| 10:00 |
PGMO Prize Ceremony, by François Clautiaux (ROADEF), Filippo Santambrogio (SMAI-MODE), and
|
| 10:15 |
Tristan Garrec, PGMO PhD Prize lecture |
| 10:45 |
Mathurin Massias, PGMO PhD Prize lecture |
| 11:15 |
Break |
| 11:20 |
Online interactive poster session |
| 12:50 |
End of meeting |
Analysis of a two-stage probabilistic programming model - Room Claude Berge
We present a simple two-stage probabilistic programming problem. Structural properties of the underlying probability function are briefly discused. Finally a global solution to the problem is explicitly identified.
René Henrion (Weierstrass Institute Berlin), Wim van Ackooij (EDF R&D)
: https://www.youtube.com/watch?v=310MKVqdp9Y&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=1
Presentation and other relevant documents :
http://www.wias-berlin.de/preprint/2783/wias_preprints_2783.pdf
Distributed Optimization with Flexible Communications - Room Jean-Jacques Moreau
In distributed optimization for large-scale learning, a major performance limitation comes from the communications between the different entities. When computations are performed by workers on local data while a coordinator machine coordinates their updates to minimize a global loss, we present an asynchronous optimization algorithm that efficiently reduces the communications between the coordinator and workers. This reduction comes from a random sparsification of the local updates. We show that this algorithm converges linearly in the strongly convex case and also identifies optimal strongly sparse solutions. We further exploit this identification to propose an automatic dimension reduction, aptly sparsifying all exchanges between coordinator and workers.
Franck Iutzeler (Univ. Grenoble Alpes)
: https://www.youtube.com/watch?v=NQiakepGXjo&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=3
Presentation and other relevant documents :
https://arxiv.org/abs/1812.03871
Optimal Control Techniques for Sampled-Data Control Systems with Medical Applications - Room Jean-Jacques Moreau
In biomechanics, recent mathematical models allow one to predict
the muscular force response to functional electrical stimulations. The main
concern of the present paper is to deal with the computation of optimized
electrical pulses trains (for example in view of maximizing the final force response). Using the fact that functional electrical stimulations are modeled as
Dirac pulses, our problem is rewritten as an optimal control
problem where the control is a piecewise constant function with a finite number of discontinuities. We propose direct and indirect schemes and a finer study consisting in analyzing the mathematical properties of the control problem.
Toufik Bakir (UBFC), Piernicola Bettiol (LMBA), Bernard Bonnard (UBFC & INRIA), Loic Bourdin (xLIM) and Jérémy Rouot (ISEN)
: not available
Presentation and other relevant documents :
Multivariate degradation with dynamic covariates and imperfect maintenance - Room Jean-Jacques Moreau
The project is centered on the degradation and predictive maintenance strategy of the steam generators of nuclear power plants. These heat exchangers are affected by clogging, which may induce safety and performance issues. Three different degradation indicators can be punctually measured in order to assess the clogging level, which depends on design and operation conditions that are heterogeneous over the different steam generators. The degradation phenomenon can be prevented by different types of maintenance actions (chemical cleaning processes).
The aim of the study is to be able to assess and predict the evolution of the degradation in order to develop an efficient predictive maintenance strategy. To that aim, we have developed a general framework for modelling multivariate degradation processes with dynamic covariates and imperfect maintenance.
Olivier Gaudoin (Université Grenoble Alpes)
:https://www.youtube.com/watch?v=btyuyTr8Z6k&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=2
Presentation and other relevant documents :
Interactive Methods and Preference Elicitation for Solving Hard Multi-Objective Combinatorial Optimization Problems - Room Claude Berge
In this project, we consider multi-objective problems where the decision maker's preferences are represented by a non-linear aggregation function whose parameters are initially not known. More precisely, we focus on rank-dependent aggregators such as ordered weighted averages (OWA) and Choquet integrals which are non-linear scalarizing functions that assign weights to ranks rather than to objectives in the aggregation process, so as to control the importance attached to the bottom performance or to any other order statistics; for instance, an OWA operator with decreasing weights helps promoting balanced solutions while ensuring overall efficiency. In this setting, we propose new interactive heuristic methods consisting in combining regret-based preference elicitation and heuristic search so as to quickly focus the search on the most promising solutions. For OWA operators and Choquet integrals, the proposed methods run in polynomial time and are guaranteed to generate no more than a polynomial number of queries. We perform numerical tests comparing our methods to different interactive solving methods in order to show the practical efficiency of our approach in terms of number of queries, computation time and gap to optimality.
Thibaut Lust (Sorbonne University)
:https://www.youtube.com/watch?v=31sYmY3WJC8&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=4
Presentation and other relevant documents :
Optimal Resource Allocation in Micro-organisms under Changing Environment - Room Claude Berge
Photosynthetic microorganisms are known to adjust their photosynthetic capacity according to light intensity. This so-called photoacclimation process may correspond at equilibrium to the optimal behavior in order to maximize growth. But its dynamics under varying condition remains less understood. To tackle this problem, we propose here a resource allocation model, at coarse-grained, to represent microalgae growth and photoacclimation. Using the Pontryagin maximum principle and numerical simulations, we determine the optimal strategy of resource allocation in order to optimize microalgal growth rate over a time horizon. We show that, after a transient, the optimal trajectory approaches the optimal steady state, a behavior known as the turnpike property. Then, the model is fitted with experimental data, resulting in a bi-level optimization problem which is solved numerically. The fitted trajectory represents well a Dunaliella tertiolecta culture facing a light down-shift. Finally, we compute the optimal trajectory under day/night cycle and show that the synthesis of the photosynthetic apparatus starts a few hours before dawn. This anticipatory behavior has actually been observed both in the laboratory and in the field. This shows the algal predictive capacity and the interest of our method which predicts this phenomenon.
Terence Bayen (University of Avignon)
:https://www.youtube.com/watch?v=gfvs0FFvxm0&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=5
Presentation and other relevant documents :
Planning energy investment under uncertainty - Room Marcel Boiteux
We consider stochastic programming models for energy optimization problems with investment decisions of the here-and-now type and generation variables of recourse. The resulting problem is coupled both along scenarios and along power plants. To achieve decomposition, we combine the Progressive Hedging algorithm to deal with scenario separability, and an inexact bundle method to hablde separability for different power plants in each scenario subproblem. By suitably combining these approaches, we obtain two primal-dual methods to compute an approximate solution to the original problem.
Felipe Atenas (Unicamp), Welington de Oliveira (MINES ParisTech, PSL – Research University, CMA – Centre de Mathematiques Appliquées), Claudia Sagastizábal (Unicamp) and Wim van Ackooij (EDF R&D)
:https://www.youtube.com/watch?v=OGiO643mW2c&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=6
Presentation and other relevant documents :
Hyperbolic polynomials : algorithms and implementations - Room Jean-Jacques Moreau
A real univariate polynomial is hyperbolic whenever all its roots are real or, in other words, if it equals the characteristic polynomial of a real symmetric matrix.
This property can be extended to the multivariate case via the classical algebraic tool of symmetric determinantal representations. By the way, not every multivariate hyperbolic polynomial admits such a representation/certificate. Hyperbolic Programming (HP) is the natural convex optimization problem asking to minimize a linear function over the hyperbolicity cone of a multivariate hyperbolic polynomial. HP generalizes Linear (LP) and Semidefinite Programming (SDP), central problems in mathematics and its applications. The goal of this project is to contribute to the following two research directions:
(1) the development of symbolic-numerical algorithms and implementations for the general HP problem, and
(2) efficient computation of hyperbolicity certificates such as symmetric determinantal representations.
The two questions above are highly related since when a polynomial has a symmetric determinantal representation or, more generally, when its hyperbolicity cone is a section of the cone of positive semidefinite matrices, then the associated HP problem reduces to a SDP problem.
Simone Naldi (Université de Limoges)
:https://www.youtube.com/watch?v=znSJixtBxCw&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=7
Presentation and other relevant documents :
https://msp.org/pjm/2019/303-1/p08.xhtml
Long-term Unit Commitment Problem with Optimal Zone Configuration: a Case Study in Martinique - Room Marcel Boiteux
Island regions are not interconnected with the mainland and the isolation of these regions has brought challenges to the electric power systems. For instance, the power grids in island regions are more fragile to the rapid variance of the electricity consumption or production than the continental grid. Meanwhile, more intermittent renewable energy, such as wind and solar, has been connected to the power grid. Some of the overseas territories of France are island territories in the Atlantic, Pacific and Indian Oceans.
In this work, we propose a long-term unit commitment (UC) model with simplified network (zone) constraints to apply in these areas.
One of the main purposes of this work is to examine whether a better zonal configuration could help to increase the performance of the long-term UC model. Also, by applying this long-term UC model, we can create a robust generation plan for these territories and would be able to test, for instance, whether a larger amount of renewable energy could be introduced in these areas and whether the performance of the system could be improving by constructing a new transmission line.
Hong Cai, Loïc Quéval, Martin Hennebel (CentraleSupélec, University Paris-Saclay)
:https://www.youtube.com/watch?v=vfxN3Ru2K5M&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=8
Presentation and other relevant documents :
Sequential Learning over Implicit Feedback for Robust Large-Scale Recommender Systems - Room Jean-Jacques Moreau
In this work, we propose a theoretically founded sequential strategy for training large-scale Recommender Systems (RS) over implicit feedback mainly in the form of clicks. The proposed approach consists in minimizing pairwise ranking loss over blocks of consecutive items constituted by a sequence of non-clicked items followed by a clicked one for each user. Parameter updates are discarded if for a given user the number of sequential blocks is below or above some given thresholds estimated over the distribution of the number of blocks in the training set. This is to prevent from updating the parameters for an abnormally high number of clicks over some targeted items, mainly due to bots; or very few user interactions. Both scenarios affect the decision of RS and imply a shift over the distribution of items that are shown to the users. We provide a proof of convergence of the algorithm to the minimizer of the ranking loss, in the case where the latter is convex. Furthermore, experimental results on five large-scale collections demonstrate the efficiency of the proposed algorithm concerning the state-of-the-art approaches, both regarding different ranking measures and computation time.
Aleksandra Burahnikova, Yury Maximov (Skoltech Institut) and Massih-Reza Amini (University Grenoble Alpes)
:https://www.youtube.com/watch?v=f86wvGP21IY&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=9
Presentation and other relevant documents :
Maintenance planning for hydropower plants - Room Marcel Boiteux
Presentation of a project on optimization of the maintenance planning of hydropower production systems.
