A thesis offer in hybrid approaches for robotic applications is to be filled at ISIR - Institute of Intelligent Systems and Robotics of Sorbonne University.
Thesis topic: "Hybrid AI-based approaches for the control of uncertain dynamical systems",
Candidate profile: Holder of a master's degree or equivalent degree in automation, AI / machine learning, robotics and / or modeling with very good skills in at least one of these fields,
Collaboration: with the CNAM (National Conservatory of Arts and Crafts), CEDRIC Lab.,
Application deadline: May 12, 2022.
More information and application: https://drive.google.com/ file/d/1pc8ldNqUL7E4a3fNtN19JcLH2Z7wA95d/view?usp=sharing
Background: A major challenge in robotics is to ensure that the robot is able to achieve what it was designed for with good repeatability, despite varying environmental conditions. The reasons for this variability are numerous, as illustrated by the following examples: variations in brightness affecting the performance of a perception module for an autonomous vehicle, variations in ground adhesion of a terrestrial robot operating in a natural environment, variations in aerological conditions for an autonomous drone, variation in the nature of the interaction forces between a micro-robotic system and its environment. To meet these challenges, the traditional approach in robotics consists of establishing a "nominal" model of the robot from the laws of physics, a model that is supposed to accurately describe its behavior under ideal conditions, and then using feedback loops to correct the robot's deviations due, among other things, to errors related to unmodeled environmental phenomena.
Project description and scientific objectives: The objective of this thesis is to develop hybrid methods, and to validate their effectiveness in the context of robotic applications. Several recent studies have proposed hybrid approaches in order to deal in isolation with the problems of modeling, perception, or control. Regarding the modeling aspects, we can as an example mention [1] which deals with DeLaN networks (Deep Lagrangian Networks) consisting of using neural networks with a Lagrangian structure, or [2,3] which propose models with a physical part (ODE/PDE) supplemented by a part learned by a neural network. Concerning the aspects of perception, we can for example mention [4] which deals with problems of attitude estimation from measurements of inertial power plants and uses a neural network to detect and correct vibrational effects not taken into account by the analytical model of the sensor. With regard to control aspects, the links between reinforcement learning approaches and conventional automatic approaches have been highlighted for several years now (see for example [5]) and "Model-based reinforcement learning" approaches have also been proposed (see, e.g., [6]).
The main objectives of the thesis are as follows:
1- Contribute to the development of new hybrid methods. As illustrated by the bibliographic references mentioned above, the development of hybrid methods is booming and it opens up many perspectives. The first topic that will be addressed in the thesis will be to couple approaches exploiting the geometric structure of the model (e.g. the DeLaN structure mentioned above, but this is not the only possible one), with the augmentation approach [2,3]. The goal is both to exploit strong physical structures inherent in the system, while leaving the neural network part the possibility of modeling and identifying effects that can not be captured by this structure (friction, slippage, disturbances, interaction forces, etc.).
2- Propose a complete approach, which takes into account jointly the aspects of modeling, perception, and control. Hybrid methods proposed in modelling, such as those mentioned above, most often assume that system states are measured perfectly (simulation framework), or use laboratory instrumented demonstrators to provide excellent quality measurements ([2,3]). The problem of perception is then obscured. This issue is also central to reinforcement learning approaches, which are highly dependent on the quality of the measures. In summary, moving from demonstrators in laboratories to real applications (i.e. experimental) requires taking into account the perceptual part that is at the interface between the modeling and control aspects. A first step in this direction will be to study how a model trained on simulated data can adapt to real data – and how the learned term of dynamics can compensate for errors in estimating the perception brick.
3- Validate the approaches developed on robotic applications. If the main objectives of the thesis are methodological in nature, it will be important to evaluate their effectiveness on real use cases of robotics. Here we will be able to rely on the means available at ISIR, with two fields of application already identified: autonomous drone navigation, and micro-robotics applications. Regarding drones, in accordance with the previous Point 2, we will seek to validate on an autonomous navigation task a complete hybrid approach of modeling/estimation of state/control. Regarding micro-robotics, due to the scale effect, the objects to be manipulated tend either to stick to the effectors of the robot or to be propelled with strong accelerations making the manipulation tasks at small scales highly unpredictable. A hybrid approach will make it possible on the one hand to complete the interaction models between the robot effectors and the manipulated objects and on the other hand to adapt the control laws to the experimental conditions for a better success rate of the micro-robotic manipulation tasks.
Profile sought and skills required: The candidate must hold a master's degree or equivalent degree in Automation and/or AI/Machine Learning and/or Robotics and/or Modeling with very good skills in at least one of these fields.
