Search

Objectifs : Ce travail a été réalisé dans la côte sableuse Nord du Sénégal (la grande côte) allant de Yoff à Saint-Louis dans le but de connaitre la composition spécifique, déterminer la structure de la flore (spectre taxonomique) et sa répartition tout au long de cette bande côtière. Méthodologie et résultats : Des inventaires systématiques ont été entrepris de juillet 2017 au mois d’avril 2019 soit 4 récoltes par site. Ces travaux ont permis de répertorier 21 espèces réparties dans 3 classes, 11 ordres, 15 familles et 17 genres. La classe des Florideophyceae (algue rouge) domine avec 57,14 % des espèces identifiées suivie de la classe des Ulvophyceae (algue verte) avec 23,8 % et enfin des Phaeophyceae (algue brune) avec 19,04 % des espèces. Les familles des Dictyotaceae (14,28 %), des Corallinaceae, des Gelidiaceae, des Bryopsidiaceae avec chacune 9,52 % des familles sont les plus diversifiées. L’étude a montré que 74,07 % des espèces ont été retrouvées à Yoff, 11,11 % à Kayar, 7,04 % des espèces à Mboro et 3,7 % aussi bien à Loumpoul qu’à Saint-Louis. Conclusion et application des résultats : Ce travail a permis de déterminer la structure et la répartition de la flore macroalgale de la côte Nord du Sénégal. Ces informations sont importantes pour une éventuelle exploitation des algues à potentialité économique telles que le Meristotheca senegalensis qui est une espèce retrouvée à Yoff, commercialisée puis exportée vers la chine et le japon. English abstract Objective: This work constitutes a systematic inventory of seaweeds from the northern sandy coast of Senegal, from Yoff to Saint-Louis. Methodology and results: The study was undertaken from July 2017 to April 2019 or 4 crops for station to know the specific diversity, determine the structure of the flora (taxonomic spectrum) and its distribution throughout this coastal strip. The inventory identified 21 species in 3 classes, 11 orders, 15 families and 15 genera. The class of Florideophyceae (red algae) dominates with 57.14% of identified species followed by the class of Ulvophyceae (green algae) with 23.8 % and in the end the Phaeophyceae (brown algae) with 19.04 % of identified species. The families of Dictyotaceae (14.28 %), Corallinaceae, Gelidiaceae and Bryopsidiaceae with each 9.52 % of the families are the most diversified. The study showed that 74.07 % of species were found in Yoff, 11.11 % in Kayar, 7.04 % of species in Mboro and 3.7 % in both Loumpoul and Saint-Louis. Conclusion and application of results: This work made it possible to determine the structure and distribution of the macroalgal flora of the northern coast of Senegal. This information is important for a possible exploitation of algae with economic potential such as Meristotheca senegalensis which is a species traded then exported to China and Japan. Key words: Seaweed, Diversity, division, North Coast, Senegal.

Context and objective. Acute Respiratory Infections (ARIs) are one of the first reasons for consultation in children and often result in a prescription of antibiotics. The present study aimed to evaluate knowledge of health workers on the ambulatory antibiotic prescription for ARI in children. Methods. This cross sectional study was a structured questionnaire-based survey on the knowledge, attitude and practice of health workers. The study was conducted in the Dakar medical region from July 13thto September 4th, 2018 and included health personnel that was qualified and authorized to prescribe drugs in children; trainees were excluded.Results. We interviewed 83 health professionals working in 20 health structures. Nearly 60% of them were general practitioners. Nasopharyngitis and angina were the most common ARIs. For all ARIs combined, antibiotics were prescribed in 72.8% of cases. These prescriptions were made in 100% of angina cases, 41% of rhinopharyngitis cases and 62% of bronchiolitis cases. Conclusion. Antibiotics are over-prescribed in ARF. The availability of precise algorithms for the management of ARF could reduce these antibiotic prescriptions and prevent the selection of multi-resistant germs. Contexte et objectif. Les infections respiratoires aiguës (IRA) constituent l’un des premiers motifs de consultation chez l’enfant et occasionnent souvent une prescription d’antibiotiques. L’objectif de cette étude était d’évaluer la prescription des antibiotiques en ambulatoire dans les IRA chez l’enfant.Méthodes. Il s’agissait d’une enquête type CAP (connaissance, attitude et pratique) du 13 juillet au 4 septembre 2018, menée au niveau de la région médicale de Dakar. Nous avions inclus tout personnel de santé qualifié et habilité à faire des prescriptions chez l’enfant; les stagiaires étaient exclus. Résultats. Nous avons interrogé 83 professionnels de santé exerçant dans 20 structures sanitaires. Près de 60% des professionnels interrogés étaient des médecins généralistes. La rhinopharyngite et les angines étaient les IRA les plus fréquentes. Pour toutes les IRA confondues, les antibiotiques ont été prescrits dans 72,8% des cas. Ces prescriptions étaient faites dans 100% des cas d’angine, 41% des rhinopharyngites et 62% des bronchiolites. Conclusion. Les antibiotiques sont surprescrits dans les IRA qui sont majoritairement virales. La mise à disposition d’algorithme précis sur la prise en charge des IRA pourrait diminuer ces prescriptions d’antibiotiques et prévenir la sélection de germes multirésistants.

