HAL - Lille 3
Not a member yet
5967 research outputs found
Sort by
Intelligent Proportional Differential Neural Network Control for Unknown nonlinear systems
No full textInternational audienc
Vectorization of local search for solving flow-shop scheduling problem on Xeon Phi™ MIC co-processors
No full textInternational audienceThis paper aims to propose a vectorizable cost function for the permutation flow-shop problem (PFSP) with the makespan criterion. This vectorization has been tested on a Xeon Phi core, using a local search. Indeed, Xeon Phi co-processors require vectorization in order to get the best performance from the device. Taillard's benchmark instances are used for the validation of the algorithm. The obtained results show that vectorization is more efficient as the number of jobs and the number of machines increase. Speedups up to 4.5× are recorded
On the Differential Privacy of Bayesian Inference
No full textInternational audienceWe study how to communicate findings of Bayesian inference to third parties, while preserving the strong guarantee of differential privacy. Our main contributions are four different algorithms for private Bayesian inference on proba-bilistic graphical models. These include two mechanisms for adding noise to the Bayesian updates, either directly to the posterior parameters, or to their Fourier transform so as to preserve update consistency. We also utilise a recently introduced posterior sampling mechanism, for which we prove bounds for the specific but general case of discrete Bayesian networks; and we introduce a maximum-a-posteriori private mechanism. Our analysis includes utility and privacy bounds, with a novel focus on the influence of graph structure on privacy. Worked examples and experiments with Bayesian naïve Bayes and Bayesian linear regression illustrate the application of our mechanisms
Reporting adverse events in cancer surgery randomized trials: A systematic review of published trials in oesophago-gastric and gynecological cancer patients
No full textInternational audienceBackground: Few reports describe how adverse events (AEs) are reported in cancer surgery trials.Materials and methods: We systematically reviewed 179 consecutive study reports issued between January 1, 1990 and November 15, 2014, which investigated surgery in oesophago-gastric (OG) or gynecologic (GY) cancer patients. Based on the reviewed reports, we assessed how AEs were reported according to CONSORT statement.Results: Morbidity assessment was the primary objective of 56 studies (31.3%). Postoperative AEs were described in 161 studies (90%). Definition of AEs and grading scale (NCI-CTC AE, Dindo-Clavien scale, etc …) were given in 27.3% and 16.8% of studies, respectively. AEs were reported by event and grade in 8.3% of studies. Definition of expectedness, seriousness, causality and safety population were present in 0.5%, 1.1%, 7.8%, and 7.2% of the studies, respectively. Reporting of AEs did not improve over time nor better in high-impact factor journals.Conclusion: The reporting of AEs in cancer trials investigating surgery needs to be improved
A Hybrid Method for Assigning Containers to AGVs in the Dynamic Environment of Container Terminals
No full textInternational audienceThe handling operations performed in ports require the use of equipment operating in a dynamic environment. Some tasks may not be fully carried out due to equipment failure or power breakdown that may occur particularly with the automated guided vehicles (AGV). The unavailability of equipment such as AGV has important consequences in terms of respecting the deadlines of different operations that a port should perform, such as the loading and unloading operations of ships. This situation can aggravate if there are also traffic problems in the port with some inaccessible network nodes. A part of the equipment will be blocked or the operations will take longer than expected if they don`t take the optimal path to connect the loading/unloading points and storage areas. These reasons confirm the usefulness of establishing a robust system able to resolve the problem of assigning containers in the static and dynamic environments. In a previous work, we developed a system for assigning containers in a static environment. In order to improve this method, we devote this paper to the study of the robustness of our system to the dynamic environment of the port. The numerical tests included in this paper show an adequate performance of our method for this particular dynamic environment
Model-Based Diagnosis of Multi-Track Level Crossing Plants
No full textAs is witnessed by railway statistics, Level crossing (LC) safety has always been one of the major concerns for railway stakeholders. LC safety is an issue at the crossroads between technical aspects, operational procedures and human factors, making the search for effective solutions a challenging task. This paper deals with technical aspects related to LC safety. In particular, we carry out an analysis pertaining to the diagnosability of two main failure classes that can affect the protection system at automatic LCs. In the current study, a labeled Petri net behavioral model depicting the global system function, including both the normal operation and the faulty behavior, is first established. Petri net has been used as the modeling formalism mainly for its mathematical foundations and expressiveness capabilities. Using such a mathematical notation is highly recommended to deal with dependability issues in safety-critical systems, especially in railways. Based on the established model, different model-based approaches for the diagnosis of discrete event systems (DESs) will be brought into play to investigate the diagnosability of two considered failure classes, while the obtained results will be compared. In particular, a technique that we have established, which is based on on-the-fly and incremental analysis of the model state space shows interesting efficiency, making it possible to tackle the combinatorial explosion problem, which arises particularly when considering multi-track LCs. The originality of this technique w.r.t existing DES model-based diagnosis approaches is that, in general, a partial building/investigation of the state space suffices to decide diagnosability and build an online diagnoser. Findings pertaining to LC safety are drawn based on a thorough discussion of the obtained results. In particular we show how the diagnosability analysis outputs can be taken into account in the global LC risk assessment process
Significance of Microscopically Incomplete Resection Margin After Esophagectomy for Esophageal Cancer
No full textInternational audienceOBJECTIVE:The objectives of this study were to establish if R1 resection margin after esophagectomy was (i) a poor prognostic factor independent of patient and tumor characteristics, (ii) a marker of tumor aggressiveness and (iii) to look at the impact of adjuvant treatment in this subpopulation.METHODS:Data were collected from 30 European centers from 2000 to 2010. Patients with an R1 resection margin (n = 242) were compared with those with an R0 margin (n = 2573) in terms of short- and long-term outcomes. Propensity score matching and multivariable analyses were used to compensate for differences in baseline characteristics.RESULTS:Independent factors significantly associated with an R1 resection margin included an upper third esophageal tumor location, preoperative malnutrition, and pathological stage III. There were significant differences between the groups in postoperative histology, with an increase in pathological stage III and TRG 4-5 in the R1 group. Total average lymph node harvests were similar between the groups; however, there was an increase in the number of positive lymph nodes seen in the R1 group. Propensity matched analysis confirmed that R1 resection margin was significantly associated with reduced overall survival and increased overall, locoregional, and mixed tumor recurrence. Similar observations were seen in the subgroup that received neoadjuvant chemoradiation. In R1 patients adjuvant therapy improved survival and reduced distant recurrence however failed to affect locoregional recurrence.CONCLUSIONS:This large multicenter European study provides evidence to support the notion that R1 resection margin is a prognostic indication of aggressive tumor biology with a poor long-term prognosis.TRIAL REGISTRATION:ClinicalTrials.gov NCT01927016
SLEEP OBSTRUCTIVE APNEA SYNDROME (SOAS) IS NOT ASSOCIATED WITH MARKED ERYTHROCYTOSIS. A STUDY ON 337 PATIENTS
No full text21st Congress of the European-Hematology-Association, Copenhagen, DENMARK, JUN 09-12, 2016International audienc
Computational Modeling of Human Social Intention
No full textInternational audienceIn this paper, we propose to analyze the kinematic of human arm to predict social intention (personal vs. social intention). We propose to use infrared markers placed on the arms of two human subjects. The trajec-tories of different markers are defined as shape spaces and subsequently analyzed as a Riemannian manifold for the purposes of characterizing the intention (personal vs. social). The first results show that the best results are obtained with the wrist marker (69.2%)