Link to the offer on the ISIR website: https://www.isir.upmc.fr/ contact us/oppotunites/
*Research position in NLP and Dialogue Systems https://jobs.fbk.eu/Annunci/Jobs_A_research_position_in_the_field_of_Natural_Language_Processing_and_Dialogue_Systems_222909971.htm
*Application deadline: 26 May, 2022*
The call is in the context of the expansion and consolidation of AI activities funded both at the local and at the national level and aims at selecting candidates that can help the strengthening of the FBK competences in AI and NLP of the future.
*Job Description* FBK is looking for a candidate to cover the 3-year position with a dynamic, highly motivated, researcher in the field of "Natural Language Processing" for the NLP Unit of Digital Health and Wellbeing (dHWB) Center.
In detail, the candidate will be asked to advance state-of-the-art research in the field of NLP, with particular emphasis on the development of techniques for *conversational agents (*task oriented dialogue systems, collaborative human-machine dialogues, generation of explanations).
Moreover, the candidate is expected to take on research challenges that come from the application of state-of-the-art research both in dialogue systems and in information extraction for healthcare, within the vision of the dHWB Center. On the topics above, the candidate will have the possibility to supervise PhD students and develop their own specific research directions in accordance with the strategies of the Research Unit and the dHWB Center.
The candidate will work in collaboration with other researchers of the NLP Unit and of the dHWB Center, as well as with partners involved in NLP projects. The candidate is also expected to contribute to proposals for funded activities, including reporting and dissemination of results (in both academic and popular venues). Furthermore, we expect the successful candidate to contribute to maintain a strong role of FBK in the Italian and international NLP community.
The purpose of the current call is an opportunity of working into an internationally renewed NLP group and develop their own research path in accordance with the long term strategy of the NLP Unit and the dHWB Center.
*Job requirements* *The ideal candidate should have:*
- PhD Degree in areas related to Computer Science or Computational Linguistics;
- Research Expertise in Dialogue Systems and/or Information Extraction;
- Publication track record in the field of NLP;
- Good Knowledge in application of deep learning techniques in NLP;
- Good knowledge and proficiency of the English language;
- Team working attitude;
- Good communication and relation skills;
- Autonomy in developing research and organising work activities;
- Experience in writing proposals of research projects.
*Additional requirements:*
- Experience in supervision of internship and thesis of bachelor and master level on topics related to NLP;
- Experience in working for international research projects;
*FBK actively seeks diversity and inclusion in the workplace and is also committed in promoting gender equality.*
*Employment*
*Type of contract*: fixed term contract -- full time
*Start date*: preferably from July 2022
*Duration*: 3 years
*Gross annual salary*: about € 39.500
*Workplace*: Povo - Trento (Italy)
*Application* Interested candidates are requested to submit their application by completing the online form (https://jobs.fbk.eu/). Please make sure that your application contains the following attachments (in pdf format):
- Detailed CV including list of scientific publications
- Motivation letter
- Email address contact for two referees
*Application deadline: 26 May, 2022*
Please find attached a thesis topic proposed by the LIS laboratory (DIAMS team), for which we are looking for a candidate.
Towards automatic discovery of areas of interest (ZOI) in the field of maritime transport
Application deadline: 05/06/2022
Applications should be addressed to: [email protected], [email protected] and [email protected]
Detailed CV
Cover letter
Details of the last two transcripts (including M1 and M2),
Letters of Recommendation
A Ph.D. position in Artificial Intelligence and Machine Learning is open at the CRIL lab. This Ph.D. fellowship shall be funded for 3 years from October 2022.
Title: Learning Interpretable Circuits
Overview: In recent years, there has been a growing interest in the design of statistical learning algorithms for interpretable models, which are not only accurate but also understandable by human users. A related and desirable property is explainability, which refers to the computational ability of predictive models to explain their predictions in intelligible terms.
In this setting, decision trees are of paramount importance, as they can be easily read by recursively breaking a choice into sub-choices until a decision is reached. Although the problem of learning decision trees is NP-hard, several optimization approaches have been proposed for identifying a decision tree that minimizes some regularized risk objective (e.g. [1,2,3]).
Beyond decision trees, various classes of (boolean or arithmetic) circuits are also endowed with interpretability and explainability properties. These models are indeed able to efficiently handle a wide variety of explanation queries, by exploiting the decomposability property on the logical ``and’’ gates, and the determinism of ``or’’ gates [4]. Well-known classes of interpretable circuits include ordered decision diagrams, affine decision trees, and decomposable negation normal form representations.