Louise Fournon (ENPC)
:https://www.youtube.com/watch?v=NgW9LcGT-uI&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=10
Presentation and other relevant documents :
First and second order necessary conditions in optimal control of evolution systems - Room Jean-Jacques Moreau
We provide second order necessary optimality conditions for a class of infinite dimensional optimal control problems under functional pure state constraints together with end point constraints. Using tools of second order variational analysis, we derive necessary optimality conditions in the form of a maximum principle and a second order variational inequality. These results can be applied to optimal control models involving PDEs.
Marco Mazzola (IMJ-PRG, Sorbonne Université), Hélène Frankowska (CNRS, IMJ-PRG, Sorbonne Université) and Elsa Maria Marchini (Politecnico di Milano)
:https://www.youtube.com/watch?v=6pSdu2q0OnE&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=11
Presentation and other relevant documents :
New approaches of bidding and contract problems: use of non-self Nash games and Radner equilibrium concept - Room Marcel Boiteux
This is the presentation for our PGMO-IROE project 2019-35 entitled "New approaches of bidding and contract problems: use of non-self Nash games and Radner equilibrium concept
Didier Aussel (University of Perpignan), Wim van Ackooij (EDF R&D), Juan Pablo (Federal University of Rio de Janeiro), Claudia Sagastizabal (University of Campinas, Sao Paulo) and Olivier Beaude (EDF R&D)
:https://www.youtube.com/watch?v=5lqDO20m1fc&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=12
Presentation and other relevant documents :
Regret Bounds for Kernel-Based Reinforcement Learning - Room Jean-Jacques Moreau
We consider the exploration-exploitation dilemma in finite-horizon reinforcement learning problems whose state-action space is endowed with a metric. We introduce Kernel-UCBVI, a model-based optimistic algorithm that leverages the smoothness of the MDP and a non-parametric kernel estimator of the rewards and transitions to efficiently balance exploration and exploitation. Unlike existing approaches with regret guarantees, it does not use any kind of partitioning of the state-action space. For problems with $K$ episodes and horizon $H$, we provide a regret bound of $\widetilde{O}\left( H^3 K^{\frac{2d}{2d+1}}\right)$, where $d$ is the covering dimension of the joint state-action space. We empirically validate Kernel-UCBVI in continuous MDPs with sparse rewards.
Omar Darwiche Domingues, Pierre Menard (INRIA), Matteo Pirotta (Facebook), Emilie Kaufmann (CNRS & ULille (CRIStAL), Inria Lille) and Michal Valko (DeepMind Paris)
Presentation and other relevant documents :
Online learning for min-max discrete problems - Room Claude Berge
We study various discrete nonlinear combinatorial optimization problems in an online learning framework. In the first part, we address the question of whether there are negative results showing that getting a vanishing (or even vanishing approximate) regret is computational hard. We provide a general reduction showing that many (min-max) polynomial time solvable problems not only do not have a vanishing regret, but also no vanishing approximation α-regret, for some α (unless NP=BPP). Then, we focus on a particular min-max problem, the min-max version of the vertex cover problem which is solvable in polynomial time in the offline case. The previous reduction proves that there is no (2−ϵ)-regret online algorithm, unless Unique Game is in BPP; we prove a matching upper bound providing an online algorithm based on the online gradient descent method. Then, we turn our attention to online learning algorithms that are based on an offline optimization oracle that, given a set of instances of the problem, is able to compute the optimum static solution. We show that for different nonlinear discrete optimization problems, it is strongly NP-hard to solve the offline optimization oracle, even for problems that can be solved in polynomial time in the static case (e.g. min-max vertex cover, min-max perfect matching, etc.). On the positive side, we present an online algorithm with vanishing regret that is based on the follow the perturbed leader algorithm for a generalized knapsack problem.
Evripidis Bampis (LIP6 UPMC), Dimitris Christou (NTUA), Bruno Escoffier (Sorbonne Université) and Kim Thang Nguyen (IBISC, University Paris-Saclay)
:https://www.youtube.com/watch?v=LDfceKEc65c&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=14
Presentation and other relevant documents :
https://arxiv.org/abs/1907.05944
The polyhedral structure of multistage stochastic linear problem with general cost distribution - Room Claude Berge
By studying the intrinsic polyhedral structure of multistage stochastic linear problems (MSLP), we show that a MSLP with an arbitrary cost distribution is equivalent to a MSLP on a finite scenario tree. More precisely, we show that the expected cost-to-go function, at a given stage, is affine on each cell of a chamber complex i.e., on the common refinement of the complexes obtained by projecting the faces of a polyhedron. This chamber complex is independent of the cost distribution. Furthermore, we examine several important special cases of random cost distributions, exponential on a polyhedral cone, or uniform on a polytope, and obtain an explicit description of the supporting hyperplanes of the cost-to-go function, in terms of certain valuations attached to the cones of a normal fan. This leads to fixed- parameter tractability results, showing that MSLP can be solved in polynomial time when the number of stages together with certain characteristic dimensions are fixed.
Mael Forcier (ENPC), Stéphane Gaubert (INRIA) and Vincent Leclere (ENPC)
:https://www.youtube.com/watch?v=fSKwPITRcm8&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=15
Presentation and other relevant documents :
https://hal.archives-ouvertes.fr/hal-02929361/
Online Search with a Hint - Room Claude Berge
Outline of our project and results so far.
Spyros Angelopoulos (CNRS and Sorbonne University)
:https://www.youtube.com/watch?v=n62ngR06QJ4&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=16
Presentation and other relevant documents :
Integrated Optimal pricing and location of electric vehicle charging stations - Room Marcel Boiteux
We propose a bi-level optimization model to determine optimal pricing, sizing, and location of Electric Vehicle (EV) charging stations by taking into account the behavior of EV users. This is done by adopting a recent approach called preference-list or rank-based method which characterizes users by a set of distinct ordered sets of predefined preferences or products. In the upper-level of the bi-level model, the service provider is in charge of making decisions on the price, size, and location of the charging stations to maximize its profit. In the lower level, EV users select their first available charging stations from their preference list. We present two methods, i.e., KKT-based method and cutting plane method to solve the problem. The cutting plane method is used for solving only small problem instances in order to validate the solution obtained using the KKT-based method. The bi-level model and the optimization methods are demonstrated by performing several numerical experiments on a randomly generated set of problem instances.
Alemseged Weldeyesus (University of Edinburgh), Ikram Bouras, Luce Brotcorne (INRIA Lille) and Miguel F Anjos (University of Edinburgh)
:https://www.youtube.com/watch?v=EyLIefRiel0&feature=youtu.be
Presentation and other relevant documents :
Modelling and Solving Energy Problems with SMS++ - Room Marcel Boiteux
The optimization of energy systems is a complex task which involves the
modelling of sophisticated multi-level nonlinear stochastic problems and the
development of algorithms to solve them. On the one hand, modelling these
problems is not a simple task as they present a convoluted nested structure
that must capture phenomena at time resolutions spanning from minutes to
decades. Moreover, certain versatility is desirable to consider
characteristics of different systems. On the other hand, efficiently
implementing algorithms to solve them is particularly challenging, especially
if modelling flexibility is required and distributed computation is used to
exploit HPC systems, which is necessary to tackle real-world instances. In
order to facilitate the implementation and maintenance of such models and
general and adaptable algorithms for solving them, the innovative, open-source
Structured Modelling System++ (SMS++) is being developed. In this talk, we
will present an overview of the SMS++ system, including recent improvements
concerning the handling of uncertainty and how we deal with complex long-term
optimization problems.
Antonio Frangioni and Rafael Lobato (Università di Pisa)
:https://www.youtube.com/watch?v=Y5OdaukXgLI&list=PLukFMCj2tI3eH0s3reFnplsL-tF69CZfL&index=18
Presentation and other relevant documents :
L'argumentation autour du véhicule électrique dans les corpus des grands débats nationaux - Room Marcel Boiteux
It is today a fact that the web is a privileged medium for governments around the world to consult citizens on political and societal topics and analyze the corresponding submitted data. Using Natural Language Processing (NLP) software to create synthetic views from this data which is abundant and heterogeneous remains a non-trivial challenge. In this project, we leverage data collected by the French government and civil groups in 2019: the Great National Debate (Grand Débat National), the True Debat (Vrai Débat) and Listen the France (Entendre la France). Our aim was to detect opinions expressed in those online debates on the specific topic of the electric vehicle, and extract arguments. To this end, we proceed with manual annotation of data, from which we trained and evaluated Machine Learning tools: a classifier for Opinion Mining (Bayes and SVM) and a Named Entity Recognizer (Rasa NLU) to detect claims and arguments.
Damien Nouvel and Johanna Cordova (INaLCO)
Presentation and other relevant documents :
http://helium.lab.parisdescartes.fr:2261/debats/projet.html
Complexity Results for Common Due Date Scheduling Problems with Interval Data and Minmax Regret Criterion - Room Claude Berge
We consider the problem of scheduling independent jobs with a common due date on a single machine with the objective of maximizing the number of early jobs. The processing times are uncertain and take any value from a certain job dependent interval. For measuring the quality of an algorithm for that problem with imprecise data we use the concept of minimizing the maximum regret. We present complexity results and some dominance properties.
Imed Kacem and Hans Kellerer
: not available
Presentation and other relevant documents :
Confirmed plenary speakers
(plenary talks December 03rd, 2019)
Booklet of Abstracts
Karen Aardal (Delft Institute of Applied Mathematics, Delft University of Technology) 
Integer optimization and machine learning, some recent developments
In recent years the interface between machine learning and integer optimization has quickly been developing. We will give some examples of recent developments in the use of machine learning for learning how to branch and cut, and the use of integer optimization in adversarial machine learning.
Presentation during PGMODAYS 2019
Friedrich Eisenbrand (Swiss Federal Institute of Technology Lausanne)
Dynamic Programming for IP
In this talk, I will survey some recent results on the complexity of integer programming in the setting that lends itself to dynamic programming approaches. These include the general integer programming problem in standard form with small coefficients and integer programming problems with block structure. The field of parameterised complexity has developed tools to provide lower bounds on the complexity of IP in these cases. The goal of this talk is to give an overview on recent progress and open problems.
Rémi Munos (DeepMind Paris) 
Distributional reinforcement learning, Rémi Munos (Deepmind & Inria)
I'll talk about recent work related to the distributional reinforcement learning approach where the full return distribution is learnt instead of its expectation only. We generalize Bellman equations to this setting and describe a deep-learning approach for approximating those distributions. I will report experiments on Atari games.
