Stage M2. Few shot learning and Template Matching for Nano Materials
Pollen metrology
This internship aims to evaluate the capacity of state of the art algorithms to provide an object detection model based on few examples. Current approaches at Pollen, use both classic machine learning and deep learning to provide models to the users. Even if we aimed to reduce the number of training data, the performances are limited when the training dataset is reduced to few training examples.
In our context, nano materials are coming mainly from the semiconductor industry. These objects are highly organized and similar between them but their structure is complex as several materials are involved. Another type of object comes from chemistry. These objects display a simple geometry but they are less similar between each other, as they are made with chemical processes their appearance can be greatly modified and their pose is various.
The contribution of this internship is to evaluate new approaches in order to estimate their potential in chemical and semiconductor applications for a future integration in the software. The first family of methods relies on Few Shot Learning. These methods are trained using a large dataset and then few examples are then given for the detection. Current performances of these algorithms may be a blocking factor for the integration in the software. Other approaches that may be of interest are based on templates. These approaches use descriptors for the template and the image to find the template in the image. New approaches developed more robust features for the matching. The main challenge is to being able to handle deformation of the objects through the images.
Link:
https://pollen-metrology.com/
Internship. Automation of QA
Pollen metrology
Pollen Metrology is developing the 3rd generation of the proprietary software Platypus Smart with a client-server architecture and an API interface. This new generation has new QA requirements which includes the implementation of the cutting edge technologies of testing.
In order to structure and develop its Quality department (QA) Pollen is looking for an intern in automation and/or Web Services testing. The desired skills for QA are attention to detail and rigor. These competences are oriented to find all discrepancies between product and customer requirements. The QA represents the final user, hence the product has to be compliant with the highest standards for customer use and production line.
The maturity of the software (Gen 3) and its architecture requires automated tests in order to ensure the quality of the key user functionalities. This internship will be oriented mostly on the conception, definition and maintenance of automated test repositories with a purpose of functional (user oriented and API ) and non-functional (robustness) testing.
The internship will be organized in two parts:
build and maintain a test repository of functional tests, validating the most common behaviors of the application to test with different parameters and variables
build from scratch a functional and non-functional repository dedicated to a specific customer, particularly on the interface between the application developed by Pollen and the customer infrastructure.
The contribution of this internship is to help improve the quality of the developed application, taking into account the exigence of the business sector particularities and requirements.
Link:
https://pollen-metrology.com/
Researcher (PhD-level) in the area of wireless communications.
Huawei technologies france
Within Huawei’s Mathematica and Algorithmic Laboratory in Paris, the candidate is expected to contribute to forward-looking research in the general area of radio access networks, including the physical and MAC layers, and future network architecture and applications. The role includes developing innovative techniques and demonstrating them analytically and through proof-of-concept implementation, producing patents and reports, publishing in scientific conferences and journals, performing joint research projects with academic partners.
Required skills
- Solid background in wireless communications and information theory
- Advanced knowledge of mathematics and statistics
- Proficiency in Matlab and Python, and coding in general
- Good scientific writing and oral communication skills
- Familiarity with machine learning concepts and tools
Sought profile:
- The candidate should hold a doctoral degree.
- The position is based in Boulogne Billancourt (Paris area). The candidate should be able to legally work in France.
- The position should be filled ideally by the end of 2020.
For further information, please contact Maxime Guillaud, maxime.guillaud@huawei.com.
PhD Application
Huawei technologies france
A graduate researcher (phd) position is available at the paris research center of huawei technologies, france
The hired candidate will work on new waveform design and optimization for beyond 5g networks and the project will be carried out at the wt lab under the supervision of dr. kamel tourki.
The proposed project will apply communication theory, signal processing, and machine learning tools to the analysis and optimization of the proposed waveform candidates.
More details will be provided to the hired phd student (due to confidentiality).
Requirements:
- msc, meng degree in telecom/electrical/computer engineering, or related fields.
- strong experience in signal processing and communications, mimo, multiuser detections.
- prior experience on ofdm and/or related waveforms (fbmc, ufmc, wola etc.) is necessary.
- strong mathematical background.
- ability to program and perform simulations in matlab and/or python.
- outstanding oral and written english skills.
- motivated, enthusiastic, positive person.
Preferred qualifications:
- backgrounds in 4g and 5g networks and standards.
- prior research experience and publications.
