Post-Docs Offers

Democratizing AI

2-year post-doc position for the HUMANIA ANR project, aiming to further the democratization of AI and harvest the results of challenges in the series AutoDL and meta-learning. Full time job located at Université Paris-Saclay; telecommuting possible. PhD in artificial intelligence, computer vision, machine learning, or related discipline, required.
Application deadline: April 30, 2022
Date of taking office: June-September, 2022
More information: https://guyon.chalearn.org/available-positions

Causality for Program Synthesis
A 2-years post-doc position is opened at TAU, as part of the INRIA Project Lab HyAIAI (Hybrid Approaches for Interpretable Artificial Intelligence - https://project.inria.fr/hyaiai/). The position will be split between the TAU team at INRIA Saclay (base location) and the LACODAM team at INRIA Rennes (frequent visits from Saclay, all expenses covered of course). The work will tackle learning causality from data, as a path to better explainable models, building on TAU expertise using Deep Generative models. LACODAM will provide the use-case in progam synthesis: real-time recommendation for computer usage, learned from logs of user activity.

• Details: available here
• Profile: The successful candidate holds a PhD in Machine Learning. Good programming skills are required, in Python with Scikit Learn and TensorFlow or PyTorch,
• When: from Dec. 2019
• Duration: 2 years
• Contact: Marc Schoenauer (Marc dot Schoenauer at inria dot fr) and Michele Sebag (sebag at lri dot fr)
  Please send a CV, a statement of research interest, a link to the PhD dissertation and available publications, and a list of three references.

Impact of Quality of Working life on Company Performances ?
A post-doc position is opened at the crossroad of Data Science and Humanities; the goal is to investigate the relationships between quality of working life (based on nation-wide polls) and company performances (based on data from the Ministry of Industry). The position is funded by
University Paris-Saclay.

• Profile: The successful candidate holds a PhD in Machine Learning or Econometrics, Management or Sociology of Labor. Good programming skills are required.
• When: from June 2016
• Duration: 1 year
• Contact: Philippe Caillou (caillou at lri dot fr) and Michele Sebag (sebag at lri dot fr)
  Please send a CV, a statement of research interest, a link to the PhD dissertation and available publications, and a list of three references.

Un post-doc est ouvert en Apprentissage Statistique pour les Sciences Humaines et Sociales, portant sur l'analyse des relations entre qualité de la vie au travail et performance des entreprises. Les données (disponibles) sont issues des Ministères du Travail et du Budget. Le candidat.e aura un doctorat en apprentissage machine ou en économétrie, gestion quantitative ou sociologie du travail. Le CDD sera localisé sur le campus de l'université Paris-Saclay.

Large-scale optimization: new algorithms and benchmarking

• Detailed description is available here: https://www.lri.fr/~auger/PostDocEngineerProposal.pdf
• When: from July 2015
• Contact: Anne Auger (INRIA TAO)

Bio-inspired algorithms for intrusion detection

• Official post is here
• Description of work: The postdoctoral fellow will be in charge of a survey of the recent advances in the field of intrusion detection from network streams (not limited to bio-inspired methods). She/He will select the most promising avenues, supervise their implementation by a team of engineers from the industrial partner, conduct detailed comparisons among them, and ultimately propose original approaches to address their weaknesses and tackle their limitations.
• Profile: The successful candidate should hold a recent PhD in Computer Science, with a strong expertise in bio-inspired algorithms and/or data stream mining. Knowledge of cyber-security is more than welcome, but not mandatory. On the other hand, programming skills are absolutely necessary, and software engineering experience is welcome, too.
• When: from Oct. 1. 2015
• Duration: 2 years
• Contact: Marc Schoenauer (INRIA TAO)
  Please send a complete CV, a statement of research interest, a link to the PhD dissertation and available publications, as well a list of three references.

Autonomic auto-tuning for Machine Learning

• contact Cecile Germain (Universite Paris Sud)
• more information: http://www.lri.fr/~cecile/JOBS/PostDocAutoTuning.pdf
• general goal: a fully automatic system that can adapt the configuration parameters of parallel machine learning algorithms and their hyperparameters on the fly

Machine Learning and Optimization for Long Term Investment Planning (around March 2013)

• contact Olivier Teytaud (INRIA-LRI)
• More info: http://www.lri.fr/~teytaud/metis.html
• Optimization; Energy; Machine Learning; MDP solving

Machine Intelligence for Manufacturing and Design (Sept. 2011)

• Funded by the INRIA-Saclay and Intercim SA., contact Michèle Sebag (CNRS-LRI)
• more details

Application of Machine Learning methods to Search and Optimization (Sept. 2011)

• Funded by the Microsoft-INRIA joint Lab., contact Youssef Hamadi (Microsoft Research Cambridge) and Marc Schoenauer (INRIA Saclay)
• more details

Collaborative development in planing

• Funded by the European project MASH, contact Olivier Teytaud
• more

Bandits or Monte-Carlo Tree Search

• Funded by ANR project EXPLORA, contact Olivier Teytaud
• more

Traffic modelling and inference (was Jan. 2009)

• Funded by ANR project Travesti, located in Armines (Paris)
• ANR Travesti Details

Deep Networks (was March 2010)

• Funded by ANR project ASAP, contact: Hélène Paugam-Moisy
• details

Evolutionary Planning (was Jan. 2010)

