UCL

PhD Students

BORCKMANS Pierre (Starting date : October 2008)
Derivative-Free Optimization : Swarm Intelligence and multimodal problems
Thesis advisors : P. Van Dooren, P.-A. Absil
Classical optimization methods often rely on the use of the derivatives of the objective function. As this is not always suitable in practice, some methods have been developed that do not use these derivatives. Among these derivative-free optimization (DFO) methods, I will study Particle Swarm Optimization (PSO). This stochastic population-based algorithm was first proposed in 1995 and some variations have now been introduced. I will focus on two aspects of this method that need to be investigated further : multi-modal problems and non-euclidian search spaces (e.g. manifolds, ...). New formulations of the PSO algorithm dealing with these problems will be proposed. In this process, various practical applications will be studied (e.g. model reduction, vision, genetic analysis,...).

BOUMAL Nicolas (Starting date : September 2010)
Design and analysis of algorithms for smooth curve fitting on Riemannian manifolds
Thesis advisors : V. Blondel, P.-A. Absil
We focus on designing and analyzing practical algorithms to fit smooth curves to data on manifolds. The regression tools we intend to develop will be useful in a number of applications, providing the means for data resampling and denoising. We specifically aim to provide efficient software to achieve these tasks on tractable and practically important Riemannian manifolds.

BROWET Arnaud (Starting date : 15 September 2008)
Image segmentation and Video tracking
Thesis advisors : P. Van Dooren, P.-A. Absil
Computer vision has been a very active topic in the last few decades but there is still progress needed in robust, reliable and stable algorithms (for handling problems such as noise, occlusion, deformable objects, etc..). We will focus on the real-time video tracking problem and explore different image segmentation techniques from a wide variety of domains such as graph cuts, community detection in graphs, snakes or active contours, partial differential equation solutions for energy minimization. We plan to adapt those techniques and develop new ones to address real-time video tracking in various fields such as remote sensing for collision detection and video surveillance.

CASON Thomas (Starting date : October 2006)
Differential-Geometric Methods in Graph Theory
Thesis advisors : P.A. Absil, P. Van Dooren
Several problems in graph theory require a mathematical analysis of their geometric structure. In particular, we recently focused on problems of graph matching using optimization techniques on manifolds.

CHANG Chia-Tche (Starting date : September 2008)
The joint spectral radius: from computation problems to continuous switching systems
Thesis advisor : V. Blondel
The joint spectral radius (JSR) characterizes the maximal asymptotic growth rate of a set of matrices. This quantity appears in many applications such as system theory, graphs and networks, coding theory or combinatorics on words, but has been proven to be NP-hard to approximate. We would like to investigate the JSR computation problem in a general framework on the one hand, and to explore properties of linear continuous switching systems using the JSR, on the other hand.

COPPE Sébastien (Starting date : September 2007)
The mechanism of prediction in motor control
Thesis advisors : Ph. Lefèvre, M. Missal
In our every day life, we always pursue some objects moving in our visual field. In many cases, the object we were tracking disappears for a moment (which could be short or longer). It is very important to be able to predict the reappearance of the target (spatially and temporally) for tracking it as best as possible as soon as it will reappear. This prediction also requires some internal models in the brain. The goal of my thesis is to better understand all of these mechanisms of prediction.

DECUYPER Adeline (Starting date : September 2011)
Dynamics and mathematics of time-evolving social networks
Thesis advisor : V. Blondel
The project focuses on the study of the dynamical aspects of large time-evolving graphs as representations of underlying social networks. The goal is then to develop mathematical models that reproduce properties observed on real social networks. Among other applications, this project will investigate the spreading dynamics of e.g. information or infectious diseases in social networks through numerical simulations and analysis of empirical data of real social networks.

DELHAYE Benoît (Starting date : September 2009)
Dexterous Manipulation in Microgravity: Mechanical properties of the fingertip-object contact
Thesis advisors : Ph. Lefèvre, J-L Thonnard
The aim of the project is to study the mechanical properties of the fingertip and their influence on the control of grip force when manipulating objects. More specifically, we will study how information from skin afferent s allows characterising the properties of an object when grasping it. This information is of prime importance to prevent the object from slipping with a minimal grip force as a function of the friction properties of the fingertip-object interface. Moreover, we will focus on the influence of a change in gravity on these mechanisms.