- background in machine learning and artificial intelligence.
- experience in testbed development.
how to apply:
email to kamel.tourki@huawei.com
email subject: 'phd application - ' together with your cv, academic
(undergraduate and master) transcripts, master/pfe thesis summary.
R&D Compiler Engineer Internship
Huawei technologies france
Develop compiler optimization techniques and related software components for our internal
customers, with special dedication to AI-oriented compilation toolchains.
• Study target architectures such as Huawei AI chips to establish efficient optimization strategies.
• Work closely with other teams involved in the AI toolchain from high-level languages to chip
design for determining requirements and relevant contributions to improve the competitiveness
of Huawei products.
• Participate to MindSpore AI framework’s Open Source ecosystem, both in terms of R&D and
of community life.
• Conduct research on compiler optimization, including automatic parallelization, data locality
optimization, code generation, vectorization, memory management, scheduling, etc.
• Disseminate research results in major scientific venues and in internal community of experts.
Generate intellectual property when appropriate.
Watch on latest international research achievements on parallelism and compiler optimization,
and evaluate their value to solve Huawei challenges. Accurately grasp the long-term technological trends in the field.
• Follow the directions of the team, create and maintain strong research collaborations with
top European academic research institutes and industry partners, supervise Master and PhD
students if needed.
Post doctorat : Génération automatique de code pour l’algèbre linéaire rapide sur GPU
ONERA
L'ONERA travaille à la génération automatique de code pour un futur logiciel de simulation en mécanique des fluides. Par ailleurs, Google AI travaille à une approche générique pour la génération de code qui consiste à utiliser une approche hiérarchique à plusieurs niveaux de représentation ("intermediate representation").
Ce travail en collaboration entre l'ONERA et Google AI, visera à la résolution rapide de problèmes d'algèbre linéaire issus des équations de la mécanique des fluides en écoulements compressibles discrétisées sur des maillages structurés. Nous chercherons à générer automatiquement un code optimisé pour l'exécution sur GPU. Pour cela nous mettrons en place et étendrons potentiellement le formalisme du dialecte LinALG du framework MLIR sur des problèmes d'algèbre linéaire de difficulté croissante. Pour guider la génération automatique ultérieure, nous devrons réaliser une analyse des goulots d'étranglement à chaque niveau de parallélisme (warps, blocs, threads, vectorisation). Nous étudierons les décisions d'optimisation qui peuvent être prises pour générer du code, les choix des différents niveaux
de parallélisme, les choix d’ordonnancement des boucles, les choix de tuilage (cacheblocking) ou de vectorisation de boucles. Après l'étude de cas simples, nous étendrons l'approche à la résolution par méthode de Gauss-Seidel (itération de l'application à un vecteur d'une décomposition LU de la matrice initiale), dans une approche de type wavefront qui sera adaptée aux spécificités des GPU.
Profil et compétences recherchées Docteur en sciences dans le domaine de la simulation numérique ou des langages de programmation/compilation, Expérience de la programmation parallèle en mémoire partagée sur CPU multi-coeurs ou GPU,
Compétences en algèbre linéaire numérique fortement appréciées.
INGENIEUR CFD (CDD 18 mois)
ONERA
Vous participez aux travaux de l'unité Conception et production de Logiciels pour les Ecoulements de Fluides (CLEF) du DAAA. La mission de cette unité est de réaliser des grandes plateformes
logicielles de simulation en Mécanique des Fluides, aptes à répondre aux besoins de recherche et d'applications de l'ONERA, de laboratoires de recherche partenaires et des utilisateurs de l'industrie.
Dans cette unité, vous prendrez en charge des travaux au sein du logiciel de simulation en aérodynamique elsA ou de son successeur. Ces travaux concernent le développement de conditions
aux limites de non-réflexion qui ont pour objectif de limiter les réflexions artificielles de chocs ou d’ondes sur des frontières ou des conditions aux limites. Vous vous inscrirez dans un effort important mené en collaboration entre plusieurs unités de l’ONERA et des membres de Safran sur l’amélioration de ce type de conditions aux limites. Le travail comprendra des travaux bibliographiques, le développement de modèles analytiques, ainsi que des développements au sein
du logiciel elsA/son successeur et de sa chaine logicielle, et enfin des validations sur des cas tests simplifiés et des cas tests d’échelle industrielle.