Although the problem of learning circuits has been studied in the literature (e.g. [5,6]), very little is known about learning interpretable circuits. The focus of this thesis proposal is to address this issue, by exploring several directions of research. One of them is to identify the learnability of several subclasses of interpretable circuits. Another important perspective is to exploit constraint-based or MILP solvers for extracting optimal - or quasi-optimal - interpretable circuits. Finally, since decision circuits can easily be extended to semi-ring networks, a final topic is to extend results about binary classification to other learning tasks, such as regression or density estimation.
Profile: The candidate should have a Master or similar degree in computer science. As the thesis proposal lies at the intersection of explainable artificial intelligence, combinatorial optimization, and machine learning, the candidate should have a strong background in some of these topics.
Contacts: Frederic Koriche ([email protected]), Jean-Marie Lagniez ([email protected]) and Stefan Mengel ([email protected]).
I share an open thesis topic within Orange in collaboration with CNRS-LAAS and IRIT:
https://orange.jobs/jobs/offer.do?lang=&joid=111932
Full title: Adaptativity of the response of a connected living space to the interactional and behavioral context of its occupants: design of an intelligent technological infrastructure
Profile sought: Master 2 level or engineer. Sensor data collection. Artificial Intelligence / Machine Learning. Human-Machine Interaction / Adaptation models of interactive systems.
Experience in processing and analyzing sensor experimentation/manipulation data and multi-component software integration, middleware.
If possible experience in developing an experimentation protocol.
Enthusiastic and enterprising with a sense of results, teamwork, rigorous, serious, autonomous. Know how to lead a project. Ability to work orally and in writing in French and English. Languages: English level B2/C1.
Practical details: Type of contract: CIFRE Orange thesis
Laboratory: LAAS CNRS in collaboration with IRIT (Toulouse)
Place of work: IUT Blagnac (Smart Home platform)
Start: October 2021 but fixed-term contract possible before
Thesis supervision: Eric Campo, University Professor (LAAS), [email protected]
Frédéric Vella, Research Fellow (IRIT), [email protected]
Industrial supervision: Jean-Léon Bouraoui [email protected] and André Bottaro, [email protected], Orange
Our School of Computer Science at National University of Ireland, Galway, Ireland is hiring 4 Lecturers Above the Bar/Assistant Professors in Computer Science: 3 Permanent and 1 Fixed-Term.
The positions are in diverse areas of Computer Science including: Cloud Computing, Software Engineering, Cyber Security, Machine Learning, Data Science.
Great opportunity to join a dynamic, research-intensive school.
Apply by June 09th at https://nuigalway.ie/about-us/jobs/
Lecturer Above the Bar/Assistant Professor in Computer Science – 3 Posts , Full Time Permanent, Contract Type B
Applications are invited for an appointment as Lecturer Above the Bar/Assistant Professor in Computer Science at the National University of Ireland Galway (NUI Galway). Up to three new posts will be filled as a result of this recruitment. For these posts, we are seeking candidates who can demonstrate a world-class research profile, or its potential from industry experience, as well as teaching expertise in one in the following specialisms:
A. Cloud Computing; B. Software Engineering; C. Cybersecurity; D. Data Science; E. Machine Learning.
In their applications, candidates are requested to identify which one of these specialisms they wish to be considered for, and provide evidence of expertise in the specialism.
The successful candidates will actively contribute to the University’s strategic vision for CS, and will participate in research, teaching, student project supervision, and programme administration across a complementary portfolio of our undergraduate and postgraduate programmes.
Research contribution is of fundamental importance for these posts. The appointees must be capable of acting as principal scientific investigator on large-scale externally funded projects and will be expected to provide leadership and research supervision to the members of their research groups. They will be expected to attract R&D funding, and demonstrate clear plans to apply for major funding awards, for example, from Science Foundation Ireland and the European Research Council (ERC) Grant programmes. The appointees will also be required to develop national and international collaborations with academia and industry. The new appointees will be expected to disseminate their work through high quality peer-reviewed journals, high-impact conferences, and workshops.
The successful candidates will contribute to teaching of fundamental and applied computer science on our portfolio of undergraduate and taught postgraduate programmes, as well as participating in academic programme management, marketing and outreach activities. The appointed candidates will be expected to develop new curricula, prepare and deliver materials for online and classroom modes of delivery, supervise undergraduate and postgraduate students in their project work, and conduct related administrative duties.