DEVILLE Pierre (Starting date : October 2011)
Detection and Stability Analysis of Communities in Large Social Networks
Thesis advisor : V. Blondel
Social, technological and information systems can often be described in terms of complex networks. The analysis of the structural and statistical properties of such networks is one of the major foci of complex systems science at the moment. The main objective of this research is to design methods that aim to quantify and visualize the stability of communities in large social networks. Communities in a network are almost connected subnetworks. Their analysis may have several applications as well as impact at different levels of our society (economical, sociological and political) and their interest is numerous.

DEVOLDER Olivier (Starting date : October 2009)
Infinite-dimensional structured convex optimization''
Thesis advisors : F. Glineur, Y. Nesterov
The goal of this thesis is to generalize results and algorithms developed for finite-dimensional structured convex optimization to the infinite-dimensional case, i.e. to problems with an infinite number of variables and/or constraints.

EGO Caroline (Starting date : September 2009)
Visual tracking in cerebral-palsied children: influence on learning and exploration of new avenues for rehabilitation
Thesis advisor : Ph. Lefèvre
Cerebral palsy (CP) is a group of permanent disorders of movement and posture due to a non-progressive lesion of the fetal or infant brain. The motor disorders in cerebral-palsied children are often accompanied by several disabilities including sensation, perception or cognition impairments. In particular, previous studies showed a high prevalence of learning disorders in cerebral-palsied children. The goal of the present study is to point out subclinical deficits in oculomotor performances of children with low level of severity of CP by comparing their performances to those recorded in age-matched control children between 6 and 14 years. Indeed, subtle oculomotor deficits such as impaired saccade pursuit interaction or poor anticipation and prediction skills in oculomotor strategies may be an important cause of learning disabilities. The underlying purpose of this study is to investigate new paths for rehabilitation in cerebral-palsied children.

GIARD Thibault (Starting date : September 2010)
Dextrous Manipulation in Microgravity: Motor Planning and Interaction with Robots
Thesis advisors : P. Lefèvre, J-L Thonnard
With the thumb opposing the index, humans have the ability to grasp and manipulate objects. More specifically, this grip permits to modulate the force developed to hold objects stable in our hands. Indeed, the force required to hold an object depends on many factors, including the object weight, the mechanical and frictional properties of the skin/object interface and the inertial constraints induced when the object is accelerated. In most situations encountered on daily bases, manipulating an object also implies the anticipation of static and inertial torques. In this project, two tools are used in order to study those aspects : parabolic flights and robotic arms (the Phantom by SensAble and the Kinarm by b-kin).

HOLLANDERS Romain (Starting date : September 2010)
On the Policy Iteration Algorithm for PageRank Optimization
Thesis advisors : J-C Delvenne, R. Jungers
In search engines, it is critical to be able to sort web pages according to their relative importance (or PageRank). Here, we study the specific problem of optimizing the PageRank of a node (i.e. a web page) in a directed graph (i.e. the web graph) when the presence of some edges (i.e. hyperlinks) is uncertain. We explain how this problem can be formulated as a Markov Decision Process. Classical algorithms can then be used to solve this new formulation of the problem. Among these, one of them, built following the principle of Policy Iteration, seems to perform very well in practice. Yet, its theoretical behavior is unknown. In the thesis, we study the complexity of this algorithm and we show that it converges in polynomial time under the natural assumption that the random walk restarts with a fixed probability called damping. We also try to extend our result to the case in which there is no damping and we give some clues towards new results in this direction.

KLEPPER Audrey Modelling and control of chemical processes
Thesis advisor : D. Dochain

KRINGS Gautier (Starting date : September 2007)
Modeling and analysis of large-scale evolving networks
Thesis advisor : V. Blondel
The topic of this thesis is the analysis of large evolving networks, such as mobile phone communication networks. Several results have been obtained lately in network theory, allowing a better understanding of the structure of social networks. However, most of them only apply to static graphs. The goal of this thesis is to analyze how social networks evolve, how edges are created and deleted and how our vision of social networks is affected by the time scales of the collected datasets.