Presentation during PGMODAYS 2019
Shai Shalev-Shwartz (Mobileye, an Intel company and Hebrew University of Jerusalem)
Teaching self-driving cars to be responsible: A mathematical formalism of the duty of care
An important element in making self-driving cars a real product, rather than a science project, is to provide clear safety guarantees. We argue that statistical guarantees give a very weak safety and propose instead a white-box, interpretable, mathematical model for safety assurance, which we call Responsibility-Sensitive Safety (RSS).
(plenary talks November 20, 2018)
Noureddine EL KAROUI - Criteo Al Lab & UC, Berkeley
"Auction theory from the bidder standpoint"
Most classical auction theory has been developed from the standpoint of the seller, trying to understand how to optimize auctions to maximize seller revenue for instance. Billions of auctions are now run on the internet everyday and this creates a need to better understand auctions from the bidder perspective. In this talk I will present some recent results on this question, showing for instance that auctions that are reputed to be truthful are not truthful anymore when the seller optimizes the auction format based on bidders' past bids, provide explicit (and simple to implement) shading strategies that improve bidders' utility and discuss various equilibrium questions. No prior knowledge of auction theory is assumed and the talk will be self-contained. Based on joint works with Thomas Nedelec, Marc Abeille, Clément Calauzènes, Benjamin Heymann and Vianney Perchet while doing research at Criteo.
Damien ERNST - University of Liège
"Reinforcement learning, energy systems and deep neural nets "
Reinforcement leaning is a highly-successful subfield of artificial intelligence where an agent is ought to interact with its environment to maximize a sum of rewards. In this talk, Damien Ernst will argue that this learning paradigm can be very powerful to solve many decision-making problems in the energy sector, as for example investment problems, the design of bidding strategies for playing with the intraday electricity market or problems related to the control of microgrids. He will also describe some very recent progresses in the field on deep reinforcement learning that could be used to foster the performances of reinforcement learning agents when confronted with environments that can exhibit sudden changes in their dynamics, as it is often the case with energy systems.
Brigitte JAUMARD, Concordia University Montreal
"Optimization Models and Algorithms for Network Reconfiguration"
In future cognitive networks, thanks to Software Defined Networking (SDN), network reconfiguration and traffic migration will be triggered by intelligent software tools. This requires efficient algorithms, but also clear objectives in order to know went trigger and how to conduct network reconfiguration. In this talk, we will discuss seamless network reconfiguration, as well as various minimum disruption objectives, when a seamless reconfiguration cannot be conducted, i.e., minimize \textit{(i)} the number of disruptions, \textit{(ii)} the maximum number of concurrent disruptions, \textit{(iii)} the overall duration of the disruptions, and \textit{(iv)} the maximum disruption duration. We survey these four different minimum disruption objectives for the RWA (Routing and Wavelength Assignment) and RSA (Routing and Spectrum Assignment) lightpath migration in optical networks. For each of them, the defragmentation problem can be reduced to a graph theory problem (Minimum Feedback Vertex, Vertex Separation, Graph Bandwidth, Minimum Sum Cut) or formulated as an Integer Linear Program. We investigate the most efficient available exact algorithms. For seamless reconfiguration, we will discuss a nested decomposition scheme. Extensive numerical experiments are conducted on different traffic and network instances, in order to compare the level of disruption entailed by the different objectives.
Daniel KUHN - Ecole Polytechnique Fédérale de Lausanne
"Data-driven Distributionally Robust Optimization Using the Wasserstein Metric: Performance Guarantees and Tractable Reformulations"
We consider stochastic programs where the distribution of the uncertain parameters is only observable through a finite training dataset. Using the Wasserstein metric, we construct a ball in the space of (multivariate and non-discrete) probability distributions centered at the uniform distribution on the training samples, and we seek decisions that perform best in view of the worst-case distribution within this Wasserstein ball. The state-of-the-art methods for solving the resulting distributionally robust optimization problems rely on global optimization techniques, which quickly become computationally excruciating. In this paper we demonstrate that, under mild assumptions, the distributionally robust optimization problems over Wasserstein balls can in fact be reformulated as finite convex programs - in many interesting cases even as tractable linear programs. Leveraging recent measure concentration results, we also show that their solutions enjoy powerful finite-sample performance guarantees. Our theoretical results are exemplified in mean-risk portfolio optimization, uncertainty quantification and machine learning.
Rekha THOMAS - University of Washington
"Algebraic Vision"
Algebraic Vision is an emerging viewpoint of problems in computer vision that aims to examine polynomial models in vision through the lens of algebra. A key problem in computer vision is the estimation of the three-dimensional shape of a world scene from images and the parameters (position, orientation, etc.) of the cameras that captured them. This problem, studied under the name structure-from-motion or multi-view geometry, has its origins in photogrammetry and perspective drawings. The modeling language for these problems is projective geometry, which naturally leads to polynomial models with rich and beautiful structure that beg for algebraic tools. In this talk I will show some of the surprising structure, properties, and algorithmic successes that have emerged from this algebraic viewpoint.
|
PROGRAM |
INVITED SPEAKERS | BOOK OF ABSTRACTS |
|
|
Aharon BEN-TAL, Technion, Tel-Aviv Roberto COMINETTI, UAI, Santiago Frauke LIERS, FAU Erlangen-Nürnberg Andrzej RUSZCZYŃSKI, Rutgers Henri SANSON, Orange
Invited and contributed sessions on November 14th, 2017 |
 |
Aharon BEN-TAL (Technion, Tel-Aviv)
"Regaining tractability in some large scale/uncertain engineering optimization problems"
The need to solve real-life optimization problems poses frequently a severe challenge, as the underlying mathematical programs threatens to be intractable. The intractability can be attributed to any of the following properties: large dimensionality of the design dimension; lack of convexity; parameters affected by uncertainty. In problems of designing optimal mechanical structures (truss topology design, shape design, free material optimization), the mathematical programs typically have hundreds of thousands of variables, a fact which rules out the use of advanced modern solution methods, such as Interior Point method. Some Signal Processing and Estimation problems may result in nonconvex formulations. In the wide area of optimization under uncertainty classical approaches, such as chance (probabilistic) constraints, give rise to nonconvex NP-hard problems. Nonconvexity also occurs in some Robust Control problems.
In all the above applications we explain how the difficulties were resolved. In some cases this was achieved by mathematical analysis, which converted the problems (or its dual) to a tractable convex program. In other cases novel approximation schemes, based on Robust Optimization, are used to address intractable probability inequalities. In the case of large-scale convex programs, novel first order algorithms were employed. In the Robust Control example, a parameterization scheme is developed under which the problem is converted to a tractable deterministic convex program.
Roberto COMINETTI (UAI, Santiago) 
"Optimization and Games in Congested Networks"
In this talk we will survey the use of optimization and game theory models in urban transport and telecommunications. After a brief review of equilibrium flows in congested networks, we will describe our recent work on stochastic traffic equilibrium, risk-averse route choice, dynamic equilibrium, and the limiting behavior of the Price-of-Anarchy under heavy congestion.
Frauke LIERS (FAU Erlangen-Nürnberg) 
"Robust Solution Approaches for Challenging Network Optimization and Air-Traffic Management Problems"
One way of protecting against uncertainties that occur in real-world applications is to apply and to develop methodologies from robust optimization. The latter takes these uncertainties into account already in the mathematical model. Then, the task is to determine solutions that are feasible for all considered realizations of the uncertain parameters, and among them one with best guaranteed solution value. In this talk, we give an overview over some robust approaches for real-world applications, and also talk about recent advances.
In particular, we study dynamic network flows with uncertain input data under a robust optimization perspective, where protection is sought against uncertain travel times. For operating gas networks, we present robust approaches for protecting against uncertain physical parameters. Already in the stationary case, gas network operation is complex as nonconvex quadratic optimization taks need to be solved. Robust approaches are presented for their operation that differ in the level of adjustability of the physical state variables such as pressures and flows in the network.
Finally, we will also cover an application of robust combinatorial optimization in air-traffic management. For the runway scheduling problem, we will point out modelling and solution techniques using recoverable robust solution approaches.
(Robust dynamic flows is joint with C. Gottschalk, A. Koster, B. Peis, D. Schmand, and A. Wierz. Robust gas networks is joint with D. Aßmann, M. Stingl, and J. Vera. Robust runway scheduling is joint with A. Heidt, M. Kapolke, and A. Martin.)
Andrzej RUSZCZYNSKI (Rutgers University) 
"Risk-Averse Control of Markov Systems"
We shall focus on risk-averse control of discrete-time and continuous-time Markov systems. We shall refine the concept of time consistency for such systems, introduce the class of Markovian risk measures, and derive their structure. This will allow us to derive a risk-averse couterpart of dynamic programming equations. Then we shall extend these ideas continuous-time Markov chains and derive the structure of risk measures and dynamic programming equations in these cases as well. In the last part of the talk, we shall discuss risk-averse control of diffusion processes and present a risk-averse counterpart of the Hamilton--Bellman--Jacobi equation. Finally, we provide brief information o n current research on partially-observable problems in discrete and continuous time.
Henri SANSON (ORANGE) 
"Stakes and overview of research works in Artificial Intelligence and Operations Research at Orange"
The object of the presentation is to explain how Artificial Intelligence and Operations Research techniques permit to answer operator Orange ’s big stakes in terms of infrastructure performances, operational efficiency and customer relationship. After introducing these stakes, we will review some of our major research works in Artificial Intelligence and Operations Research contributing directly to these objectives, mainly in the field of networks optimization, data mining as well as content and interaction technologies.
November 8-9, 2016 at EDF'Lab Paris Saclay
See abstracts below
Download Book of Abstracts
Download PGMO Presentation 2016
Giuseppe Calafiore (Politechnico di Turino)
“Foundations and Recent Developments in Scenario Optimization”
Scenario optimization has become a standard tool for tackling many robust design problems that are difficult to attack via standard, deterministic worst-case methods. In scenario design, a standard convex optimization problem with N randomly sampled constraints (the “scenarios”) is solved, and the scenario theory guarantees that the resulting solution is robust in a probabilistic sense against future realizations of the uncertainty. In this talk we will give an overview of the foundations of scenario optimization theory, and also discuss some recent developments and applications.