NUI Galway has been inspiring minds since 1845 as a research-led university. NUI Galway is one of the oldest and largest universities in Ireland. The campus community includes over 21,000 students and staff and 110,000 alumni located in over 100 countries across the world. NUI Galway is counted among the Top 300 universities in the world, positioned in the top 2% in QS global rankings. NUI Galway is an international university with global ambition, but with deep roots in the region and nationally. Our university is at the heart of a distinct and vibrant region, renowned for its unique culture, creative industries, medical technologies, marine ecology, tech sector, and innovation.
In Computer Science research and teaching, NUI Galway holds a strong international reputation. In the 2022 Times Higher Education world university rankings by subject, Computer Science at NUI Galway is ranked joint first in Ireland.
Research engineer/Post-Doc confirmed in Automatic Language Processing: Development of a virtual teaching assistant
Type of contract: 1-year fixed-term contract Information
General: Place of work: CNRS/LISN, rue du Belvédère, University Paris-Saclay university
campus Type of contract: FIXED-term contract
Duration: 12 months
Expected date of employment: From June 2022
Work quota: Full-time
Remuneration: According to profile
Desired level of study: Doctor/Engineer
Desired experiences: work in the field of processing
Context We are looking for an experienced research engineer/post-doctoral fellow to work in the LISN laboratory (cnrs-Université Paris-Saclay joint laboratory), with researchers specialized in Automatic Language Processing (NLP).
This research is part of the 18-month maturation program between the company Professorbob.ai, leader in adaptive learning, SATT Paris Saclay (Société d'Accélération du Transfert de Technologies) and cnrs. The posts are located in the premises of the CNRS LISN (Saclay, 91).
This is to work on a virtual teaching assistant project dedicated to education and training, which is the subject of a collaboration between the laboratory and the company working on the development of Professorbob.ai (https://professorbob.ai/)
This assistant will have to be able to help students in their learning: • Answering questions related to course topics • Offering tools for knowledge anchoring • Personalizing learning through "adaptive learning" methods .
The creation of the virtual assistant requires advanced technical knowledge and mastery on models and issues in natural language processing. More specifically, we will focus on the issues of text generation, information retrieval , language evaluation and domain transfer. Recent advances in language processing allow us to consider the construction of such a system, especially through neural approaches for the generation of questions or the search for information.
Unfortunately, while the most efficient models can obtain satisfactory results in the English language , few pre-exist models for the French language. Also, even if there are publicly accessible corpora for the task of generating questions, these corpora only partially correspond to the types of questions desired for a course assistant.
To overcome this lack of data, we are working to set up a corpus of French-language course questions by the date of the start of the contract. The main issues studied in the proposed position will relate to the generation of questions and answers. Question generation. With the previously collected data, the objective is to create/generate questions via question generation models based on neural approaches. To measure the quality of the questions generated by the model, it is desirable to consider innovative metrics.
Indeed, approaches based on n-grams are still largely dominant today[1]. Nevertheless, recently, several approaches have been updated: approaches that do not compare words but contextualized vector representations of subwords (Bert score)[2]; approaches proposing to use context to verify the relevance of generated texts1[3, 4] . It should also be noted that these approaches still achieve worse results today than a human evaluation. Generation and selection of responses. A second important step of the project concerns the selection of response considering a user query
This search for information therefore corresponds to the improvement and specialization of search engines. This approach can be considered both through the use of search engines based on word frequencies or approaches via neural networks. Once the relevant documents have been found, it is necessary to produce a response, either by content extraction (extractive response) and/or by generation (abstract response) [5]. In the case of the generation of answers, it is then necessary to verify the relevance and veracity of the facts transcribed by the model, thus, work on the evaluation of the generation is to be considered [2, 6, 7].
Activities The overall goal of the project is to assist a teacher by helping him answer many of the learners' repetitive questions. It is therefore necessary to learn to answer the questions, based on reliable data, provided by the teachers. Building on recent work in the field of NLP, it is known that it is possible to improve conventional and basic systems for answering questions. However, the data in which the answers will have to be found are not the classic data used in evaluation campaigns, but data related to the discipline being learned.
It will first be asked to process the question and answer data collected during the annotation campaign. The job will therefore be to format and clean the available data. In a second step, the work will focus on the generation of questions, but also on their evaluation.
To do this, it will be necessary to evaluate which models and metrics are the most appropriate, but also to set up an evaluation protocol to validate the proposed approaches.
It will then be necessary to be able to deploy these approaches on the system.
Finally, the approaches to selection/generation of answers will be studied and implemented in order to allow significant improvements to the assistant. It should also be noted that the evaluation issues studied for the generation of questions may also be useful in this last step.
The work will be done in a collaborative framework with 2 other researchers and will have to take into account the team's research axes: transfer learning, continuous learning and conversational AI.