PROFIL
Docteur et/ou ingénieur/docteur avec spécialisation en simulation numérique
Goût pour la recherche (méthodes numériques, développement algorithmique, modélisation physique…)
Expérience souhaitée du développement de logiciels de calcul scientifique en environnement HPC de préférence dans le domaine de la CFD
Compétences techniques requises : Fortran (>=90), C++ (>=11), MPI,Python, Numpy
Link:
https://www.onera.fr/fr/rejoindre-onera/offres-emploi/detail?jobId=907&jobTitle=INGENIEUR%20CFD%2018%20mois
DEVELOPPEUR LOGICIEL( CDD 12mois)
ONERA
Vous participez aux travaux de l'unité Conception et production de Logiciels pour les Ecoulements de Fluides (CLEF) du DAAA. La mission de cette unité est de réaliser des grandes plateformes
logicielles de simulation en Mécanique des Fluides, aptes à répondre aux besoins de recherche et d'applications de l'ONERA, de laboratoires de recherche partenaires et des utilisateurs de l'industrie.
Dans cette unité, vous prendrez en charge des travaux en lien avec le logiciel de simulation en aérodynamique elsA. Vos travaux se porteront sur les bases de validation et de non régression de ce logiciel. Ces deux bases possèdent actuellement un environnement et un mode de fonctionnement différent, et votre objectif, en coordination avec les membres de CLEF, sera d'harmoniser les outils permettant d'utiliser ces bases. Nombre de cas (au sens des configurations testées) sont présents dans les deux bases, pourtant les scripts utilisés comme la façon de lancer les cas sont différents.
Ces travaux doivent également permettre de réorganiser la base de non régression afin de pouvoir lancer des "suites" de cas en fonction du périmètre recherché (cas elsA python purs, cas elsA CGNS,
cas de couplage, ...), ce qui s’avère fastidieux aujourd'hui. Enfin cette restructuration devra se faire dans l'optique de pouvoir lancer ces bases de manière automatique et autonome pour répondre au
besoin de lancement commandé par action (commit) ou programmé (lancement quotidien).
PROFIL
Ecole d’ingénieur et/ou doctorat universitaire
Goût pour les méthodes numériques, le développement algorithmique, la modélisation physique
Expérience souhaitée du développement de logiciels de calcul scientifique en environnement HPC de préférence dans le domaine de la CFD
Link:
https://www.onera.fr/fr/rejoindre-onera/offres-emploi/detail?jobId=1103&jobTitle=INGENIEUR%20DEVELOPPEUR%20LOGICIEL%2012%20mois%20H%2FF
Post-doc Machine Learning
Pollen metrology
Pollen Metrology is a deeptech company specialized in the creation of intelligent software (AI) for the production of high-performance materials. Pollen has developed a unique artificial intelligence technology to automatically analyze all types of images from scanning or transmission electron microscopes. In the context of the launch of a new product range, particularly in the United States and Asia, Pollen is recruiting new collaborators to strengthen its research team in order to work with our customers in the semiconductor ecosystem.
Your missions
You will be in charge of prototyping and developing machine learning algorithms mainly for classification and analyzing metrological data of various types. You will use frameworks such as TensorFlow or PyTorch, and internal technologies that you will help develop.
You will also participate in the integration of your algorithms in our Platypus Smart product, support in the implementation of integration tests, and you will have the opportunity to validate your developments directly with our customers (USA, Europe, Asia).
Required skills
-Very good knowledge of deep learning.
-Good knowledge of classification tasks.
-Good knowledge of statistics.
-Good knowledge of Python (TensorFlow or PyTorch, scikit-learn, etc).
-Knowledge in image processing.
-Fluency in English.
-You are eager to learn.
Link:
https://pollen-metrology.comi
PhD CIFRE Offer at Huawei in the area of statistical and machine learning
Huawei technologies france
A graduate researcher (PhD) position is available at the Paris Research Center of Huawei Technologies, France
The hired candidate will work on developing theory and algorithms for distributed statistical learning settings, with a particular application to the problem of AI at the Edge in wireless cellular networks.
The PhD thesis will be carried out under the joint supervision of Denis Trystram, ENSIMAG from Université Grenoble-Alpes, Abdellatif Zaidi from Huawei’s Paris Research Center and guyen Kim Thang from Université Evry.