Mérouane Debbah (Huawei)
"Random Matrices for Engineering 5G"
The asymptotic behavior of the eigenvalues of large random matrices has been extensively studied since the fifties. One of the first related result was the work of Eugène Wigner in 1955 who remarked that the eigenvalue distribution of a standard Gaussian hermitian matrix converges to a deterministic probability distribution called the semi-‐circular law when the dimensions of the matrix converge to infinity. Since that time, the study of the eigenvalue distribution of random matrices has triggered numerous works, in the theoretical physics as well as probability theory communities. However, as far as communications systems are concerned, until the mid 90's, intensive simulations were thought to be the only technique to get some insight on how communications behave with many parameters. All this changed in 2000 when large system analysis based on random matrix theory was discovered as an appropriate tool to gain intuitive insight into communication systems. In particular, the self-‐averaging effect of random matrices was shown to be able to capture the parameters of interest of communication schemes. Since then, the results led to very active research in many fields such as MIMO systems or Ultra-‐Dense Networks. This talk is intended to give a comprehensive overview of random matrices and their application to the latest design of 5G Networks.
Patrick Jaillet (MIT)
"Online Optimization for Dynamic Matching Markets"
There are many situations in which present actions must be made and resources allocated with incomplete knowledge of the future. It is not clear in this setting how to measure the quality of a proposed decision strategy. Online optimization compares the performance of a strategy that operates with no knowledge of the future (on-line) with the performance of an optimal strategy that has complete knowledge of the future (off-line). In some cases some probabilistic information about the future may be available. In this talk, we provide an overview of results obtained from that perspective on problems arising from dynamic matching markets such as (i) online auctions, (ii) display advertisements, and (iii) kidney exchange programs.
Claire Mathieu (ENS)
"The average diameter of social networks "
"The friends of my friends are my friends". This well-known observation underlies a probabilistic model of social networks. With that model, one may explain the small world phenomenon: if one considers two a priori unrelated persons, in reality they can be related to each other by a short chain of friends. In a variant of the forest fire model, it is possible to prove this phenomenon rigorously.
Michel PRENAT (Thales Optronique)
"Thales Optronique : Optimization and related Data Analysis Problems"
Thales Optronique SAS, part of Thales, is a leading company in the fields of optronic systems for defense applications (land, naval, airborne) and civil applications (high power lasers dedicated to scientific research and industrial processes), and radiofrequency systems for missile electronics applications. After a review of the applications and related skills, we show some optimization stakes in the fields of complex systems design, complex automatic functions, real time heterogeneous embedded computers, interface with the human operator, maintenance policy, databases acquisition. Some of these problems are a little bit more developed, showing a strong interaction between optimization and data analysis and processing.
Welcome coffee (starting at 08:30)
09:00 Introduction by Jean-Paul CHABARD (Scientific Director EDF R&D) and Pierre Pansu (FMJH)
09:20 Scientific Conference by Mérouane Debbah (Huawei)
10:10 PGMO PhD prize ceremony part 1
10:20 PhD Prize : talk 1 by Pauline Sarrabezolles
10:50 Coffee Break
11:20 Scientific conference by Giuseppe Calafiore (Politecnico di Torino) - Slides presentation
12:10 Industrial conference by Michel Prenat - Thales Optronique
Lunch / Coffee (13:00 — 14:30)
14:30 Scientific conference by Claire Mathieu (ENS) - Slides presentation
15:20 Scientific conference by Patrick Jaillet (MIT) - Slides presentation
16:10 Coffee Break
16:40 PGMO PhD prize ceremony part 2
16:45 PGMO PhD Prize : talk 2 by Bruno Ziliotto
17:15 Review of the PGMO program, and of the IROE and IRSDI research initiatives, by Stéphane Gaubert, Gilles Stoltz, Sandrine Charousset,
Georges Hébrail and Michel Prenat.
Breaks will be the opportunity for discussion around posters describing the activities of PGMO projects.
Download PGMODAYS 2016 - Book of Abstracts
9:00 / 10:30
Amphithéâtre 1 : Mean Field Games (invited Session, organized by Daniela Tonon)
Daniela Tonon - On the variational formulation of some Mean field game systems - Slides Presentation
Yves Achdou - A Mean Field Equilibrium as a Model for Mining Industries
Luis Briceño-Arias, Dante Kalise, Francisco José Silva Alvarez - Proximal methods to solve stationary mean field game systems with local couplings - invited session on mean field games
Room A140 : Decomposition methods
Halil Sen, Boris Detienne, Ruslan Sadykov, Francois Vanderbeck - Revisiting Benders decomposition
Shahin Gelareh - BENMIP-- A Generic Benders Decomposition Solver: Challenges and Progress - Slides Presentation
Kostas Tavlaridis-Gyparakis, Antonio Frangioni - SMS++: a Structured Modeling System with Application in Energy Optimization
Amphithéâtre 2 : Optimization, Games and Renewable Energy (OGRE1) (invited Session, organized by Michel De Lara)
Michel De Lara - Charting optimization problems with multiple agents and information - Slides Presentation
El-Ghazali Talbi - A Multiobjective Evolutionary Algorithm for Household Appliances Scheduling - Slides Presentation
Yezekael Hayel - Efficient Coupled Energy-Transport Management for Smart Cities
Room A139 : OR1-Graphs and Optimization(Invited Session, organized by Marie-Christine Costa, Sourour Elloumi)
Axel Parmentier - Lower bounds in resource constrained shortest path algorithms
Viet Hung Nguyen, Paul Weng - Efficient algorithm for fair perfect matching
Youcef Magnouche, Ali Ridha Mahjoub, Sébastien Martin - The multi-terminal vertex separator problem
Room A116 : Large Scale Optimization and Statistics 1
Vianney Perchet, Francis Bach - Highly-Smooth Zero-th Order Online Optimization - Slides Presentation
Dmitry Ostrovsky, Anatoli Juditsky, Zaid Harchaoui - Adaptive filtering by convex optimization - Slides Presentation
Rémy Degenne, Vianney Perchet - Combinatorial semi-bandit with known covariance
Room 128 : Control 1
Dario Prandi - Neuro-geometry of vision and applications to image processing
Jean-Baptiste Caillau, Jean-Baptiste Pomet, Jeremy Rouot - Averaging for minimum time control problems and applications - Slides Presentation
Achille Sassi, Hasnaa Zidani, Jean-Baptiste Caillau, Emmanuel Trélat, Max Cerf - Kernel Density Estimation applied to the chance-constrained Goddard problem
11:00 / 12:30
Amphithéâtre 1 : Games 1 (Invited Session of GdR "Théorie des jeux", organized by Guillaume Vigeral)
Rida Laraki - Acyclic Gambling Games - Slides Presentation
Saeed Hadikhanloo - Learning in Anonymous Games with Actions in Banach Spaces: Application in First Order Mean Field Games
Panayotis Mertikopoulos, Yannick Viossat - Imitation dynamics with payoff shocks
Room A140 : Industrial applications (Invited Session, organized by Sandrine Charousset)
Eric Gourdin - Location and routing problems in modern telecommunication networks
Mathieu Leconte, Ioannis Steiakogioannakis, Georgios Paschos - Distributed multi-commodity routing optimization in Software-Defined Networks Slides Presentation
Bayram Kaddour, Olivier Beaude - Enedis strategic planning
Amphithéâtre 2 : Optimization, Games and Renewable Energy (OGRE2) (Invited Session, organized by Michel de Lara)
Bernardo Pagnoncelli, Tito Homem-De-Mello, Rodrigo Carrasco, Mailyng Carrasco - Microgrid energy management with renewables and storage
Didier Aussel, Anton Svensson - Multi-leader-follower-games: a state of art of applications to energy problems and of their (global vs local) reformulations
Luce Brotcorne, Léonard von Niederhäusern - Energy Pricing Problems for Demand side and Revenue Management - Slides Presentation
Room A139 : OR2-Mathematical programming in Energy (Invited Session, organized by Marie-Christine Costa, Sourour Elloumi)
Sandra Ulrich Ngueveu - Piecewise linear bounding of energy conversion functions and resulting MILP-based solution methods - Slides Presentation
Thomas Ridremont, Cédric Bentz, Marie-Christine Costa, Pierre-Louis Poirion - Mixed-integer Bilevel programs for designing robust networks. An application to wind power collection. Slides Presentation
Pierre-Louis Poirion, Cédric Bentz, Marie-Christine Costa, Thomas Ridremont, Camille Zakhour - Optimal design of a robust wind-farm under power constraints.
Room A116 : Large Scale Optimization and Statistics 2
Eugene Ndiaye , Olivier Fercoq , Alexandre Gramfort , Joseph Salmon - Gap Safe Screening Rules for Sparse-Group Lasso - Slides Presentation
Emilio Carrizosa, Amaya Nogales Gómez, Dolores Romero Morales - Clustering categories in support vector machines - Slides Presentation
Quoc Tran-Dinh, Olivier Fercoq, Volkan Cevher - A Smooth Primal-Dual Optimization Framework for Nonsmooth Composite Minimization
Slides Presentation
Room A128 : Semidefinite programming (Invited Session, organized by Xavier Allamigeon)
Didier Henrion, Simone Naldi and Mohab Safey El Din - Exact Algorithms for Linear Matrix Inequalities
Hamza Fawzi - Semidefinite programming lifts and sparse sums of squares - Slides Presentation
Xavier Allamigeon, Stephane Gaubert, Mateusz Skomra - Solving Generic Nonarchimedean Semidefinite Programs using Stochastic Game Algorithms
13:45 - 15:15
Amphithéâtre 1 : Games2 (Invited Session of GdR "Théorie des jeux", organized by Guillaume Vigeral)
Marie Laclau, Frédéric Koessler, Tristan Tomala - Competitive Information Design - Slides Presentation
Stefano Lovo - Markov Perfect Equilibria in Stochastic Revision Games - Slides Presentation
Sylvain Sorin - Limit value of dynamic zero-sum games with vanishing stage duration
Room A140 : Numerical methods for optimal transport (Invited Session, organized by Quentin Mérigot)
Jean-Marie Mirebeau - Monotone and consistent discretizations of diffusion PDEs using Obtuse Superbases
Marco Cuturi - Algorithmic Wasserstein Distances and Applications to Histogram Regression
Boris Thibert - Semi-discrete optimal transport and applications in non-imaging optics
Amphithéâtre 2 : Energy1
Claudia Sagastizábal, Wim van Ackooij, Rene Henrion, Alexander Kruger, Michel Thera,Welington de Oliveira - Bilevel Models for Unit Commitment with Demand-Side Management
Fabio D'Andreagiovanni, Giovanni Felici, Fabrizio Lacalandra - Zero-price Energy Offering by (Multiband) Robust Optimization
Mahbubeh Habibian, Golbon Zakeri, Anthony Downward, Miguel Anjos, Michael Ferris - Co-optimization of Demand Response and Reserve Offers
Room A139 : OR3 - Combinatorial Optimization (Invited Session, organized by Marie-Christine Costa and Sourour Elloumi)
Bruno Escoffier - Reoptimization algorithms for combinatorial problems - Slides Presentation
Guillaume Sagnol, Thibaut Sauvage - A Completely-positive representation of the cone of flow matrices - Slides Presentation
Spyros Angelopoulos, Christoph Dürr, Thomas Lidbetter - The expanding search ratio of a graph - Slides Presentation
Room A116 : Black-Box optimization
Benjamin Doerr, Carola Doerr, Timo Kötzing - Optimizing Multi-Valued Decision Variables with Evolutionary Algorithms - Slides Presentation
Anne Auger - Revisiting the Benchmarking of Derivative Free Optimization Algorithms
Carola Doerr, Johannes Lengler - Cost of Elitist Selection in Discrete Black-Box Optimization - Slides Presentation
Room A128 : Icode (Invited Session of the Institute for Control and Decision of the Idex Paris Saclay, organized by Yacine Chitour)
Paolo Mason - Optimal control methods for the stability of switched systems - Slides Presentation
Antoine Girard - Robust controllers for nonlinear systems with performance guarantees: an approach based on discrete abstractions.
