be overcome, based on the expertise of the supervising team in sequential learning: the input objects for policies are graphs, the actions are non-trivial in the case of reaction patterns, and the underlying problem is a multi-objective optimization problem. However, this type of application lends itself well to transfer or progressive learning (curriculum learning [20]), with the possibility of learning a policy on a simpler problem such as constraint optimization or synthesizability heuristics, and then improving this policy to optimize more complex objectives that include these synthesizability issues.

Beyond the intended application, the development of reinforcement learning (RL) techniques for policies conditioned on objective graphs, as suggested by the synthesis of molecules using available reactions, is an important theme for the Machine Learning community. While multi-task RL, specifically goal-conditioned RL, is expanding in the literature of the field, very few approaches deal with complex graph structures. In this context, the search for invariants in the manipulated structures will be a key lever for establishing effective policies. Additionally, automatic curriculum learning, which has seen numerous recent developments [20,21,22,23] by dynamically determining task specification distributions adapted to the current level of the learning agent, has not yet been deployed to our knowledge in environments with known dynamics, as in the case of our application. The exploitation and adaptation of Monte Carlo Tree Search (MCTS) planning algorithms for this framework seems to be a particularly promising research direction that we intend to develop in this thesis.

Références
[1] J. Leguy, T. Cauchy, M. Glavatskikh, B. Duval, et B. Da Mota. EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation. Journal of Cheminformatics, 2020.
[2] T. Cauchy, J. Leguy et B. Da Mota. Definition and exploration of realistic chemical spaces using the connectivity and cyclic features of ChEMBL and ZINC. Digital Discovery, Royal Society of Chemistry, under review.
[3] J. Leguy, M. Glavatskikh, T. Cauchy, et B. Da Mota. Scalable Estimator of the Diversity for De Novo Molecular Generation Resulting in a More Robust QM Dataset (OD9) and a More Efficient Molecular Optimization. Journal of Cheminformatics, 2021.
[4] Voršilák et al. SYBA: Bayesian estimation of synthetic accessibility of organic compounds. Journal of Cheminformatics, 2020.
[5] Bühlmann et al. ChEMBL-Likeness Score and Database GDBChEMBL. Frontiers in Chemistry, 2020.
[6] Thakkar et al. Retrosynthetic accessibility score (RAscore) – rapid machine learned synthesizability classification from AI driven retrosynthetic planning. Chemical Science, 2021.
[7] Bradshaw et al. Generating Molecules via Chemical Reactions. Workshop DeepGenStruct, ICLR 2019.
[8] Scarselli et al. The Graph Neural Network Model. IEEE Transactions on Neural Networks, 2009.
[9] Kipf et al. Semi-Supervised Classification with Graph Convolutional Networks. 5th International Conference on Learning Representations (ICLR), 2017.
[10] Yun et al. Graph Transformer Networks. Advances in Neural Information Processing Systems, NeurIPS, 2019.
[11] J. Leguy, T. Cauchy, B. Duval, et B. Da Mota. Goal-directed generation of new molecules by AI methods. Chapter of Computational and Data-Driven Chemistry Using Artificial Intelligence, Elsevier, 2022.
[12] Thölke et al. TorchMD-NET: Equivariant Transformers for Neural Network based Molecular Potentials. arXiv preprint arXiv:2202.02541, 2022.
[13] Zhang et al. Comparative Study of Deep Generative Models on Chemical Space Coverage. Journal of Chemical Information and Modeling, 2021.
[14] Liu et al. Constrained Graph Variational Autoencoders for Molecule Design. Advances in Neural Information Processing Systems, 2019.
[15] S. Lamprier, T. Scialom, A. Chaffin, V. Claveau, E. Kijak, J. Staiano, B. Piwowarski. Generative Cooperative Networks for Natural Language Generation. ICML, 2022
[16] T. Scialom, S. Lamprier, B. Piwowarski, J. Staiano. Answers Unite! Unsupervised Metrics for Reinforced Summarization Models. EMNLP/IJCNLP, 2019.
[17] Jumper et al. Highly accurate protein structure prediction with AlphaFold. Nature, 2021
[18] Khemchandani et al. DeepGraphMolGen, a multi-objective, computational strategy for generating molecules with desirable properties: a graph convolution and reinforcement learning approach. Journal of Cheminformatics, 2020.
[19] Zhou et al. Optimization of Molecules via Deep Reinforcement Learning. Scientific Reports, Nature, 2019.
[20] N. Castanet, S. Lamprier, O. Sigaud. Stein Variational Goal Generation For Reinforcement Learning in Hard Exploration Problems. CoRR 2022
[21] Andrychowicz et al. Hindsight Experience Replay. arXiv preprint arXiv:1707.01495, 2017.
[22] Florensa et al. Automatic goal generation for reinforcement learning agents. International conference on machine learning, PMLR, 2018.
[23] P-A Kamienny, J. Tarbouriech, S. Lamprier, A. Lazaric, L. Denoyer: Direct then Diffuse: Incremental Unsupervised Skill Discovery for State Covering and Goal Reaching. ICLR 2022