Requirements:
• MSc and/or MEng degree in statistics, theoretical computer science, electrical engineering or related
fields
• Strong mathematical background and a wish to investigating/understanding theoretical aspects of machine
learning (especially distributed settings of it) as well as practical approaches.
• Knowledge in communication theory and networks is a plus.
• Ability to program and perform simulations in Python
• Outstanding oral and written English skills
How to apply:
Email to abdellatif.zaidi@huawei.com
Internship Position on Mobile Cellular Traffic Analysis with Machine Learning
Huawei technologies france
The Global Technical Service research team of the Mathematical and Algorithmic Sciences Lab, Huawei France Research Center, located in the Paris area, is looking for highly motivated candidates for an intern on Mobile Cellular Traffic Analysis with Machine Learning.
Scope
The past decade has witnessed the rapid growth of global mobile cellular traffic demands due to the popularity of mobile devices. While accurate traffic prediction becomes extremely important for stable and high-quality mobile service, the performance of existing methods is still poor due to three challenges: complicated temporal variations including burstiness and long periods, multi-variant impact factors such as the scenario/use case and day of the week, and potential spatial dependencies introduced by the movement of population.
The goal of this intern is to study methods based on machine learning that work on soft information (scenario/use case, base station 3D positions/locations) to identify group of cells whose network key performance indicators are related and to exploit this information to characterize and predict the network traffic. In additional, labels in the data-set can be out-of-date and machine learning can be used to identify anomalies and update this information.
Specific Requirements
Ideal candidates should be in the final year of a Master degree in Telecommunications. They should have a background in Machine Learning, telecommunications, and applied mathematics.
English: Operational
Contacts
- Huawei FRC: Dr. Antonio De Domenico (antonio.de.domenico@huawei.com)
To apply please send a complete CV and a motivation letter.
Transfer learning researcher
Huawei technologies france
Huawei Noah's Ark Lab was established in 2012 and is one of the world's leading industrial AI laboratories with offices in Shenzhen, Hong Kong, Beijing, Shanghai, Xi'an, North America, and Europe. Driven by long-term, high impact projects and innovative research, we are committed to promote innovation and technological development in the AI field and provide technical support for Huawei’s products and services. We are now looking for new talents to join our Paris team , specializing in reinforcment learning and control, time series forecasting, anomaly detection, structured data, and optimization. New internship positions are opened for transfer learning.
- Role Description and Responsibilities:
Perform research on one of the preferred topics, publish high quality research results on top conferences, improve the visibility for lab and company in academic communities.
Requirements:
1. Master/PhD degree in computer science, statistics, applied mathematics, information engineering, electronics engineering or have equivalent research experiences;
2. Good knowledge in transfer learning (especially in domain adaptation, multi-task learning, semi-supervised learning and weak supervised learning, small sample learning), causal inference and time series data analysis
3. High quality publications in machine learning related conferences or journals;
4. Strong programming skills, experienced in one or more programming languages including but not limited to: Python (preferential), C/C++, and Java.
Efficiency of mobile gaming system for selected 3D Graphics effects
Huawei Technologies
Internship Summary:
HiSilicon Kirin chipset team proposes exciting internship opportunities for future electronic and computer science engineers aiming to innovate for industry leading mobile chip creation. Internships are available in Sophia-Antipolis, France.
Today premium mobile phones embed high performance 3D Graphics Processing Unit (GPU) enabling mobile game vendors to embed photo-realistic computer graphics effects. The internship goal is focused on surveying, analyzing and evaluating the efficiency of mobile gaming systems for such algorithms. Interest in 3D Graphics technology and processor architecture will be key for successful internship. This internship may lead to full time engineering position offer.
Internship Duration:
6 months.
Responsibilities and tasks:
The successful candidate will have the following responsibilities during the internship:
- Analyze the structure of selected popular mobile 3D Graphics games on an existing platform. Identify, select and analyze photo-realistic effects which are commonly used (lighting, shadows, Motion blurr, depth of field, Bokeh, bloom, etc.).
- Create a performance evaluation framework for the algorithm. Analyze resulting performance on the system. Identify bottlenecks. Propose improvements.
Position Qualifications:
- Last year Master degree student in Electronics/Computer Science.
- Fluent in English, efficient learner, good in communication.
- Good programmer (C, C++, Python, etc).
- Interest in 3D Graphics and GPU architectures.