Frédéric Jean - Inverse Optimal control: the sub-riemannian case - Slides Presentation
15:45 - 17:45
Amphithéâtre 1 : Games and Probabilities
Rene Henrion, Wim van Ackooij, Pedro Perez Aros - Aspects of Nonsmoothness for Gaussian Probability Functions - Slides Presentation
Piernicola Bettiol - Differential Games confined to path-wise constraints
Charles Bertucci - Optimal stopping in mean field games
Room A140 : OR 5
Margaux Nattaf, Christian Artigues, Pierre Lopez - An exact method for a continuous scheduling problem with concave efficiency functions
Slides Presentation
Stephane Chretien, Nathalie Herr, Jean-Marc Nicod, Christophe Varnier - Scheduling independent parallel machines with convex programming
Stephane Chretien, Paul Clarkson, Alistair Forbes - A convex relaxation of the optimal sensor placement problem in power grids
Jia Liu, Immanuel Bomze, Abdel Lisser, Jianqiang Cheng, Peter Dickinson - New formulations, relaxations and penalisations for mixed binary Qps
Amphithéâtre 2 : Energy Management and economic valuation (Invited Session, organized by Jean-Philippe Chancellier)
Teemu Pennanen - Asset valuation and optimal investment - Slides Presentation
François Pacaud - Optimization of a district microgrid
Tristan Rigaut - Energy Efficiency Investment and Management for Subway Stations
Amphithéâtre 2 : Energy 2 (17h15-17h45)
Rodolphe Griset - Nuclear Power Plant Outage Planning : an exact optimization approach to test robustness to demand variations and power plant availability
Room A139 : OR4 - Robust and bilevel optimization (Invited Session, organized by Marie-Christine Costa and Sourour Elloumi)
Pierre-Louis Poirion, Sonia Toubaline, Claudia D'Ambrosio, Leo Liberti - On the Binary Bilevel Problem
Dritan Nace, Akli Fundo, David Savourey, Fatjon Gjata - The robust flight level assignment problem - Slides Presentation
Céline Gicquel, Safia Kedad-Sidhoum, Quan Vu - A multi-stage stochastic programming approach for remanufacturing planning under uncertainty
Slides Presentation
Matteo Fischetti, Ivana Ljubic, Michele Monaci, Markus Sinnl - Interdiction Games and Monotonicity - Slides Presentation
Room A116 : Tropical methods (Invited Session, organized by Xavier Allamigeon)
Anne Bouillard - Computing optimal worst-case performance bounds with Network Calculus
Laure Daviaud - About the description of functions computed by max-plus automata
Marianne Akian, Mustapha Bouhtou, Jean Bernard Eytard, Stéphane Gaubert - Price incentives in mobile networks: a tropical approach
Room A116 : Applications (17h15-17h45)
Bayarbaatar Amgalan - An integrated method for uncovering key effects on cancer-activated multi-dimensional network - Slides Presentation
Room A128 : Control 2
Dominikus Noll - Robust control for mixed system uncertainty - Slides Presentation
Philippe Mahey, Jonas Koko, Arnaud Lenoir, Luc Marchand - Information relaxation and spatial decomposition of a stochastic optimal control problem
Francisco José Silva Alvarez, Joseph Frédéric Bonnans, Justina Gianatti - On the convergence of discrete optimal controls
Michael Orieux, Jean Baptiste Caillau, Jacques Féjoz - Problème de transfert en temps optimal - Slides Presentation
Welcome (starting at 08:30)
09:00 Introduction by Elisabeth Crepon (Ensta ParisTech) et Pierre Pansu (FMJH)
09:15 Scientific Conference by Tamás Terlaky : Mixed-Integer Second-Order Cone Optimization (MISOCO): Disjunctive Conic Cuts (DCCs) - Theory and Experiments
10:00 Results of the PGMO PhD prize by Benjamin Martin.
10:10 PHD Prize : talk 1
10:40 Coffee Break
11:10 Scientific conference by Coralia Cartis : Regularization methods - varying the power, the smoothness and the accuracy
Slides (pdf, 1.1 Mo)
11:55 Scientific conference by Alexandre Boissy (Air France) : Optimize an airline: the Air France KLM case. Why OR is a key player in concrete daily operational decisions as well as for very strategic stakes
Lunch / coffee (12:40 — 14:30)
14:30 Scientific conference by Anatoli Juditsky : Optimization problems and Variational Inequalities on domains given by Linear Minimization Oracles
Slides (pdf, 1.32 Mo)
15:15 Scientific conference by Jean-Paul Chabard (EDF) : The electrical system : an overview of some new challenges for optimization and operational research.
16:00 Coffee Break.
16:30 PHD Prize : talk 2 by Samuel Vaiter.
17:00 Review of PGMO projects by Stéphane Gaubert and Sandrine Charousset.
Breaks will be the opportunity for discussion around posters describing the activities of PGMO projects.
Parallel sessions on PGMO main topics
9h00-10h30 - Combinatorial 1
• Emiliano Traversi (LIPN Université Paris 13) :
QPLIB, a library of Quadratic Programming Instances
• Frédéric Meunier (CERMICS, Ecole des Ponts) :
New complexity results for colorful linear programming
Slides (pdf, 328 Ko)
• Rosa Figuereido (LIA, Université d'Avignon) :
Relevance of Negative Links in Graph Partitioning: A Case Study Using Votes From the European Parliament
Slides (pdf, 1.2 Mo)
11h00-12:30 - Combinatorial 2
• Julien Darlay and Frédéric Gardi (Innovation 24 & LocalSolver) :
Extension of the ADSL network of Bouygues Telecom
Slides (pdf, 1.6 Mo)
• Nicolas Beldiceanu (Mines Nantes, TASC team (CNRS/INRIA) :
Using Finite Transducers for Describing and Synthesising Structural Time-Series Constraints for Constraint Programming and Linear Programming
• Said Hanafi (Université de Valenciennes) :
Convergent Scatter Search Algorithms for 0–1 Mixed Integer Programs
14h30-16h00 - Quadratic Programming
• Denis Cornaz :
Equilibrated Vertex k-Separator Problem
Slides (pdf, 2.6 Mo)
• Sourour Elloumi (Cedric-ENSIIE) and Amélie Lambert (Cedric-Cnam) :
Exact solution of general quadratic programs through quadratic convex relaxation
• Immanuel Bomze (Department of Statistics and Operations Research (ISOR), University of Vienna) :
New formulations, relaxations and penalisations for mixed binary Qps
16h30–18h - Combinatorial – Energy problems
• Sandra Ulrich Ngueveu (INP Toulouse & LAAS-CNRS) :
Integration of reversible energy sources in combinatorial energy optimization problems
Slides (pdf, 833 ko)
• Etienne Gaillard de Saint Germain (EDF R&D) :
Application of branch and price to solve the operational scheduling problem at ERDF
• Miguel Munoz Zuniga (IFPEN) :
Optimization under probabilistic contraints of complex systems – Application to the anchoring of offshore wind-turbines
9h-10h30 - Stochastic 1
• René Henrion (Weierstrass Institut Berlin, Germany) :
(Sub-) gradient formulae for probability functions with Gaussian distribution
Slides (pdf, 976 Ko)
• Wim van Ackooij (EDF R&D) :
Probabilistic optimization via approximate P-efficient points and bundle methods
• Emilie Chouzenoux (Institut Gaspard Monge et Univ. Paris Est) :
Some recent results on proximal primal dual methods for convex optimization
11h-12h30 - Robust
• Dominikus Noll (Institut de Mathématiques de Toulouse) :
Optimization for the design of robust feedback controllers
• Michael Poss (LIRMM, CNRS) :
A Dynamic Programming Approach for a Class of Robust Optimization Problems
Slides (pdf, 458 Ko)
• Pierre-Louis Poirion ( LIX-Ecole Polytechnique):
On a robust inventory problem
14h30-16h - Stochastique 2
• Céline Gicquel (Université Paris Saclay / Laboratoire de Recherche en Informatique) :
A stochastic programming approach for planning remanufacturing activities under uncertain demand and returns forecasts
Slides (pdf, 1 Mo)
• Francisco J. SILVA (XLIM-DMI, Université de Limoges) :
The Sakawa-Shindo Algorithm in Stochastic Control
• Dang Phuong Nguyen (LIP6 UPMC) :
Stochastic graph partitioning: Quadratic versus SOCP formulations
16h30-18h – Stochastic – Energy Problems
• Pierre Carpentier (Ensta ParisTech) and Jean-Philippe Chancellier (ENPC) :
Abstract : Dual approximate dynamic programming for large scale hydro valley
Slides (pdf, 674 Ko)
• Valentin Foucher (Ecole Polytyechnique, stagiaire au CERMICS-ENPC) :
Abstract : SDDP for Energy Management in Microgrid
Slides (pdf, 850 Ko)
• Henri Gerard (Cermics ENPC) :
Abstract : Extending decomposition methods to stochastic optimization under risk
Slides (pdf, 495 Ko)
9h-10h30 - Games (Session proposed by "GdR Jeux", organised by G. Vigeral)
• Sylvain Sorin (CNRS, Université Pierre et Marie Curie) :
Convergence of gradient-like and evolutionary dynamics
• Tristan Tomala (HEC Paris) :
Stochastic Revision Games : Existence of Markov Perfect Equilibria
• Vianney Perchet (Université Paris Diderot - INRIA) :
Approachability and Stochastic Games with Incomplete Information
Slides (pdf, 398 Ko)
11h-12h30 - Markets
• Luce Brotcorne (INRIA) :
How to determine new energy service prices by integrating customers behaviour?
