ORBi Collection: Computer science
http://hdl.handle.net/2268/85
The Collection's search engineSearch this channelsearch
http://orbi.ulg.ac.be/simple-search
STEREO MATCHING METHOD USING MULTIPLE ANGULAR SIGNATURES MATCHED BY MULTIDIMENSIONAL DYNAMIC TIMEWARPING (MD-DTW)
http://hdl.handle.net/2268/173963
Title: STEREO MATCHING METHOD USING MULTIPLE ANGULAR SIGNATURES MATCHED BY MULTIDIMENSIONAL DYNAMIC TIMEWARPING (MD-DTW)
<br/>
<br/>Author, co-author: Grogna, David; Guerri, Géraldine; Verly, Jacques
<br/>
<br/>Abstract: We consider the problem of matching interest points (IPs) extracted
from the two images of a stereoscopic pair. We describe
each IP via a descriptor consisting in one angular signature for
a grayscale image, three signatures for a color image, and, more
generally, N signatures for N coregistered images of different modalities.
Each angular signature is generated by spinning a “steerable
x-tapered, y-derivative, half-Gaussian filter” 360° around
each IP. The novel contribution of the paper is the use of multidimensional
dynamic time warping (MD-DTW) for (1) producing a
single, common warping for all pairs of corresponding signatures
in the two descriptors of two IPs that are candidate for matching,
and (2) for computing a single related distance between the two
sets of signatures. Preliminary experimental results and a comparison
with the results obtained with SIFT descriptors and matching
via the smallest Euclidean distance in parameter space demonstrates
the value of this novel approach.Active network management for electrical distribution systems: problem formulation and benchmark
http://hdl.handle.net/2268/173942
Title: Active network management for electrical distribution systems: problem formulation and benchmark
<br/>
<br/>Author, co-author: Gemine, Quentin; Ernst, Damien; Cornélusse, Bertrand
<br/>
<br/>Abstract: In order to operate an electrical distribution network in a secure and cost-efficient way, it is necessary, due to the rise of renewable energy-based distributed generation, to develop Active Network Management (ANM) strategies. These strategies rely on short-term policies that control the power injected by generators and/or taken off by loads in order to avoid congestion or voltage problems. While simple ANM strategies would curtail the production of generators, more advanced ones would move the consumption of loads to relevant time periods to maximize the potential of renewable energy sources. However, such advanced strategies imply solving large-scale optimal sequential decision-making problems under uncertainty, something that is understandably complicated. In order to promote the development of computational techniques for active network management, we detail a generic procedure for formulating ANM decision problems as Markov decision processes. We also specify it to a 75-bus distribution network. The resulting test instance is available at http://www.montefiore.ulg.ac.be/~anm/ . It can be used as a test bed for comparing existing computational techniques, as well as for developing new ones. A solution technique that consists in an approximate multistage program is also illustrated on the test instance.Systematic analysis of site-specific yield distributions resulting from nitrogen management and climatic variability interactions
http://hdl.handle.net/2268/173632
Title: Systematic analysis of site-specific yield distributions resulting from nitrogen management and climatic variability interactions
<br/>
<br/>Author, co-author: Dumont, Benjamin; Basso, Bruno; Leemans, Vincent; Bodson, Bernard; Destain, Jean-Pierre; Destain, Marie-France
<br/>
<br/>Abstract: At the plot level, crop simulation models such as STICS have the potential to evaluate risk associated with management practices. In nitrogen (N) management, however, the decision-making process is complex because the decision has to be taken without any knowledge of future weather conditions. The objective of this paper is to present a general methodology for assessing yield variability linked to climatic uncertainty and variable N rate strategies. The STICS model was coupled with the LARS-Weather Generator. The Pearson system and coefficients were used to characterise the shape of yield distribution. Alternatives to classical statistical tests were proposed for assessing the normality of distributions and conducting comparisons (namely, the Jarque-Bera and Wilcoxon tests, respectively). Finally, the focus was put on the probability risk assessment, which remains a key point within the decision process. The simulation results showed that, based on current N application practice among Belgian farmers (60 60 60 kgN ha-1), yield distribution was very highly significantly non normal, with the highest degree of asymmetry characterised by a skewness value of -1.02. They showed that this strategy gave the greatest probability (60%) of achieving yields that were superior to the mean (10.5 t ha-1) of the distribution.Mathematical Modeling of HIV Dynamics After Antiretroviral Therapy Initiation: A Review
http://hdl.handle.net/2268/173614
Title: Mathematical Modeling of HIV Dynamics After Antiretroviral Therapy Initiation: A Review
<br/>
<br/>Author, co-author: Rivadeneira, Pablo; Moog, Claude; Stan, Guy-Bart; Brunet, Cécile; Raffi, François; Ferré, Virginie; Costanza, Vicente; Mhawej, Marie-José; Biafore, Federico; Ouattara, Djomangan; Ernst, Damien; Fonteneau, Raphaël; Xia, Xiaohua
<br/>