- Interest in using 3D Graphics APIs (Vulkan, DirectX, etc.).
- Familiarity with Android/Linux is a plus.
Ray tracing 3D Graphics algorithm for mobile GPU system
Huawei Technologies
Internship Summary:
HiSilicon Kirin chipset team proposes exciting internship opportunities for future electronic and computer science engineers aiming to innovate for industry leading mobile chip creation. Internships are available in Sophia-Antipolis, France.
Today gaming desktop and announced gaming console use Ray Tracing algorithm for their 3D Graphics titles. The internship is focused on selecting, designing and evaluating ray tracing 3D Graphics algorithm for mobile Graphics Processing Unit (GPU) systems. Interest in programming 3D Graphics and processor architecture will be key for successful internship. This internship may lead to full time engineering position offer.
Internship Duration: 6 months.
Responsibilities and tasks:
The successful candidate will have the following responsibilities during the internship:
- Survey existing 3D Graphics ray tracing algorithms. Estimate and rank their complexity for mobile systems. Select one of them for implementation.
- Create a performance evaluation framework for the algorithm. Analyze resulting performance on the system. Identify bottlenecks. Propose improvements.
Position Qualifications:
- Last year Master degree student in Electronics/Computer Science
- Fluent in English, efficient learner, good in communication.
- Good programmer (C, C++, Python, etc).
- Interest in 3D Graphics and GPU architectures.
- Interest in using 3D Graphics APIs (Vulkan, DirectX, etc.).
- Familiarity with Android/Linux is a plus.
Hybrid image processing pipeline learning strategies
Huawei Technologies
Objective: Modern image processing systems are composed of traditional image processing algorithms as well as CNN based blocks. Traditionally, the machines learning based components of the pipeline are trained using supervised learning. However, it is difficult to obtain suitable training data for deeply embedded modules and the overall performance is sub-optimal when training of components is performed independently. However, traditional image processing algorithms are non-differentiable making end-to-end training of the system complex
Responsibilities:
Under the supervision of machine learning and image processing experts, the intern will have the following responsibilities:
- study different state of the art techniques for hybrid pipeline training & parameter optimization
- benchmark different methods on a simple "toy" pipeline
- define, implement & test a training strategy for a typical isp pipeline
Position Requirements:
- Very good practical knowledge of image and signal processing, and machine learning techniques and frameworks
- Good programmer, can use development tools efficiently (e.g. Python/Matlab/C/C++) and work autonomously
Improve color enhancement for complex photography and video scenes using classical and AI/CNN techniques
Huawei Technologies
The Objective of this internship is to leverage cutting edge research to improve the color enhancement algorithms of an Image Signal Pipeline in very complex scenes (low light or HDR scenes for example) for video and capture. The challenge is to improve the colors in an image psychovisually so that they are both correct and subjectively pleasing while also keeping the computational cost to a minimum. This is an opportunity for R&D in a new advanced high-tech domain using image processing and machine learning techniques.
Responsibilities:
• Study start of the art classical and AI based color enhancement research (e.g. papers from top universities worldwide)
• Get familiar with commercial color enhancement algorithms (DarkTable…)
• Generate a database of SDR and HDR images with a latest generation Huawei phone. Start with MIT-5K database. Search for good subjective based (professional photographer) CNN training databases.
• Develop a Deep Reinforcement Learning algorithm and compare/benchmark with other machine learning algorithm or classical techniques or combinations that can improve the accuracy of the colour. Starting point would be the MIT-5K datasets.
• Understand the theory of the classical or machine learning algorithms and architecture for color enhancement (which layer/stage does which color task etc…)
• Simplify the classical or machine learning algorithm to achieve an implementable solution in a smartphone environment
Position Requirements:
- Very good practical knowledge of image and signal processing, and machine learning techniques and frameworks
- Basic knowledge of camera sensors
- Good programmer, can use development tools efficiently (e.g. Python/Matlab/C/C++) and work autonomously
Smartphone application for interactive photo / video post processing
Huawei Technologies
The objective of this internship is to develop a user friendly Android application where a typical user could easily control the image processing pipeline to match his/her aesthetic choices.
