Selected publications and realizations
Olivier Teytaud --- TA
(General audience: in french: Videos or audio Europe1 or audio on RSR or popularization on "Le Monde";
in Chinese (Taiwan): Nov, 12th 2010 and Nov. 15th 2010 )
(personal blog; nobody else than me is responsible for it)
Energy management forum in Tainan (Taiwan)

Most importantly, working only on really important problems around planning, machine learning (ML), optimization, operational research (OR). I like both mathematical analysis and experimental research.
Application side: I've spent a lot of energy in the past around environment. This remains a main motivation to me. My main topics have been:
- Energy planning;
- Water resources;
- Meteorological previsions for clean water;
- Drainage basins;
- Sanitary sewers;
- Energy consumption estimation.
- Games
For each of these topics I have mainly a point of view of computer scientist - I know nothing about a nuclear power plant or a hydroelectric plant, and I've never met a drainage basin, but I have experience around their modelization and use in a planning problem. Also, I've worked only of aspects of meteorological previsions that can be of some use for preserving or improving environment.
Parallel algorithms (based on the feeling that the number of people at the intersection of OR/ML and parallelization is too small).
High scale problems (high dimension, or highly ill-posed optimization problems), and for this contacts with industry.

Some publications (complete list since 2005)

Around Planning in high-dimensional contexts: (description of our main technique here in french , seminaire Le Modele et L'algorithme, Inria or here in english (Birs2011) ).
- The computational intelligence of MoGo (IEEE Transactions on CIAIG).
- Consistency modifications for automatically tuned Monte-Carlo Tree Search (LION4).
- At last some wins as black in 9x9 Go komi 7.5 against top players (ICGA journal, december 2010).
Around noisy optimization
- Bernstein races are great for Adaptive noisy optimization (EvoStar 2010).
- Bandit-based genetic Programming (EuroGP 2010)
Around parallel optimization and complexity bounds in optimization
- Lower bounds for Comparison-Based Evolution Strategies using VC-dimension and Sign Patterns . (tightness and explicit algorithms in the paper below), Algorithmica, Accepted.
- Theory of Parallel Evolutionary Algorithms: Tightness and Automatic Parallelization (submitted).
- Continuous Lunches are free plus the design of optimal optimization algorithms (Algorithmica 2010)
Others: quasi-random numbers are very efficient in evolutionary optimization, both for EMNA, CMA , restarts, ...

Some realizations and awards

The nash equilibrium of 2-player game with partial information is undecidable, even with finite state space.

We got the gold medal in TAAI 2010 in three categories: 9x9 Go, 13x13 Go, 19x19 Go!
Our AI is ranked in the top 1\% of humans at Urban Rivals, a free internet game.
2010: we got the 2009 chess base award for the biggest contribution to games.
We won the first ever (and only) game against a top professional player in 9x9 Go with komi 7.5 as black, more here.
We are the only team which realized in the game of Go
- a win with H7 against a top pro in 19x19 Go;
- a win with H6 against a pro player in 19x19 Go.
The algorithm presented in C. Hartland, S. Gelly, N. Baskiotis, M. Sebag, O. Teytaud (2006). Multi-Armed Bandit, Dynamic Environments and Meta-Bandits. Workshop "On-Line Trading of Exploration and Exploitation", Nips-Workshop 2006, paper has won the Exploration Versus Exploitation challenge here. Accepted in CAP'07 (acceptance rate 40 \%, with title "Change Point Detection and Meta-Bandits for Online Learning in Dynamic Environments", 12 pages.

Some contracts

High-dimensional planning with applications to energy: (coordinateur)IOMCA (white ANR project with Taiwan and Artelys) around high-dimensional planning with applications in energy (introductory talk)
MASH : robotics, vision and collaborative development for planning (responsible WP6 - planning)
OMD and its follow-up OMD2 : multi-disciplinary optimization. Includes academics + Industrial partners around expensive optimization (in particular surrogates and parallelization)
EXPLO-RA (exploration / exploitation dilemna).

Brief info

1995: Normale Sup Lyon (what was termed at that time "concours M' (math)" )
2001: ph.D.
2002-2004: consultant
- Planification de production electrique
- Planification/simulation dans le domaine de la sante
- Planification pour les ecoulements d'eau
- Prevision meteorologiques locales
- Statistiques pour prevision d'emplois, de transports, de logements
2004-now: research position in Inria Saclay (CR1)
- Optimization (theory and practice; parallel, noisy cases included) ...... (33 %)
- Planning (theory and practice; including games and parallelism) ....... (33 %)
- Machine learning (theory and practice; include parallelism) ............... (33 %)

Students

I'm very happy that all my students have done great publications and have good positions; for any of them, the HAL server should provide a list of publications.

ph.D. students
- S. Gelly (1st Prix de la Chancellerie (important ph.D. award in France); 2nd at the "Specif Gilles Kahn" award), now at Google.
- J. Mary, currently assistant professor in Univ. Lille
- F. Teytaud, post-doc Paris Dauphine
- A. Rimmel (post. doc univ. Alberta, post-doc Paris Dauphine, industry)
- N. Omont (now consultant in www.artelys.com)
- J.B. Hoock (ph.D. in progress)
- P. Rolet (now "chef de projet" in industry)
- A. Couetoux (ph.D. in progress)
- J. Decock, Ilab with Artelys

2 post-doc:
- Nataliya Sokolovska, since January 2010,
- David Auger, since October 2010).
2 engineers (J.B. Hoock, 2 years, H. Doghmen, 2 years)
8 internships (all have publications)
- 100% Atsushi Takahashi (engineering school)
- 100% Remy Bergasse (licence ENS)
- 80% Merve Amil (master 2)
- 50% Yann Bonnemay (engineer
- 33% Yizao Wang (X internship)
- 100% Christophe Fiter (Ecole Centrale)
- 100% Louis Chatriot (X internship)
- 100% Vincent Berthier (engineer)
4 engineers' projects (Epita).