<br/>Abstract: This review shows the potential ground-breaking impact that mathematical tools may have in the analysis and the understanding of the HIV dynamics. In the first part, early diagnosis of immunological failure is inferred from the estimation of certain parameters of a mathematical model of the HIV infection dynamics. This method is supported by clinical research results from an original clinical trial: data just after 1 month following therapy initiation are used to carry out the model identification. The diagnosis is shown to be consistent with results from monitoring of the patients after 6 months. In the second part of this review, prospective research results are given for the design of individual anti-HIV treatments optimizing the recovery of the immune system and minimizing side effects. In this respect, two methods are discussed. The first one combines HIV population dynamics with pharmacokinetics and pharmacodynamics models to generate drug treatments using impulsive control systems. The second one is based on optimal control theory and uses a recently published differential equation to model the side effects produced by highly active antiretroviral therapy therapies. The main advantage of these revisited methods is that the drug treatment is computed directly in amounts of drugs, which is easier to interpret by physicians and patients.Mathematical Modeling of HIV Dynamics After Antiretroviral Therapy Initiation: A Review
http://hdl.handle.net/2268/173614
Title: Mathematical Modeling of HIV Dynamics After Antiretroviral Therapy Initiation: A Review
<br/>
<br/>Author, co-author: Rivadeneira, Pablo; Moog, Claude; Stan, Guy-Bart; Brunet, Cécile; Raffi, François; Ferré, Virginie; Costanza, Vicente; Mhawej, Marie-José; Biafore, Federico; Ouattara, Djomangan; Ernst, Damien; Fonteneau, Raphaël; Xia, Xiaohua
<br/>
<br/>Abstract: This review shows the potential ground-breaking impact that mathematical tools may have in the analysis and the understanding of the HIV dynamics. In the first part, early diagnosis of immunological failure is inferred from the estimation of certain parameters of a mathematical model of the HIV infection dynamics. This method is supported by clinical research results from an original clinical trial: data just after 1 month following therapy initiation are used to carry out the model identification. The diagnosis is shown to be consistent with results from monitoring of the patients after 6 months. In the second part of this review, prospective research results are given for the design of individual anti-HIV treatments optimizing the recovery of the immune system and minimizing side effects. In this respect, two methods are discussed. The first one combines HIV population dynamics with pharmacokinetics and pharmacodynamics models to generate drug treatments using impulsive control systems. The second one is based on optimal control theory and uses a recently published differential equation to model the side effects produced by highly active antiretroviral therapy therapies. The main advantage of these revisited methods is that the drug treatment is computed directly in amounts of drugs, which is easier to interpret by physicians and patients.Meta-learning of exploration-exploitation strategies in reinforcement learning
http://hdl.handle.net/2268/173595
Title: Meta-learning of exploration-exploitation strategies in reinforcement learning
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]Batch mode reinforcement learning based on the synthesis of artificial trajectories
http://hdl.handle.net/2268/173593
Title: Batch mode reinforcement learning based on the synthesis of artificial trajectories
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]Learning for exploration-exploitation in RL. The dusk of the small formulas’ reign
http://hdl.handle.net/2268/173591
Title: Learning for exploration-exploitation in RL. The dusk of the small formulas’ reign
<br/>
<br/>Author, co-author: Ernst, DamienBeyond function approximators for batch mode reinforcement learning: rebuilding trajectories
http://hdl.handle.net/2268/173590
Title: Beyond function approximators for batch mode reinforcement learning: rebuilding trajectories
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]Model-free Monte Carlo-like policy evaluation
http://hdl.handle.net/2268/173589
Title: Model-free Monte Carlo-like policy evaluation
<br/>
<br/>Author, co-author: Ernst, DamienLower bounds in reinforcement learning: the intelligent agent dream is getting closer
http://hdl.handle.net/2268/173586
Title: Lower bounds in reinforcement learning: the intelligent agent dream is getting closer
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]A reinforcement learning approach for designing artificial autonomous intelligent agents
http://hdl.handle.net/2268/173583
Title: A reinforcement learning approach for designing artificial autonomous intelligent agents
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]Clinical data based optimal STI strategies for HIV: a reinforcement learning approach
http://hdl.handle.net/2268/173582
Title: Clinical data based optimal STI strategies for HIV: a reinforcement learning approach
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]Computing near-optimal policies from trajectories by solving a sequence of standard supervised learning problems
http://hdl.handle.net/2268/173581
Title: Computing near-optimal policies from trajectories by solving a sequence of standard supervised learning problems
<br/>
<br/>Author, co-author: Ernst, Damien
<br/>
<br/>Commentary: [Talk given at several other places]Supervised learning based sequential decision making
http://hdl.handle.net/2268/173580
Title: Supervised learning based sequential decision making
<br/>
<br/>Author, co-author: Ernst, DamiengammaMAXT: a fast multiple-testing correction algorithm
http://hdl.handle.net/2268/173465
Title: gammaMAXT: a fast multiple-testing correction algorithm
<br/>
<br/>Author, co-author: Van Lishout, François; Gadaleta, Francesco; Moore, Jason H.; Wehenkel, Louis; Van Steen, Kristel
<br/>