Responsibilities:
Under supervision by members of our ISP team, he intern will have the following responsibilities:
- Get familiar with state of the art image enhancement techniques, similar to what photographers would use when post processing their images with professional SW tools
- Study related image enhancement and semi-automatic object selection algorithms
- Implement a prototype application on a smartphone
The candidate will be closely integrated into a team of image processing algorithm and architecture experts and will thus benefit from the knowledge and experience of the inter-disciplinary team based in Sophia Antipolis.
Position Requirements:
- Passionate by photography & good knowlegde of commercial photo post processing tools (e.g. darktable, lightroom, gimp, photoshop, …)
- Basic understanding of a typical image enhancement pipeline (RGB2RGB)
- Basic knowledge of Android / iOS smartphone application development
- Good programmer, fast learner and creative
Study the influence of image preprocessing on the accuracy of computer vision algorithms
Huawei Technologies
The objective of this internship is to study the influence of image pre-processing on computer vision algorithm effectiveness. Indeed, the RAW image from the camera typically undergoes a preprocessing phase done by the ISP which aims to produce a high quality YUV image. Traditionally, ISP algorithms and tuning parameters have been optimized for human viewers and are typically inefficient for machine vision. Indeed, algorithms could be too complex for the needs of machine vision algorithms and therefore require too much compute power. Or they may alter the information contained in the scene to produce a more appealing picture to a human viewer which would negatively affect computer vision algorithms.
Responsibilities:
Under supervision by members of our ISP team, he intern will have the following responsibilities:
- With the help of local experts, define the test methodology (test image database, computer vision algorithms to use, benchmark)
- Implement several state of the art computer vision algorithms & accuracy benchmark & integrate them with a simple ISP pipeline
- Study the effect of algorithm alterations (e.g. remove processing blocs, change tuning parameters, replace the noise filter or other modules with a different ones, retrain CNN based algorithms on altered images, change the compute precision, …)
- Summarize the conclusions
Position Requirements:
- Basic knowledge ISPs, image processing algorithms, computer vision algorithm, machine learning, CNNs and camera sensors
- Good programmer, can use development tools efficiently (e.g. Python/Matlab/C/C++) and work autonomously
- Highly motived to surpass herself/himself, eager to learn, self-driven and excellent communication skills
Use Machine Learning to reduce image signal processing algorithm complexity
Huawei Technologies
Objective:
With the fast evolution of smartphone camera technologies and application scenarios, the ISP is becoming an increasingly crucial and complex component, and image processing algorithms are constantly updated as research makes progress and creates possibilities for new camera features but those need to be further optimized to fit in thermal and area budgets.
Responsibilities:
• Study start of the art (e.g. papers from top universities worldwide) black box optimization and exploration techniques,
• Get familiar with the current exploration techniques and tools used internally, to understand the limitations and bottlenecks,
• Integrate additional exploration strategies so as to achieve better optimization consisting of reduction of variable precision, replacement of operators with imprecise operators,
• Improve simulation speed of current techniques (optimize processing code, add more parallelism to the code),
• Remove some of the tool limitations
Position Requirements:
Very good practical knowledge of optimisation techniques and frameworks (such as Bayesian Optimisation, Evolutionary algorithms, …)
Good programmer, can use development tools efficiently (e.g. Python/C/C++/Java) and work autonomously
Stage M2 statistiques Identification d'interactions Gène×Environnement par méta-analyse en génétique des plantes
INRAE
Le stage consistera à développer une stratégie de
méta-analyse prenant en compte la similarité entre panels étudiés et à évaluer les performances de la stratégie proposée (en terme de puissance de détection des marqueurs in fluant sur le caractère étudié, et de contrôle de faux positifs) à l'aide d'une étude de simulations.
La procédure d'analyse devra être implémentée de manière efficace pour être appliquée à de gros jeux de données (panels d'individus génotypés pour ≈ 106 marqueurs, évalués dans une vingtaine d'environnements). Dans un deuxième temps la méthodologie sera appliquée
à plusieurs jeux de données publiques portant sur une espèce modèle (Arabidopsis) et deux espèces d'intérêt économique (maïs et blé).
Link:
https://www6.inrae.fr/mia-paris/content/download/4924/45796/version/1/file/StageMetaAnalyse.pdf
M2 intership Applied Statistics Is what we know about the structure of pollination networks robust to sampling effects ?