Excessive weight gain during pregnancy is associated with higher risks of complications in the mother and in the newborn. We carried out this work, the objective of which was to study the influence of excessive maternal weight gain during pregnancy. This was a prospective study carried out in Dakar over a period of 5 months. We included all women with a normal body mass index in early pregnancy. These parturients were divided into two groups: the first group consisted of women whose weight gain during pregnancy was normal between 11.5 and 16 kg and the second group were those who had an excessive weight gain greater than 16 kg. We analyzed maternal data and neonatal parameters. During the period, 194 women had a normal weight gain (PNP) and 104 an excessive weight gain (PEP), i.e., a prevalence of PEP of 10.5%. There were no differences in the two groups for marital status, geographic origin, level of education, drinking habits, maternal size and pathologies during pregnancy (p>0.05). On the other hand, maternal age, parity and absence of a professional activity were significantly associated with excessive weight gain during pregnancy. Regarding neonatal data, we did not note any significant difference with regard to gender, trophicity, Apgar score, stillbirth, postnatal transfer and neonatal mortality between the 2 groups.

Introduction: Breastfeeding is the benchmark for feeding infants during the first months of life, regardless of their birth weight. The objective of our study was to compare the weight growth of a cohort of low birth weight (LBW) newborns in the first 6 months of life according to the type of breastfeeding adopted by the mother. Methodology: It was a prospectiveand analytical study of a cohort of newborns from LBW. It took place over a period of 17 months at the level of a reference structure in Dakar. We have included all newborns born “in-born” weighing less than 2500 g with a threshold of 1000 g. The data was analyzed with SPSS version 17 software. Results: The average birth weight of newborns was 1821.8 g with extremes ranging from 1000 g to 2400 g and the median weight of 1840 g. The majority of our population had a birth weight ≥ 1500 g (83%) with more than half of them being premature (58.9%). Depending on the type of breastfeeding adopted by the mother, 33/112 newborns were on exclusive breastfeeding (EBF). During the first 6 months of life, the average weight gain was significantly better in infants on EBF (747.9 g) compared to infants on mixed breastfeeding (728.9 g) and artificial breastfeeding (676.6 g). Conclusion: Better awareness of mothers should be made during hospitalization and during follow-up consultations for newborns of LBW to encourage the continuation of the EBF which remains the best food for this infant population.

8 pages; National audience; La factorisation biaisée de matrices offre une grande qualité de prédiction pour les systèmes de recommandation. Mais elle attribut un comportement uniforme à tous les utilisateurs et produits. Ce papier proposons une amélioration de la factorisation biaisée de matrices. Notre proposition utilise des groupes de produits similaires. Elle attribue à chaque utilisateur un biais pour chaque groupe de produits. Nos expérimentations sur des jeux de données réels montrent l'efficacité de notre approche.

12 pages; National audience; La factorisation de matrices offre une grande qualité de prédiction pour les systèmes de recommandation. Mais sa nature statique empêche de tenir compte des nouvelles notes que les utilisateurs produisent en continu. Ainsi, la qualité des prédictions décroît entre deux factorisations lorsque de nombreuses notes ne sont pas prises en compte. La quantité de notes écartées est d'autant plus grande que la période entre deux factorisation est longue, ce qui accentue la baisse de qualité. Nos travaux visent à améliorer la qualité des recommandations. Nous proposons une factorisation de matrices utilisant des groupes de produits et intégrant en ligne les nouvelles notes des utilisateurs. Nous attribuons à chaque utilisateur un biais pour chaque groupe de produits similaires que nous mettons à jour. Ainsi, nous améliorons significativement les prédictions entre deux factorisations. Nos expérimentations sur des jeux de données réels montrent l'efficacité de notre approche.