• Funded by ANR project DESCARWIN, contact: Marc Schoenauer
• details

Automatic Parameter Tuning (was Sept. 2009)

• Open position in the Adaptive Search team at Microsoft Research - INRIA joint lab.
• Contact: Marc Schoenauer

Symbolic Learning in Swarm Robotics (was April 2009)
Position at Kyushu University (Japan) funded by Japan Science Technology Agency
Post-Doc Kyushu 2009

Learning in collective robotics (was October 2008)
Details

Supervised Machine Learning for EEG (was March 2008)
Details

Automatic tuning of search and optimization algorithms: (July 2007).
See details here

Machine Learning for Evolutionary Robotics:
(May 1. 2007). Funded by INRIA - Details and application

Automatic Evolutionary Generation of Test Data:

[+]

• Date: Nov. 2006
• -Advisor Marc Schoenauer
  Funded by the European FP6 project EvoTest (Automatic Generation of Test Data for Complex Systems, http://evotest.iti.upv.es).
• The first part of the work will be to design a tool for the automatic generation of an Evolutionary Algorithm from test data structure, by extending an existing graphical prototype. The research area is that of representation-independent variation operators.
  The second part of the work deals with the automatic tuning of EA parameters. Research will concentrate on both off-line algorithms (including meta-EAs, ANOVA experiments, learning accross runs), and on-line algorithms (racing algorithms, self-adaptation, Sequential Parameter Optimizatin, entropy-based mechanisms, ...).
• Knowledge of Evolutionary Algorithms is of course highly desirable. Experience and skill in Java and C++ programming are mandatory, as the project requires imlementation of prototypes as well as interface with industrial software frameworks. Additional knowledge in Machine Learning and/or Program Testing will be appreciated.

Magneto-Encephalography Data Mining for Brain Computer Interface:

[+]

• deadline August 31. 2006. One year, starting Fall 2006. INRIA funding
  • Research context :
    • The goal of BCI is to create a translator from neural signals to intentions. Neuro-imagery inversely observes the neural signals associated to stimuli, in order to find the scenario of activation of brain areas, associated to a given experiment (e.g. looking at a moving ball). These neural signals could be used to "code", in a sense to be defined, the intentions in the BCI. For instance, the patient might decide that "thinking of a moving ball" means, moving the cursor up.
    • A major difficulty of BCI lies in the closed-loop adaptation: the machine should decipher the patient's intentions from his/her neural activations (and accordingly, e.g., move a cursor, action an artificial muscle, etc); but the patient is adapting to the machine in the meanwhile, and the (patient, machine) system does not easily stabilize, both parts of the system being a black box for one another.
    • TAO is a partner of the ACI Nouvelles Interfaces des Maths, Project Neurodyne (coordinator S. Baillet, Laboratoire de Neurosciences Cognitives et Imagerie Cérébrale, La Pitié Salpétrière, 2003-2006), concerned with the data mining of Magneto-Encephalography Data (MEG), a non-invasive technique for measuring the neuronal emissions at the level of cell assemblies, with an unprecedented temporal resolution (milli-second) ... and generates masses of temporal data.
  • Description of the activity of the postdoc : During a recent stage in the Tao Group, a new approach was proposed for mining MEG data, based on multi-objective evolutionary computation (work submitted). The goal of the present work is to improve this approach, and to create an alphabet of neural patterns, to be used for constructing a BCI protocol.
  • Required skills : The candidate should know the bases of Statistical Learning and Evolutionary Computation. Knowledge in Signal Processing is a bolus. In any case, programming skills in C++, and inclination to programming are mandatory.
  • Contact persons : Michèle Sebag, Tao team, INRIA Futurs. Michele.Sebag at lri.fr

Evolutionary generation of mesh topologies from positive examples only.

[+]

• 6 months starting asap (May 2006). INRIA funding.
  • Research context : The design of mesh topologies for complex 3D objects is a time-consuming process (typically requiring 6 man-months of highly trained engineer for a real complex 3D object). Its automation is thus a strategic objective for industries like automobile and aeronautics. The experts use a design tool to build a topology from geometrical and computational data, and the basic idea of the whole project is to optimize some script program for that generative tool using Evolutionary Computation. However, the only available data are existing topologies built by the experts, and not the sequence of instructions they used to construct them. Hence a mandatory first step is to learn to characterize the good topologies from the existing sample topologies. A second step is to use such characterization inside an automatic programming tool that will thus optimize the quality of the output of the generative program when fed by candidate script programs.
  • Description of the activity of the postdoc : A first work within TAO team has designed a representation for topologies, that make the very different sample topologies (e.g. by the number of blocks they contain) amenable to Machine Learning. The task of the post-doc will be to build on this work to actually learn to charaterize good topologies, and to validate the results on real-world mesh databases. If time permits, the second part of the work will be to use an Evolutionary Algorithm to actually build a mesh topology from geometrical informations. A population of scripting programs for the generative tool will be evolved, and its fitness will be computed using the learned quality measure after the script has actually been turned into an actual mesh topology.
  • Required skills : The candidates should have a strong background in Statistical Machine Learning and Statistics. A good knowledge of Evolutionary Computation is a plus. Programming skills are mandatory, from C++ to any language for data handling (perl. R, python, ...).
  • Contact persons : Marc Schoenauer and Michèle Sebag, Tao team, INRIA Futurs. Marc.Schoenauer at inria.fr - Michele.Sebag at lri.fr