INRAE
The aim of this internship is to develop statistical tools to tackle sampling effects in the analyses of plant-pollinator interactions. This means accounting for different species detectability and abundances in the inference of statistical model (Latent Block Models) and in the computation of the metrics (modularity and nestedness). The observed networks will be then considered as incomplete observations of « true » plant-pollinator networks. Statistical models for these observation processes should be proposed and the inference of latent block models should incorporate this additional layer of uncertainty on the
data. Moreover, the computation of the metrics will come with an uncertainty assessment derived from observation processes.
These tools will be used to re-analyze a set of about 100 plant-pollinator networks and to investigate how accounting for the sampling effect provides a new insight in the structure of these networks.
More specifically, studies often show that plant-pollinator networks are nested but this structure simply arises from differences in species abundance and sampling effort. This hypothesis could then be tested with this new tool.
Moreover, these methods could be used on data from a citizen science program aiming at monitoring plant-pollinator interactions in France, with more than 300,000 records across seasons.
Link:
https://www6.inrae.fr/mia-paris/content/download/5078/47192/version/1/file/sujetStageEchantillonnageReseau.pdf
M2 intership Applied Statistics Is what we know about the structure of pollination networks robust to sampling effects ?
INRAE
The aim of this internship is to develop statistical tools to tackle sampling effects in the analyses of plant-pollinator interactions. This means accounting for different species detectability and abundances in the inference of statistical model (Latent Block Models) and in the computation of the metrics (modularity and nestedness). The observed networks will be then considered as incomplete observations of « true » plant-pollinator networks. Statistical models for these observation processes should be proposed and the inference of latent block models should incorporate this additional layer of uncertainty on the
data. Moreover, the computation of the metrics will come with an uncertainty assessment derived from observation processes.
These tools will be used to re-analyze a set of about 100 plant-pollinator networks and to investigate how accounting for the sampling effect provides a new insight in the structure of these networks.
More specifically, studies often show that plant-pollinator networks are nested but this structure simply arises from differences in species abundance and sampling effort. This hypothesis could then be tested with this new tool.
Moreover, these methods could be used on data from a citizen science program aiming at monitoring plant-pollinator interactions in France, with more than 300,000 records across seasons.
Link:
https://www6.inrae.fr/mia-paris/content/download/5078/47192/version/1/file/sujetStageEchantillonnageReseau.pdf
Post-doc 18 mois - Couplage des régimes thermiques rivière/zone humide/lac et validation des réseaux de capteurs mis en place
Université Clermont Auvergne
La personne recrutée en post-doctorat pour une durée de 18 mois travaillera en interaction avec les deux laboratoires GEOLAB et LMGE qui sont investis dans l’équipement et le monitoring du site d’Aydat. Il s’agira dans un premier temps de tester la validité et la performance des réseaux de capteurs mis en place sur la rivière Veyre, la zone humide, le lac d’Aydat. Ces outils de monitoring haute fréquence recouvrent plusieurs domaines et concernent à la fois la température, l’hydrologie, la météorologie et la dynamique du phytoplancton. Le croisement entre les relevés ponctuels de terrain effectués régulièrement par les membres des laboratoires impliqués et les données issues de ces capteurs permettra en partie de valider le dispositif déployé. Ensuite, l’analyse fine de l’ensemble des données hautes fréquences recueillies sur plusieurs mois servira à la production de modèles spatialisés reflétant le couplage des régimes thermiques des trois sous-systèmes (rivière/zone humide/lac). L’évaluation de l’effet thermique de la zone humide sur le fonctionnement du lac d’Aydat constituera un point important du travail. L’un des objectifs à long terme est de déterminer la part de la variabilité thermique dans les dynamiques phytoplanctoniques et plus particulièrement concernant les proliférations de cyanobactéries.
Routing Algorithms for IP Networks
Huawei technologies france
In the recent years, a new routing paradigm called Segment Routing (SR) have emerged and is now being deployed in operational networks. Segment Routing provides complete control over the forwarding paths by combining simple network instructions maintained at the ingress node. These instructions consist in a list of inflexion points (links / nodes) that can be reached segments by segments using different sets of shortest paths (typically, a set per traffic-engineering objective – low latency for real-time traffic for instance).
Current extensions are being developed to handle 1) time-critical traffic with tight QoS requirements in terms of packet loss, latency, etc, and 2) various requirements in term of node / link inclusions. In this context, the main challenge is to design a scalable and stable routing solution. The main objective of this internship will be to design algorithms based on OR techniques (linear programming, robust optimization) for path computation and routing stability.