<br/>Abstract: The purpose of the maxT algorithm (1993) is to control the family-wise error rate (FWER) when assessing significance of multiple tests jointly. However, the requirements in terms of computing time and memory of this procedure are proportional to the number of investigated hypothesis. The memory issue has been solved by Van Lishout’s implementation of maxT (2013), which makes the memory usage independent from the size of the dataset. This algorithm is implemented in MBMDR-3.0.3, a software that is able to identify genetic interactions, for a variety of SNP-SNP based epistasis model,s in an effective way. However, that implementation turned out to be less suitable for genome-wide interaction analysis studies, due to the prohibitive computational burden. Here, we present gammaMAXT, a novel algorithm which is part of MBMDR-4.2.2. We show that, in the abscence of interaction effects, test-statistics produced by the MB-MDR methodology follow a mixture distribution with a point mass at zero and a shifted gamma distribution for the top 10% of the strictly positive values. We show that the gammaMAXT algorithm has a power comparable to maxT and maintains FWER, but requires less computational resources and time. MBMDR-4.2.2 can be downloaded at http://www.statgen.ulg.ac.be.Acceleration of Affine Hybrid Transformations
http://hdl.handle.net/2268/173234
Title: Acceleration of Affine Hybrid Transformations
<br/>
<br/>Author, co-author: Boigelot, Bernard; Herbreteau, Frédéric; Mainz, Isabelle
<br/>
<br/>Abstract: This work addresses the computation of the set of reachable configurations of linear hybrid automata. The approach relies on symbolic state-space exploration, using acceleration in order to speed up the computation and to make it terminate for a broad class of systems. Our contribution is an original method for accelerating the control cycles of linear hybrid automata, i.e., to compute their unbounded repeated effect. The idea consists in analyzing the data transformations that label these cycles, by reasoning about the geometrical features of the corresponding system of linear constraints. This approach is complete over Multiple Counters Systems (MCS), and is able to accelerate hybrid transformations that are out of scope of existing techniques.Impact de l’anisotropie du milieu dans les études de profilage géographique
http://hdl.handle.net/2268/172997
Title: Impact de l’anisotropie du milieu dans les études de profilage géographique
<br/>
<br/>Author, co-author: Trotta, Marie; Deprez, Cécile; Donnay, Jean-Paul
<br/>
<br/>Abstract: Cette communication a pour objectif d’évaluer le biais introduit par l’usage de la distance euclidienne dans le cadre d’analyses de profilage géographique. Elle procède par une comparaison des distances routières et euclidiennes calculées sur près de 1 000 couples d’origines-destinations (sites de crimes – point d’ancrage du criminel) issus de la base de données de la police fédérale belge. Les comparaisons sont différenciées selon la densité du réseau de voirie (milieux rural et urbanisé) et les paramètres statistiques des distributions résultantes sont exploités dans l’application du modèle de décroissance avec la distance, principal outil de profilage géographique.; This communication aims at assessing the bias introduced by the use of the Euclidean distance in the context of geographic profiling analyzes. It proceeds by a comparison of road and Euclidean distances of about 1 000 pairs of origin-destination (crime sites – criminal’s anchor point) from the database of the Belgian Federal Police. Comparisons are differentiated by the density of road network (rural and urban areas) and the statistical parameters of the resulting distributions are exploited in the application of the most common geographic profiling model called the distance decay model.Lipschitz robust control from off-policy trajectories
http://hdl.handle.net/2268/172988
Title: Lipschitz robust control from off-policy trajectories
<br/>
<br/>Author, co-author: Fonteneau, Raphaël; Ernst, Damien; Boigelot, Bernard; Louveaux, Quentin
<br/>
<br/>Abstract: We study the minmax optimization problem introduced in [Fonteneau et al. (2011), ``Towards min max reinforcement learning'', Springer CCIS, vol. 129, pp. 61-77] for computing control policies for batch mode reinforcement learning in a deterministic setting with fixed, finite optimization horizon. First, we state that the $\min$ part of this problem is NP-hard. We then provide two relaxation schemes. The first relaxation scheme works by dropping some constraints in order to obtain a problem that is solvable in polynomial time. The second relaxation scheme, based on a Lagrangian relaxation where all constraints are dualized, can also be solved in polynomial time. We theoretically show that both relaxation schemes provide better results than those given in [Fonteneau et al. (2011)]Simultaneous perturbation algorithms for batch off-policy search
http://hdl.handle.net/2268/172982
Title: Simultaneous perturbation algorithms for batch off-policy search
<br/>
<br/>Author, co-author: Fonteneau, Raphaël; Prashanth L.A.
<br/>
<br/>Abstract: We propose novel policy search algorithms in the context of off-policy, batch mode reinforcement learning (RL) with continuous state and action spaces. Given a batch collection of trajectories, we perform off-line policy evaluation using an algorithm similar to that in [Fonteneau et al. (2010)]. Using this Monte-Carlo like policy evaluator, we perform policy search in a class of parameterized policies. We propose both first order policy gradient and second order policy Newton algorithms. All our algorithms incorporate simultaneous perturbation estimates for the gradient as well as the Hessian of the cost-to-go vector, since the latter is unknown and only biased estimates are available. We demonstrate their practicality on a simple 1-dimensional continuous state space problem.