• Miroslav Pistek (The Institute of Information Theory and Automation of the CAS, Prague, Czech Republic ) :
Nash Equilibrium in Pay-as-bid Electricity Market
Slides (pdf, 392 Ko)
• Yezekael Hayel (Université d'Avignon) :
Strategic design optimization problem under stochastic user equilibrium constraints
14h30-16h - Games (Session proposed by "GdR Jeux", organised by G. Vigeral)
• Guillaume Vigeral (Université Paris-Dauphine, CEREMADE) :
An operator approach to zero-sum games with varying duration
• Yannick Viossat (Université Paris-Dauphine) :
Imitation dynamics and dominated strategies
Slides (pdf, 478 Ko)
• Antoine Hochart (INRIA and CMAP, Ecole Polytechnique, CNRS) :
Solvability of the ergodic equation for zero-sum games
16h30-18h - Smart-Grids
• Ana Busic (Inria Paris-Rocquencourt ) :
Ancillary Service to the Grid Using Intelligent Deferrable Loads
• Didier Aussel (Université de Perpignan) :
Water and energy integration in Eco-Industrial Parks Using a Multi-Leader-Follower Approach
• Riadh Zorgati : (EDF R&D Dept. OSIRIS) :
Two Experiments for Centralized/Decentralized Management of Energy
Slides (pdf, 1.1 Mo)
9h-10h30 - Online Optimisation
• Marc Renault (LIAFA, Université Paris Diderot-Paris 7) :
Star search problems
• Cedric Gouvernet (EDF R&D) :
Warmstarted Augmented Lagrangian for online short-term unit-commitment
Slides (pdf, 995 Ko)
• Kim Thang Nguyen (IBISC, Université d'Evry Val d'Essonne) :
Lagrangian Duality in Online Scheduling
11h-12h30 - Control
• Helene Frankowska (Institut de Mathématiques de Jussieu - Université Pierre et Marie Curie) :
Deterministic Control under State Constraints
• Hasnaa Zidani (UMA ENSTA ParisTech) :
Stochastic control problems with state constraints
• Dario Prandi :
A sub-Riemannian Santaló formula with applications to isoperimetric inequalities and Dirichlet spectral gap of hypoelliptic operators
Slides (pdf, 327 Ko)
14h30-16h - Unit-Commitment / Learning
• Vincent Leclère (CERMICS Ecole des Ponts) :
Sparse optimization methods for EDF intra-day problem
Slides (pdf, 436 Ko)
• Kostas Tavlaridis-Gyparakis (LIX, Ecole Polytechnique) :
MILP Approaches for the EdF Unit Commitment Problem
Slides (pdf, 220 Ko)
• Said Hanafi (Université de Valenciennes) :
A variable neighborhood descent for short-term hydro scheduling
16h30-18h - Optimal control and beyond
• Terence Bayen (Laboratoire Montpelliérain Alexander Grothendieck, Université Montpellier) :
About Moreau-Yosida regularization of the minimal time crisis problem
Slides (pdf, 2.2 Mo)
• Matthieu Bonnivard (Université Paris 7) :
Phase-field approximation of optimal networks
• Axel Kroener (Inria-Saclay and CMAP, Ecole Polytechnique, France) :
Optimal control of a bilinear system with application to the Schroedinger equation
9h-10h30 - Black-Box and derivative free
• Cedric Malherbe (CMLA, ENS Cachan) :
On the complexity of derivative free methods
• Carola Doerr (CNRS, LIP6 Université Pierre et Marie Curie) :
Analyzing the Drawbacks of Elitist Behavior in Black-Box Optimization
• Ider Tseveendorj (Université de Versailles Saint Quentin en Yvelines) :
Use of spherical sets for nonconvex optimization
Slides (pdf, 312 Ko)
11h-12h30 - Statistics and Optimization
• Alexandre Gramfort (LTCI, CNRS, Télécom ParisTech, Université Paris-Saclay) :
Mind the duality gap: safer rules for the Lasso
• Emile Contal (CMLA, ENS Cachan) :
Optimization with gaussian processes via chaining
Slides (pdf, 1.5 Mo)
• Michel Barlaud (Université de Nice Sofia Antipolis) :
Constrained Neyman Pearson Classification
14h30-16h - Polynomial systems and Optimization
• Jean-Bernard Lasserre (LAAS-CNRS & Institute of Mathematics, University of Toulouse) :
A bounded degree SOS hierarchy for polynomial optimization
Slides (pdf, 415 Ko)
• Alain Jacquemard (POLSYS team INRIA Paris LIP6 UPMC) :
Algebraic issues in the contrast problem in Nuclear Magnetic Resonance.
Slides (pdf, 799 Ko)
• Vianney Boeuf (INRIA & CMAP – Ecole Polytechnique) :
Performance Evaluation of an Emergency Call Center: Tropical Polynomial Systems applied to Timed Petri Nets
16h30-18h - Energy
• Roberto Wolfler (LIPN université Paris 13) :
Column and row generation for stochastic optimization
• Arnaud Lenoir :
A computational framework for decomposition-based algorithms in energy management
Executive Board :
Scientific Comitee :
Plan your trip using the information below :
- Coming to Ensta : click here
- List of hotels near Ensta : click here
- Map of RER/TGV Massy Palaiseau station : click here
- Bus schedules : click here
- schedules : click here
- schedules => Paris : click here
- schedules => St Rémy-les-chevreuses : click here
- Taxis : 01 69 20 12 01 / 01 60 14 33 33
For further information, contact : pgmodays2015@gmail.com
The Gaspard Monge Program for Optimization and operations research (PGMO), launched by EDF and the Jacques Hadamard Mathematical Foundation (FMJH), is a new type of corporate patronage whose aim is is to foster, liven up a mathematical community of researchers coming from academia and industry, working in the field of optimization and operation research from academic themes to industrial issues.
The Conference on Optimization and Practices in Industry (COPI) is devoted to applied optimization techniques used in a wide range of industrial problems (energy, banking, transportation and network management). The purpose is to bring together Academics and Industrials to share different experiences and discuss future trends related to this area.
This year, a joint conference PGMO-COPI’14 will be held on October 28th to 31st 2014 in Ecole Polytechnique, Paris-Saclay. Both theoretical and practical issues will be considered.
See the participants list : click here
Download the detailed program with parallel sessions : click here
Download the proceedings : click here
Agenda
Monday 27th : PGMO-COST Workshop on Validation and Software
The detailed program of the workshop.
Tuesday 28th
The detailed program of PGMO-COPI'14 (pdf, 1.13 Mo)
The proceedings of PGMO-COPI'14 (pdf, 2.65 Mo)
Welcome (starting at 08:30)Welcome (starting at 08:30)
09:00 Introduction (Patrick Le Quéré, Ecole Polytechnique, Marc Ringeisen, EDF R&D) - Amphithéatre Faurre
09:15 Scientific Conference by Viorel Barbu, Al. I. Cuza University, Iasi, Romania : "Optimal control approach to the optical flow problem" (Amphithéatre Faurre)
10:00 Results of the PGMO PhD prize - Amphithéatre Faurre
10:10 PGMO PHD prize : First Conference : Daniel HOEHENER - Amphithéatre Faurre
Daniel HOEHENER prepared his PhD thesis at Université Pierre et Marie Curie (Paris 6) on "Conditions d’optimalité pour des problèmes de contrôle optimal avec contraintes d’états'' (Optimality conditions for some optimal control problems with state constaints), supervised by Hélène Frankowska. His thesis includes several original results about second order conditions, proven under more general assumptions than the results already existing in literature, expressed in primal form, and involving both state and control constraints. The jury committee was impressed by the novelty of the results in a very technical domain, which also gave rise to excellent publications.
10:40 Coffee break.
11:10 Scientific Conference by Daniel Bienstock, Columbia University : "Progress on solution of OPF problems" (pdf, 317 Ko) (Amphithéatre Faurre)
Abstract : OPF, or Optimal Power Flow, is a procedure of paramount importance in the operation of electrical power transmission systems. Introduced by Carpentier more than fifty years ago, OPF problems are run with some frequency (as often as every five minutes) to set generator output levels to be used during the next time window. The goal is to satisfy forecast levels of loads (demands), with some flexibility, and at minimum cost. In an accurate rendition, these problems are QCQPs -- quadratically constrained quadratic programs, and therefore, potentially extremely difficult computational problems. In routine operation of a grid, however, near-optimal solution of these problems is in fact routine, relying on linearizations, logarithmic barrier methods, and Newton's method. However, the analysis of 'what-if' scenarios, and the real-time computation of recourse actions in the event of an emergency, expose the weakness of the routine methods. Often, they fail to converge, or they converge to a solution that is later revealed to be quite suboptimal. This shortcoming makes precise risk analysis problematic. New research by Lavaei and Low has offered some hope that semidefinite relaxation-based methods may prove successful. Although subsequent research is challenging this view, a new wave of research on modern techniques for solving OPF problems has emerged. In this talk we will outline various ongoing research thrusts as well as connections to standard problems of interest to the optimization community.
11:55 Scientific Conference by Francis Bach, INRIA, Paris, France : "Beyond stochastic gradient descent for large-scale machine learning" (Amphithéatre Faurre)
Abstract : Many machine learning and signal processing problems are traditionally cast as convex optimization problems. A common difficulty in solving these problems is the size of the data, where there are many observations ("large n") and each of these is large ("large p"). In this setting, online algorithms such as stochastic gradientdescent which pass over the data only once, are usually preferred over batch algorithms, which require multiple passes over the data. In this talk, I will show how the smoothness of loss functions may be used to design novel algorithms with improved behavior, both in theory and practice: in the ideal infinite-data setting, an efficient novel Newton-based stochastic approximation algorithm leads to a convergence rate of O(1/n) without strong convexity assumptions, while in the practical finite-data setting, an appropriate combination of batch and online algorithms leads to unexpected behaviors, such as a linear convergence rate for strongly convex problems, with an iteration cost similar to stochastic gradientdescent. (joint work with Nicolas Le Roux, Eric Moulines and Mark Schmidt)
12:40 Lunch
14:00 Coffee
14:30 Review of PGMO projects 2013-2014 - Sandrine Charousset and Stéphane Gaubert - Amphithéatre Faurre
15:00 Conference by Marc Ringeisen, EDF R&D : "The stakes for edf in the energy transition." - Amphithéatre Faurre
15:45 Scientific Conference by Jean-Bernard Lasserre, LAAS-CNRS, France : "The moment-LP and moment-SOS approaches in optimization." (Amphithéatre Faurre)
We review and compare basic properties of the moment-LP and moment-SOS hierarchies for polynomial optimization problems P: min {f(x): x in K} where "f" is a polynomial and "K" is a compact basic semi-algebraic set.