Engineer in Heterogeneous Computing SoC for Networking
Huawei technologies france
The research team in Grenoble focus on next-generation ARM-based SoCs for network equipment (firewall, access router etc), including the processor, interconnect and memory subsystems and hardware accelerators. The main target is to identify the bottleneck, optimize the multiprocessor SoC architecture to improve the performance, and verify the ideas based on the system simulation.
Wearable device testing engineer
Huawei technologies france
Conduct testing for various wearable mobile health devices.
1. Investigating testing and regulatory procedures for specific health features.
2. Draft testing / evaluation protocol.
3. Build up testing envirement.
4. Conduct perfomance evaluation for certain device / prototype against comparative or golden standard creteria.
1. Bachelor or master of science candidate.
2. Backgroud of electrical engineering or computer science, or similar discipline.
3. Preferred for medical engineering or wearable device project background.
Wearable device testing engineer
Huawei technologies france
Conduct testing for various wearable mobile health devices.
1. Investigating testing and regulatory procedures for specific health features.
2. Draft testing / evaluation protocol.
3. Build up testing envirement.
4. Conduct perfomance evaluation for certain device / prototype against comparative or golden standard creteria.
1. Bachelor or master of science candidate.
2. Backgroud of electrical engineering or computer science, or similar discipline.
3. Preferred for medical engineering or wearable device project background.
Stage de fin d'études
Liebherr
Intelligence Artificielle pour le contrôle du processus de fabrication additive
Link:
https://www.math.univ-toulouse.fr/spip.php?article943
Stage de fin d'études
ONERA
Probabilités et Statistique pour l’industrie : estimation d’indices de Shapley pour l’analyse de sensibilité fiabiliste
Link:
https://www.math.univ-toulouse.fr/spip.php?article943
Stage de fin d'études
CEA
Inversion fonctionnelle des incertitudes de modèle en simulation
Link:
https://www.math.univ-toulouse.fr/spip.php?article943
Stage de fin d'études
Institut de Mathmatiques de Toulouse
Probability distribution of excursion hights in a Lindley process
Link:
https://www.math.univ-toulouse.fr/spip.php?article943
Thèse
IRT Saint-Exupéry
Développement d’un formulation bi-niveau sous incertitudes pour l’optimisation de conception multidisciplinaire d’un avion
Link:
https://www.math.univ-toulouse.fr/spip.php?article946
Graphical User Interface for the IFinder Verification Tool
Huawei technologies france
Trigger the generation of invariants and return the results,
Provide a view to configure the different verification parameters,
Retrieve and display the invariants and verification results generated by the core module of the tool,
Guide the user to follow the verification flow,
Mastering of the Java programming language and the design of graphical user interfaces.
Familiarity with the Eclipse environment and the development of plugins.
Good English communication and writing skills.
Stage de fin d'études en Data Science
Metafora Biosystems
Le stage aura pour mission d’optimiser l’algorithme au cœur de notre plateforme technologique. Cet algorithme permet une analyse topologique de données complexes (en grandes dimensions, bruitées). Il conviendra d’étudier l’impact des hypers-paramètres de l’algorithme sur les résultats d’aide au diagnostic, de trouver l’espace optimal de représentation des données et de tester différents calculs de distance pour l’analyse topologique. Une fois cette étude réalisée, les modifications apportées à l’algorithme seront mises en production.
Le candidat pourra s’appuyer pour ses tests sur une base de données hébergée sur les serveurs de Métafora.
Profil du candidat :
• Titulaire d’un Master 2 en mathématiques / data science (ou équivalent),
• Avec une expérience en programmation (python) et une connaissance des bonne pratiques de développement (utilsation des outils de versionnage de code gitlab, svn etc. , révision par pers ,...)
• Expertise dans le domaine
◦ des statistiques multidimensionnelles,
◦ de la classification et segmentation de données,
◦ de la modélisation supervisée et non-supervisée,
• Familiarité avec
◦ le traitement de mégadonnées,
◦ les bases de données,
◦ les bibliothèques logicielles de Machine Learning,
• Anglais scientifique et technique minimum,
• Esprit d’équipe,
• Curiosité scientifique/technique,
• Autonome et faisant preuve de rigueur et d’initiative, enthousiaste, force de proposition.
Rémunération : 800 euros brut