Sujet :
The first objective of the thesis will be to propose and implement reinforcement learning methods adapted to the problem, and then conduct a methodological study on toy problems, domain benchmarks, and realistic applications proposed by our chemistry partner. Learning a policy for generating molecules can be studied within the framework of classical reinforcement algorithms. Since the inputs to the policies are molecular graphs, it will be possible to use the descriptors proposed in our previous work or neural networks (Graph Neural Networks) from recent research that need to be adapted to the characteristics of molecular graphs.

The second objective of the thesis will be to transform a list of known reactions into actions applicable to molecular graphs, and then learn to sequence these chemical reactions to synthesize a target molecule (i.e., goal-conditioned RL, in the perspective of innovative conditioned bottom-up generation, rather than the usual top-down approaches involving complex and non-generalizable retrosynthesis calculations). It will also be possible to derive a fine-grained estimator of synthesizability. This objective will be the main focus of the research effort in machine learning, with valuable contributions to the statistical learning community (e.g., adaptation of planning approaches to the automatic curriculum framework).

The third and final objective will be to apply the sequencing of chemical reactions within the algorithms and developments proposed in the first part of the thesis, and then study the use of these actions in terms of performance and synthesizability criteria. The secondary benefit of this method is the possibility of not only proposing an optimized target but also justifying our proposal through the sequence of reactions that led to its elaboration.

Profil du candidat :
– H/F
– Master 2 ou école d’ingénieur

Formation et compétences requises :
– Bac+5
– Apprentissage par renforcement
– Deep Learning / Machine Learning
– Python

Adresse d’emploi :
UFR Sciences, 2 Bd de Lavoisier, 49000 Angers

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : Lorraine Research Laboratory in Computer Science a
Durée : 36 mois
Contact : sabeur.aridhi@loria.fr
Date limite de publication : 2023-06-30

Contexte :
We propose a fully funded 3-years PhD position in computer science with application to biomolecule analysis. The proposed position is funded by the Lorraine Université d’Excellence (LUE) through a multidisciplinary project that involves 2 researchers in computer science and 4 researchers in microbiology. A PhD thesis in microbiology will be conducted in parallel.

Context and motivations

Bacteriocins are antimicrobial peptides of bacterial origin with a very high economic potential in the agri-food sector. They are used in biopreservation/biocontrol applications to fight against undesirable microorganisms in the agronomy and food industry. LIBio has recently developed a technology based on the selection of two strains of the lactic acid bacterium Carnobacterium maltaromaticum producing anti-Listeria monocytogenes bacteriocins [9]. These strains inhibit the growth of this pathogen in cheeses when added to the manufacturing milk to produce the antimicrobial agents in the cheese matrix. These remarkable properties have led to a patent [10] that has very recently been licensed to a ferment producer. However, like the vast majority of biopreservation technologies, the effect is at best bacteriostatic: there is little or no decay of the pathogenic bacteria, which can then be maintained at low concentrations in the food. The biopreservation technologies described in the literature are based on engineering approaches that do not take advantage of the properties of the microbial communities forming the microbiomes of food products. Yet microbiome engineering is among the 12 promising technologies that could transform food systems over the next decade [11]. Indeed, in the case of biopreservation, assemblies of microorganisms could allow obtaining communities producing multiple antimicrobial agents and moreover being able to occupy the ecological niche of the undesirable microorganism to exclude it more efficiently. However, knowledge in the field of microbial community engineering is insufficient to fully exploit their potential. Indeed, due to the complexity of microbial communities, there is no available method to predict microbial community structure based on the knowledge of the ecological properties of microorganisms. Moreover, assembling microorganisms whose properties is to produce antimicrobial agents is a major difficulty because these agents can lead to the mutual exclusion of the microorganisms producing them.