Skills Required
Good command of NLP tools:
Deep Learning models: theoretical knowledge and advanced manipulation of RNNs, Auto-encoders, Transformers (BERT / Roberta / T5,..), etc. especially templates of Question Answering, Question Generation, etc.
Deep Learning/Machine Learning libraries and frameworks like Pytorch, Tensorflow, Keras, NLTK, Spacy, Scikit-learn, etc.
Algorithmics: very good knowledge and practical mastery of classical algorithms on texts, trees, graph - Statistics: knowledge of sampling techniques • Experience in development and debugging in Python
Mastery of the Data Science approach: definition of tasks, definition of performance metrics, technology watch, analysis of scientific publications, implementation, fine-tuning and evaluation of models
Fluent scientific English • Ability to communicate and work in a team Additional skills desirable
Search engines and text processing: indexing, use of ElasticSearch, Lucène / SolR, formalization and search for regular expressions Contact: Anne VILNAT LISN ([email protected]) AND at Professorbob.ai: Samy LAHBABI ([email protected])
(english announcement at https://anr-grifin.telecom-sudparis.eu/post/2022-04-26-phd-position-offer/ )
as part of the ANR GRIFIN project, we are recruiting a PhD student to work in the field of security of next-generation networks (IoT-based use cases) and in particular on the intelligent and automated selection and deployment of countermeasures, and verification of their implementation:
https://www.adum.fr/as/ed/voirproposal.pl?site=adumR&matricule_prop=43017
The thesis is co-supervised by Télécom SudParis, LORIA and LIP6.
The application deadline is June 30, 2022.
The requested documents are as follows: - a CV (up to date)
- a letter of motivation
- a copy of the Master's or engineering degree (if available) as well as M1/M2 statements - letters of recommendation (training manager, researcher, internship supervisor, etc.)
- a copy of an identity document (if you do not have French nationality)
3 PhD fellowships in applied Machine Learning, Information Retrieval and Natural Language Processing
The Information Retrieval Lab of the Department of Computer Science at the University of Copenhagen (DIKU) is offering 3 fully funded PhD Fellowships in applied Machine Learning, Information Retrieval, and Natural Language Processing, commencing 1 September 2022 or as soon as possible thereafter.
The fellows will conduct research, having as starting point the following broad research areas:
a fully-funded PhD in interpretability of applied machine learning;
a fully-funded PhD in overparameterization and generalizability in deep neural architectures;
a fully-funded PhD in web & information retrieval;
We are looking for candidates with a MSc degree in a subject relevant for the research area. The successful candidate is expected to have strong grades in Machine Learning and/or Information Retrieval and/or Natural Language Processing. Successful candidates should have a preliminary research record as witnessed by a master thesis or publications in the area.
The deadline for applications is 19 May 2022, 23:59 GMT +2.
Full job description: https://employment.ku.dk/phd/? show=156455
You will find below a job offer CDD of 18 months for a position of ETL engineer at the University of Grenoble, Laboratory of Computer Science (LIG).
Computer Science Laboratory of Grenoble – University of Grenoble (38) Full-time, 18-month fixed-term contract Remuneration between 22k and 35k depending on experience Level Research or study engineer (BAC + 5) Desired knowledge of semantic web technologies (RDF, SPARQL)
The LIG, Laboratoire d'informatique de Grenoble, is a research structure in computer science recognized nationally and internationally by the University of Grenoble Alpes (Top 5 French universities and top 150 in the Shanghai ranking).
He/She will be attached as an expert collaborator to the SIDESLAB project, funded by the National Research Agency. This project is interested in advancing research in medical education and training through large-scale experiments. SIDESLAB relies on the largest French learning platform for assessment and training: SIDES, used by all medical students.
He/She will be in charge of defining, developing and deploying an access interface to an RDF data lake allowing project partners to define on-demand extraction queries for their large-scale experiments, and this in a "user friendly" approach.
In this context, he/she will have to manage the tracing and injection into the data lake of the data and metadata resulting from the experiments. He/she will actively participate in the necessary evolutions of the ontology at the center of the data lake in RDF format. Profile sought:
• Have if possible experience in ETL design, analysis and development • Know the tools and programming techniques of the semantic web (triple store RDF, SPARQL, etc.)
• Master programming techniques and languages both in Front and Back with experience in application development and deployment
• Know the world of learning platforms (Moodle knowledge desirable)
• Have an appetite for modeling
• Know how to transcribe a business request in technical data
• Have good interpersonal skills and the habit of teamwork
• Know the methods of Agile project management Send your Curriculum vitae accompanied by a cover letter to [email protected] before June 30, 2022.