Thanks to powerful positivity certificates from real algebraic geometry, both approaches provide a sequence of lower bounds which converge to the global minimum of P: We also describe an alternative hierarchy based on a different characterization of nonnegativity on K which now yields a sequence of upper bounds that converges to the global minimum of P (provided that the (non necessarily compact) set K is such that one may compute moments of a measure whose support is exactly K).
16:30 Coffee break.
17:00 PGMO PHD prize : Second Conference : Miquel OLIU BARTON - Amphithéatre Faurre
Miquel OLIU BARTON also prepared his PhD thesis in Paris 6, under the supervision of Sylvain Sorin. The dissertation, on "Jeux dynamiques à information incomplète en temps discret et continu" (Dynamical games with incomplete information in dicrete and continuous time) is mainly devoted to the long-time behavior in differential and repeated games and control problems, and to games with lack of information. The jury committee appreciated the impressive amount of works and the novelty of the results, and in particular the new proof that M. Oliu Barton gave of the existence of a limit for the value in discounted games, with new and self contained techniques.
17:30 Presentation of PGMO projects 2014-2015 - Sandrine Charousset and Stephane Gaubert - Amphithéatre Faurre
18:00 End of the day.
Coffee breaks and Lunches will be the opportunity for discussion around posters describing the activities of PGMO projects.
Wednesday 29th
Welcome (starting at 08:30)
09:00 Scientific Conference by Claudia Sagastizábal, IMPA, Rio de Janeiro, Brasil : "On Modelling and Solving Risk-Averse Stochastic Equilibrium Problems" (Amphithéatre Faurre)
Abstract : Over the last decades and in a worldwide trend,many industries that were considered as "natural monopolies'' in the 1970s (electricity, telecommunications, natural gas, water supply) were restructured to introduce competition into one or more of their horizontal segments. Nevertheless, even in the most liberalizing countries, some strategical sectors continue to be subject to regulations in quality, price and entry. Such is the case for networks transmitting and distributing electricity or transporting natural gas.
For these oligopolistic industries, the regulatory mechanisms have important implications not only for supporting wholesale and retail competition but also to maintain the network reliability. To have a full understanding of the market behaviour, it is necessary to understand the impact of distortions introduced by the
regulator when capping prices, or when applying rewards and incentives for efficient production, like emission allowances.
In this talk we consider equilibrium problems for a market with competing risk-averse agents that try to maximize profit subject to coupling constraints resulting from regulatory interventions or market clearing conditions.
Each agent in the market solves a one-stage risk-averse optimization problem with both here-and-now (investment) variables and (production) wait-and-see variables. A shared constraint couples almost surely the wait-and-see decisions of all the agents.
An important characteristic of our approach is that the agents hedge risk in the objective functions (on costs or profits) of their optimization problems, which has a clear economic interpretation. This feature is obviously desirable, but in the risk-averse case it leads to variational inequalities with set-valued operators -- a class of problems for which no established software is currently available. We discuss how to overcome this difficulty without losing decomposable structures in the problem, and present numerical results, including on part of the real-life European gas network.
Joint work with J.P. Luna (UFRJ, Rio de Janeiro) and M. Solodov (IMPA, Rio de Janeiro)
09:45 Scientific Conference by René Henrion, Weierstrass Institute, Germany : "Gradient formulae in probabilistic programming" (Amphithéatre Faurre)
Abstract : Probabilistic Programming is a discipline dealing with optimization problems under probabilistic constraints. The latter have proven to be useful in many engineering problems, in particular in the management of hydro reservoirs. A major challenge in the solution of such problems is to calculate values and gradients of parameter-dependent probability functions, where the parameter is give by the decision te be optimized. The talk discusses this issue for the important class of linear probabilistic constraints under Gaussian distribution with a perspective to nonlinear constraints and Gaussian-like distributions.
10:30 Coffee break.
11:00 Parallel Sessions : (detailed program with parallel sessions : click here)
Contributed Session - Markets (Amphi. Sauvy)
Contributed Session - Hydropower (1/2) (Amphi. Faurre)
Contributed Session - Unit Commitment (1/2) (Amphi. Poisson)
Invited Session - Game Theory (1/2) (Amphi. Painlevé)
Contributed Session - Global optimisation (Amphi. Curie)
12:40 Lunch
14:00 Coffee
14: 30 Parallel Sessions : (detailed program with parallel sessions : click here)
Invited Session - Stochastic Optimisation (1/2) (Amphi. Faurre)
Invited Session - Game Theory (2/2) (Amphi. Painlevé)
Contributed Session - Nuclear (Amphi. Sauvy)
Contributed Session - Optimal Control (Amphi. Curie)
Invited Session - Recent Advances on Continuous Randomized black-box optimization (1/2) (Amphi. Poisson)
16:00 Coffee break.
16:30 Scientific Conference by Laurent El Ghaoui, University of Berkeley, USA :"Robust Sketching for Large-Scale Multi-Instance Optimization" (Amphithéatre Faurre)
Abstract :In the recent years there has been a lot of interest in approximating a data matrix by a "sketch", that is, a simpler matrix that preserves some property of interest, and on which computations can be performed faster than the original. We consider the idea in the context of solving a linear or convex quadratic program, and develop the technique of "robust sketching", which entails replacing the coefficient matrix with a sketch, but keeping track of the error thus made, via robust optimization. We examine applications of the concept in the areas of machine learning and in the context of very large LPs arising in resource management.
17:20 Cocktail, introduced by Hans Rugh, FMJH Director
Thursday 30th
Welcome (starting at 08:30)
09:00 Scientific Conference by Ivar Ekeland, Ceremade, Université Paris Dauphine, Paris, France : "Moral hazard: how to deal with people who lie and cheat" (Amphithéatre Faurre)
Abstract : In many situations, one cannot monitor an employee's effort, but one can observe his performance, which depends not only on his effort, but on luck, good or bad. The moral hazard problem is to devise a compensation scheme, relying on past performance, which will incite him to exert maximum effort. Some years ago, Sannikov found a way to reduce it to a stochastic optimization problem, and I will show his method on a simple example.
09:45 Scientific Conference by Nicole El Karoui, Université Pierre et Marie Curie and Ecole Polytechnique, Paris, France : "Dynamic utilities and Discount Curves In long terme decision making." (Amphithéatre Faurre)
Abstract : A large debate is open for several years on the standard cost-bene
t analysis to evaluate long-term policies associated with sustainable energy, resource conservation, climat change.... (Stern, Weitzman, Gollier, Ekeland, Chiappori....,). In particular, the discount rate has been at the center of discussions, given its impact on the "price of future". On the other hand, financial markets provide daily information on
financial interest rates, discount factors, yield curves, all assumed to be random. How to reconcile these different "discounting" problems? Following, Lecocq and Fourcade, discounting is a important tool to clarify the issues for the decision; but the decision has to be sequential, based on learning from the past, and driven by an adaptative criterion. In finance, Musiela and Zariphopoulou proposed to introduce dynamic utilities with no-prespecified trading horizon. With an additional consistency property with a given incomplete financial market, these dynamic utilities and its conjugate may be characterized as value function of control problem, and so as solution of Hamilton Jacobi Bellman SPEs. More interesting is the splitting property of the marginal utility in terms of monotonic optimal processes. This point of view applied to consumption problem is of large importance in long term decision making as showed by Ramsey in 1932. The notion of discount factors has to be revisited. That is a crucial point. As a first step, discount factors are evaluated as marginal indifference prices (in incomplete market) of units of cash in the future. Within this framework, the yield curve is depending of the coefficient of risk aversion and a prori of the wealth in the economy, but not a priori on the rate of preference for the present. An example based of affine yields curves well-known in finance, and classical power utilities is discussed, and simulation are provided. By mixing family of such utilities, agents heterogeneity may be easily introduced. The "mixed" discount factor may be interpreted as a consensual discount.
In conclusion, in the large and complex debate on the discount rates in economy, we have tried to conciliate economic and financial point of view, in a more stochastic way where preferences for the present are now depending on the wealth of the economy, on the diversity of the population and other economic factors.
10:30 Coffee break.
11:00 Parallel Sessions : (detailed program with parallel sessions : click here)
Invited Session - Complexity of linear programming and games (1/3) (Amphi. Sauvy)
Invited Session - Stochastic Optimisation (2/2 : applications to energy management) (Amphi. Faurre)
Invited Session - Energy Markets (Amphi. Painlevé)
Invited Session - Recent Advances on Continuous Randomized black-box optimization (2/2) (Amphi. Poisson)
12:40 Lunch
14:00 Coffee
14:30 Scientific Conference by Frédéric Roupin, Université Paris 13, France : "Solving binary quadratic problems with BiqCrunch" (Amphithéatre Faurre)
15:15 Scientific Conference by Sourour Elloumi, ENSIIE, Evry, France : "Quadratic Convex Reformulation for discrete quadratic optimization : Basic results and recent extensions" (Amphithéatre Faurre)
Abstract : We consider problem (QP) of minimizing a quadratic function subject to linear or quadratic constraints. Variables are integer and bounded. This very general problem can model many classical problems in Combinatorial Optimization.
A major difference between (QP) and integer linear programs lies in the fact that, in general, its continuous relaxation is an NP-hard optimization problem. To overcome this difficulty, the Convex Quadratic Reformulation approach transforms (QP) into a problem (QP'), equivalent to (QP), but whose continuous relaxation is a convex problem. To compute a global optimal solution to (QP) it becomes possible to solve (QP') by an implicit enumeration algorithm based on continuous convex optimization.
We make an overview of recent developments in this approach. We show in particular how positive semidefinite relaxations can be used to build the most interesting equivalent problems (QP'). In the case of binary variables, we give a vision of classical linearization as a special case of Convex Quadratic Reformulation.