Positioning

In a microbial ecosystem in which members produce antimicrobial substances like bacteriocins, three actors can be considered: the bacteriocin-producing microorganism (P) and the microorganisms sensitive (S) and resistant (R) to this bacteriocin. It was experimentally shown that in simple ecosystems mixing three such actors, all three actors are able to maintain equilibrium [12]. In these systems, S is more competitive than R because it does not pay the cost of resistance, and R is more competitive than P because it does not pay the cost of bacteriocin production. This cyclical relationship between P, S, and R is similar to that of the popular game « rock paper scissors » where no one player has an advantage over the other two: each player can overtake one player and each can be defeated by another. These simple experimental systems suggest that it is possible to implement engineering tools to predict the structure of complex communities based on the interactive properties of microorganisms. Thanks to the emergence of high throughput investigation methods, it is now possible to produce interaction data between large sets of microorganisms and thus reconstitute models of microorganism interaction networks. [6] [7] Lately, Ramia et al. [4] [5] built the interaction network corresponding to 73 Carnobacterium maltaromaticum strains. Like previously, the graph is sender-determined and also shows a highly nested structure [7], which means that it is different from a randomly built network with the same number of nodes and edges. The results also show that the competitive interaction network is very dense making C. maltaromaticum a very interesting model to develop community engineering approaches producing high performance antimicrobial substances cocktails for the fight against undesirable microorganisms. This project will use the data published in Ramia et al. [5] and will try to provide a rather computer science approach to the study of those interaction graph properties.

The originality of this project is that it will make it possible to integrate experimental variables describing the properties of interaction between microorganisms for the prediction of community structure which is not possible by existing methods.

Sujet :
Objectives of the thesis

The main goal of the thesis is to use advanced machine learning and graph-based approaches in order to predict the long-term community structure in microbiological ecosystems [3] [1]. Particularly, it aims at providing approaches to deduce diversity directly from the static, inner properties of the interaction graph the entities are involved in. The practical objectives of this interdisciplinary PhD project, which will be carried out in collaboration with researchers from the Laboratoire d’Ingénierie des Biomolécules (LIBio), are as follows:

to study existing research works on the analysis of interaction networks and long-term diversity prediction in bacteria.
to propose machine learning and graph-based approaches in order to learn models that are able to predict diversity based on the interaction graphs. In this context, regression methods could be used to learn the relation between the interaction graph properties and the diversity.
to study how graph embedding could help in predicting the level of development for each strain. In this context, we aim to study the impact of graph embedding methods on the prediction results. A specific embedding method could be proposed in the context of this project.

Profil du candidat :
Required qualification: Candidates must have a master degree in computer science. Good programming skills in a procedural language are essential. Experience of machine learning and graph mining is also desirable but not essential. A strong interest in bioinformatics would also be highly desirable.

Formation et compétences requises :
Required qualification: Candidates must have a master degree in computer science. Good programming skills in a procedural language are essential. Experience of machine learning and graph mining is also desirable but not essential. A strong interest in bioinformatics would also be highly desirable.

Adresse d’emploi :
Lorraine Research Laboratory in Computer Science and its Applications (LORIA), Nancy, France

Document attaché : 202305240952_PROJET-DE-THESE-LORIA-LUE.pdf