LECLERCQ Guillaume (Starting date : September 2007)
Visuomotor velocity transformation for visually guided manual tracking
Thesis advisor : Ph. Lefèvre
Visually guided reaching movements are common tasks in our everyday life. To achieve such tasks, the brain needs to transform the retinal information into an appropriate motor command for the muscles. The goal of this research is to investigate if the brain accounts for the complex non-linear 3-dimensional eye-head-shoulder geometry when performing visually guided manual tracking tasks. A next step is to understand how it can be achieved in the brain.

MACLEAN Heather (Queen's University, Kingston, Ontario) (Starting date : January 2009)
Plant growth modelling
Thesis advisor : D. Dochain
One of the issue of the MELISSA project of the ESA is to grow plants as a source of food for the astronauts of the space shuttles. The main objective of this thesis is to develop and validate dynamical mass balance based models of the growth of selected plants that could be used for on-line optimization and control of plant growth within the constraints of the MELISSA loop and the space shuttle operation.

MELCHIOR Samuel (Starting date : September 2008)
Multi-mesh approach for model reduction of partial differential equations
Thesis advisors : V. Legat, P. Van Dooren
Multigrid is a powerful scheme for iterative solvers, especially as preconditioner for elliptic equations. We investigate the use of several discretization of a PDE to compute efficiently its optimal reduced model with respect to the H_2 norm. The goal is to apply this to ocean models.

ORBAN DE XIVRY François-Xavier (Starting date : 16 February 2009)
Nearest stable system
Thesis advisors : P. Van Dooren, Y. Nesterov
Stability is a crucial property in dynamical systems. Here, we are interested in finding a practical algorithm that would find the nearest stable system to an unstable one. We approach this problem from an optimization point of view using convex approximations and non-linear optimization techniques. In that framework, theoretical guarantees of the approximation and complexity bounds are important questions that require an answer. Associated problems such as the problem of finding the nearest stable polynomial to an unstable one will also be investigated.

RENTMEESTERS Quentin (Starting date : 15 September 2008)
Geometric data fitting and subspace tracking
Thesis advisors : Paul Van Dooren, P.-A. Absil
Several problems in signal processing and image processing deal with corrupted data which belong to a nonlinear space. In this research, we investigate filtering and approximation techniques on nonlinear spaces to reduce the influence of the noise on the data. We are particularly interested in applying these techniques to subspace tracking problems.

SIMON Emile (Starting date : May 2008)
Perspective for optimization in systems and control: from matrix inequalities to direct search methods
Thesis advisor : V. Wertz
The aim of this work is to observe and highlight limitations of the classical approach of optimization in systems and control: the linear matrix inequalities (LMIs), and then to propose an other class of methods not afflicted by these limitations to deal with still unresolved fundamental and special problems: the direct search methods.

TRAAG Vincent (Starting date : May 2009)
Social networks with positive and negative weights
Thesis advisor : P. Van Dooren
Negative links, such as animosity, bad references or inhibiting effects are not often considered in network analysis. The goal of this research is to understand the dynamics of such links. More specifically, we will analyze a model for reputation dynamics in the context of indirect reciprocity game theory.

TREFOIS Maguy
Algorithmics of structured matrices and algebraic graph theory
Thesis advisor : J.-C. Delvenne

WANG Xuansheng (Starting date : May 2011)
Nonnegative matrix factorization techniques
Thesis advisors : F. Glineur and P. Van Dooren
Development of efficient structure-preserving linear dimensionality-reduction techniques.