Finally, we illustrate the generality and the computational efficiency of the Quadratic Convex Reformulation approach for several combinatorial optimization problems.
16:00 Coffee break.
16:30 Parallel Sessions : (detailed program with parallel sessions : click here)
Contributed Session - Game theory (Amphi. Painlevé)
Contributed Session - Scheduling (Amphi. Sauvy)
Contributed Session - Unit Commitment (2/2) (Amphi. Poisson)
Contributed Session - Industry (Amphi. Faurre)
18:00 Break
20:30 - 23:45 Social Event : Tour Montparnasse, Paris
Friday 31st
Welcome (starting at 08:30)
09:00 Scientific Conference by Jean Andre, Air liquide, Paris, France : "Forecasting & Optimization challenges for a large industrial electric energy consumer." (Amphithéatre Faurre)
09:45 Scientific Conference by Golbon Zakeri, University of Auckland, New Zealand : "Models for effectively integrating volatile renewable generation into electricity systems." (Amphithéatre Faurre)
Abstract : Integrating renewable sources of generation such as wind and solar into electricity generation is a big challenge as these sources result in intermittent and volatile supply. In this presentation we will discuss models that are designed to make the supply more robust and cost effective. We will discuss the implementation of these models in perfect and imperfect market settings.
10:30 Coffee break.
11:00 Parallel Sessions :
(Amphi Poisson) Contributed Session - Chance Constraint Programming and Quadratic Programming
(Amphi Sauvy) Invited Session - Complexity of linear programming and games (2/3)
(Amphi Faurre) Invited Session - Optimization tools for large scale industrial systems
(Amphi Painlevé) Contributed Session - Smart-Grid and energy
12:40 Lunch.
14:00 Coffee.
14:30 Conference by Patrice Gérardin, Director - Commerce Optimisation & Trading, EDF, France : "EDF scenario approach for investment choices." (Amphithéatre Faurre)
15:15 Parallel Sessions :
(Amphi Painlevé) Contributed Session - Learning and Metaheuristics
(Amphi Faurre) Contributed Session - Hydropower (2/2)
(Amphi Poisson) Contributed Session - Operational research
(Amphi Sauvy) Invited Session - Complexity of linear programming and games (3/3)
16:45 Closure of the Conference by Marc Ringeisen, EDF R&D. (Amphithéatre Faurre)
17:00 End of the day.
PGMO together with the COST Action "Mathematical Optimization in the Decision Support Systems for Efficient and Robust Energy Networks"
(see website : http://cost-td1207.zib.de/) organise prior to the COPI-PGMO'14 conference, on Monday October 27th in Ecole Polytechnique (Paris-Saclay, France) a workshop related to Validation and Software. More informations and registration : click here
2014 PGMO PhD Award : click here
Time line
Call for abstracts 2014 (pdf, 294 ko)
PGMO'days talks 2012 : click here
PGMO'days talks 2013 : click here
Copi'11 : click here
|
|
Day 1 (Tuesday, 28) |
Day 2, 3, 4 |
Social Event |
Registration |
|
Students |
Free |
100€ |
50€ |
close |
|
Academic |
Free |
300€ |
50€ |
close |
|
Industrial |
Free |
500€ |
50€ |
close |
A cocktail will be organized on Oct. 29th, 2014 after the end of the sessions of Day 2.
The social event dinner will be held in the top of Montparnasse Tower on Oct. 30th, 2014 with a panoramic view on Paris monuments:
* 20:30-21:45 : welcoming drink
- 21:45-00:00: dinner
The number of places is limited.
Registrations were closed on Oct 19th. However, we will do our best to include late registrations. If you would like to register, please send us an email with the details below. If any payment required, you will be invited to do it in the dedicated desk by check or credit card during the conference:
- First name, Famiily name and university/company
- Days of participation
- Would like to attend the cocktail of on Oct 29th?
- Would like to attend the social event on Oct 30th?
Plan your trip using the information below :
- Coming to Ecole Polytechnique, Paris-Saclay : click here
- List of hotels near Ecole Polytechnique, Paris-Saclay : click here
- Map of RER/TGV Massy Palaiseau station : click here
- Bus schedules : click here
- schedules : click here
- schedules => Paris : click here
- schedules => St Rémy-les-chevreuses : click here
- Taxis : 01 69 20 12 01 / 01 60 14 33 33
For further information, contact : pgmo.copi.14@fondation-hadamard.fr
Call for papers :
On the occasion of PGMO-COPI14, the journal Energy Systems (http://www.springer.com/engineering/energy+technology/journal/12667) will publish a special issue edited by Wim van Ackooij, Andy Philpott, Claudia Sagastizabal, and Riadh Zorgati.
GUIDELINES
- The submission deadline is December 15th, 2014
- All papers should be submitted through the journal online system (https://www.editorialmanager.com/ensy/) choosing the article type "SI - Optimization in Energy"
- When preparing the manuscript, please follow Springer's instructions for authors (http://www.springer.com/engineering/energy+technology/journal/12667?detailsPage=pltci_1060786) and use Springer’s LaTeX macro package (http://static.springer.com/sgw/documents/468198/application/zip/LaTeX.zip)
pgmo_conf2013_slides_van_ackooij.pdf
pgmo_conf2013_slides_cheng.pdf
pgmo_conf2013_slides_pfeiffer.pdf
pgmo_conf2013_slides_alais_1.pdf
pgmo_conf2013_slides_houda.pdf
pgmo_conf2013_slides_lisser.pdf
pgmo_conf2013_slides_gleixner.pdf
pgmo_conf2013_slides_traversi.pdf
pgmo_conf2013_slides_nguyen.pdf
pgmo_conf2013_slides_sarrabezolles.pdf
pgmo_conf2013_slides_bellalij.pdf
pgmo_conf2013_slides_grappe.pdf
pgmo_conf2013_slides_cassioli.pdf
pgmo_conf2013_slides_mahey.pdf
pgmo_conf2013_slides_leclere.pdf
pgmo_conf2013_slides_lenoir.pdf
pgmo_conf2013_slides_alais_2.pdf
pgmo_conf2013_slides_delara1.pdf
pgmo_conf2013_slides_delara2.pdf
pgmo_conf2013_slides_delara3.pdf
pgmo_conf2013_slides_bendotti.pdf
pgmo_conf2013_slides_dupin.pdf
pgmo_conf2013_slides_gorge.pdf
pgmo_conf2013_slides_grebille.pdf
pgmo_conf2013_slides_pistek.pdf
pgmo_conf2013_slides_dallagi.pdf
pgmo_conf2013_slides_beldiceanu.pdf
pgmo_conf2013_slides_oudjane.pdf
pgmo_conf2013_slides_figueiredo.pdf
pgmo_conf2013_slides_zaourar.pdf
pgmo_conf2013_slides_frangioni.pdf
pgmo_conf2013_slides_nikolova.pdf
pgmo_conf2013_slides2013_condat_0_partie1.pdf
pgmo_conf2013_slides2013_condat_0_partie2.pdf
pgmo_conf2013_slides2013_condat_0_partie3.pdf
pgmo_conf2013_slides_boyer.pdf
|
Room |
2.1.51 |
2.1.52 |
2.3.30 |
2.3.31 |
|
09h00 – 10h3 |
Combinatorial optimization |
Scheduling outages for nuclear plants |
Probabilistic constraints and robust optimization |
Short term optimization of electricity generation schedules |
|
11h00 – 12h3 |
Combinatorial optimization |
Scheduling outages for nuclear plants |
Probabilistic constraints and robust optimization |
Short term optimization of electricity generation schedules |
|
14h30 – 16h00 |
Combinatorial optimization |
Market Equilibrium
Optimization of hydroelectric valleys |
Large scale systems and dynamic programming |
Optimization for Imaging |
|
16h30 – 18h00 |
Combinatorial optimization |
|
Large scale systems and dynamic programming |
|
Organisation Comitee :
Scientific Comitee, advised by the PGMO scientific Council :
|
Opening Conference
18-19 September 2012 Ecole Polytechnique, and ENSTA ParisTech, Palaiseau, France
|
|
Here you can see some pictures
A movie made at the conference, on innovation.edf.com : see here
Papers published after the 18-19 september conference
A paper published in " Le Parisien"
A paper published in " Alternatives Paloises "
Posters
See here the posters presented by PGMO teams
Talks :
September 18th 2012, at Ecole Polytechnique
|
Introduction by F. Pacard, Ecole Polytechnique Deputy Director |
|
|
“Variational aspects of rank minimization problems” University Paul Sabatier, Toulouse, France |
|
|
“Evaluation complexity in smooth constrained and unconstrained optimization” University of Namur, Belgium |
|
|
“Subgradient methods for huge-scale optimization problems” University of Louvain, Belgium |
|
|
“The modeling of risk in optimization involving uncertainty” University of Washington, Seattle, USA |
|
|
Presentation of FMJH and PGMO, 2012 "call for projects" : elected projects Grégoire Allaire, Ecole Polytechnique, France Sandrine Charousset, EDF R&D, France |
|
|
“Evolutions et nouveaux besoins du système électrique” EDF R&D scientific director |
|
|
Conclusion Yves Laszlo, FMJH Director and Bernard Delpech (EDF R&D Deputy Director) |
|
September 19th 2012, at ENSTA Paristech
|
Introduction A. Reichart, ENSTA deputy director |
|
|
"On mean field games" Collège de france |
|
|
"Phase Recovery, MaxCut and Complex Semidefinite Programming" CNRS and Ecole Polytechnique, FRANCE |
|
|
"Optimisating nuclear outages : recent advances and perspectives" EDF R&D, France |
|
|
"Robust mixed integer programming with continuous recourse variables" ENSTA Paristech and CEDRIC, France |
|
|
"A tropical approach of zero-sum games" INRIA and Ecole Polytechnique, France |
|
|
"Distance constraints in Euclidean geometry" Ecole Polytechnique, France |
|
|
"Optimizing with probabilistic constraints" University Paris-Sud, France |
|
|
"Unit-Commitment in Energy Management : recent advances and perspectives" EDF R&D, France |
|
|
"Asymptotical problems in optimal location” University Paris-Sud, France |
|
|
"Second-order optimality conditions and small-time geodesics in sub-Finsler geometry" INRIA and Ecole Polytechnique, France |
|
|
"Optimal transportation under controlled stochastic dynamics" Ecole Polytechnique, France |
|