Post-Docs and Research Associates

DAVID Robert (Starting date : March 2008)
Identification of a model for the real-time application for riverflow forecasting
Host : G. Bastin
HYDROMAX is a real-time application for riverflow forecasting which is developed by CESAME and is operational at SETHY (Service d'études hydrologiques, Walloon Ministry of Public Works - Belgium) for the management of flood alarms and the on-line information of the rescue services in the Meuse river basin. HYDROMAX is connected with the telemetering network of SETHY. This network is essentially made up of water level chart recorders and rain gauges.
HYDROMAX provides in real-time:

short term predictions of riverflows based on rainfal and past riverflow measurements
long term flood forecastings based on weather forecasts

For each river basin, the predictions are produced by a mathematical model. HYDROMAX has been developed to be user friendly and to fulfill the real time forecasting requirements. HYDROMAX is successfully in operation since 1995 on the main Meuse tributaries (Semois, Ourthe, Lesse, Viroin, etc.).

DE KERCHOVE Cristobald (Starting date : September 2009)
Ranking Large Networks
Diffusion of information through large graphs. I generally work on real datasets - for example mobile phone networks - and I identify leaders who are efficient to spread some information or behavior in the network. Another trendy topic is the votes on the Web. I consider graphs of votes where users evaluate items as we can find in many web sites. A third point includes all sorts of algorithms based on some random walk on the graph. Community detection and visualization of large graphs are certainly two problems that I like looking at.

GULARTE Federico (Starting date : September 2011)
Wine-making modeling
Host : D. Dochain
The objective of this research is to use the knowledge provided by metabolic engineering, systems biology and microbiology (e.g. the role of yeast and its intra- and extra-cellular interactions during wine fermentation) in order to develop reaction kinetic models that can accurately describe links (metabolic pathways) between the cell metabolism and the product quality.

JUNGERS Raphaël (Starting date : June 2008)
Application of recent results concerning the joint spectral radius
The Joint Spectral Radius is an algebraical tool that has proven to be useful in several applications. Our aim is to study the links between these applications, especially in the case of binary matrices, and maybe in some large graphs problems.

MITRA Bivas (Starting date : July 2011)
Modeling and Analysis of the Internet
Hosts : V. Blondel, J-C Delvenne, J. Hendrickx
The explosive popularity of Internet has been accompanied by a wide range of internetworking problems. It is generally more efficient to assess solutions using analysis or simulation, provided the model is a good abstraction of the real network and applications. In this context, my research is primarily composed of two different subtasks:

Modeling Internet topology: In this subtask, we address the problem of efficient and accurate generation of the graph models for the Internet topology. We consider CAIDA trace as the reference topology and compare a set of observed properties with the proposed model. As an extension of this work, we aim to propose smart methodologies to accurately approximate the degree distribution of the Internet.
Modeling file diffusion in p2p network: The objective of the second subtask is to propose an analytical framework in order to model the popularity distribution of the files and activity of the users in the peer-to-peer networks.

MOENS Luc (Starting date : October 1994)
Identification of a model for the real-time application for riverflow forecasting
HYDROMAX is a real-time application for riverflow forecasting which is developed by CESAME and is operational at SETHY (Service d'études hydrologiques, Walloon Ministry of Public Works - Belgium) for the management of flood alarms and the on-line information of the rescue services in the Meuse river basin. HYDROMAX is connected with the telemetering network of SETHY. This network is essentially made up of water level chart recorders and rain gauges.
HYDROMAX provides in real-time:

short term predictions of riverflows based on rainfal and past riverflow measurements
long term flood forecastings based on weather forecasts

ORBAN DE XIVRY Jean-Jacques
Host : Philippe Lefèvre
Modelling of biomedical systems

ROCHA Luis (Starting date : October, 2011)
Temporal structures in empirical social networks
Host : Vincent Blondel
The interactions between people create contact networks in which infections and information can potentially spread. Within this context, my research is mostly focused on creating new methods to characterize temporal structures of empirical networks (e.g. sexual and communication networks), and on studying how the temporal and topological structures affect dynamical processes taking place on these evolving networks (e.g. epidemics or opinion formation).

SUVISESHAMUTHU Easter Selvan (Starting date : September 2010)
Optimization algorithms : A statistical signal processing perspective
Host : P.-A. Absil
We consider foundational aspects of statistical signal processing pertaining to optimization techniques, with a particular interest for Riemannian optimization. On the application side, the focus is on blind source separation techniques in various medical domains.