Introduction: Prematurity is one of the leading causes of neonatal death in Africa. The objective of this work was to assess the preterm birth at the pediatric service of the regional hospital of Ziguinchor. Material and methods: This was a prospective, descriptive and analytical study of the case of hospitalized newborns whose age was between 32 SA to 36SA + 6 days. The study was conducted from May 21, 2013 to May 21, 2014. We included all premature infants at age ≥ 32SA.Those presenting a malformation were not included. We studied maternal socio-demographic, obstetric and neonatal parameters. Results: We have identified 140 newborns out of a total of 342 Preterm births and 2292 maternity births, a prevalence of 40.9% compared to preterm infants and 6.10% in relation to all births. 51.4 percent of mothers came from urban areas, 63.5 percent were between 20 and 34 years of age, not attending school in 45.7 percent, married in 72.9 percent and without work in 94.3 percent. The average gestures represented 3.09 . Arterial hypertension was the most common medical condition (67.7%). The followup was done by a midwife (82.1%) and 23.5% had received at least 4 NPC. Eclampsia and pre-eclampsia (48.9%) and RPM (42.8%) were common obstetric pathologies. The delivery was by caesarian (51.4%), hospital (92.1%), cephalic presentation (80%), with an average weight of 1816.79 g. An RCIU (16.4%), an RPM (61.9%). At birth, a DR (16.4%) was noted due to MMH (43.7%) and transitory tachypnea (50%). Asphyxia was reported in 3.5%. During hospitalization (6.7 days on average), 6.4% had RD caused by infection (77.7%). Other complications were infection (64.4%), hypoglycemia (28.5%), digestive hemorrhage (7.01%). The fatality rate was 7.9% due to infection (63.3%), DR (18%), haemorrhage (9%). The anthropometric measurements at the exit: P 1887.9g, T: 38.3cm, PC: 30.9cm; at 1 month P: 2387.1g, T: 46.6cm, PC: 32.8cm; at 30 months P: 12.1kg, T: 89.5cm, PC: 48.4cm. Conclusion: Moderate preterm birth accounts for almost half of the cases of prematurity in our series. Their optimal management would go through a better obstetric-neonatal collaboration but above all by the installation of kangaroo mother units

In this thesis, we address the scalability problem of recommender systems. We propose accu rate and scalable algorithms. We first consider the case of matrix factorization techniques in a dynamic context, where new ratings..are continuously produced. ln such case, it is not possible to have an up to date model, due to the incompressible time needed to compute it. This happens even if a distributed technique is used for matrix factorization. At least, the ratings produced during the model computation will be missing. Our solution reduces the loss of the quality of the recommendations over time, by introducing some stable biases which track users' behavior deviation. These biases are continuously updated with the new ratings, in order to maintain the quality of recommendations at a high leve for a longer time. We also consider the context of online social networks and tag recommendation. We propose an algorithm that takes account of the popularity of the tags and the opinions of the users' neighborhood. But, unlike common nearest neighbors' approaches, our algorithm doe not rely on a fixed number of neighbors when computing a recommendation. We use a heuristic that bounds the network traversai in a way that allows to faster compute the recommendations while preserving the quality of the recommendations. Finally, we propose a novel approach that improves the accuracy of the recommendations for top-k algorithms. Instead of a fixed list size, we adjust the number of items to recommend in a way that optimizes the likelihood that ail the recommended items will be chosen by the user, and find the best candidate sub-list to recommend to the user.; Cette thèse s'intéresse à la problématique de passage à l'échelle des systèmes de recommandations. Dans ce cadre, nous proposons deux algorithmes de recommandation passant à l'échelle tout en délivrant une bonne qualité de recommandation. Dans nos premiers travaux, nous considérons l'utilisation de la factorisation de matrice pour prédire les avis des utilisateurs dans des contextes dynamiques où les utilisateurs soumettent sans cesse de nouveaux avis sur des objets. Il est difficile d'y tenir un modèle de prédiction à jour. Nous proposons un modèle de factorisation utilisant plusieurs biais locaux décrivant de façon plus élaborée les comportements des utilisateurs. Leur coût de calcul faible permet de les ajuster à la volée, lorsque de nouvelles notes arrivent. Ils assurent ainsi la robustesse du modèle dans un contexte dynamique, tout en garantissant une meilleure qualité au cours le temps. Nous nous sommes aussi intéressés à la recommandation de tags dans les réseaux sociaux. Notre proposition s'appuie sur l'algorithme des plus proches voisins. Cependant, nous déterminons dynamiquement le nombre optimal de voisins à utiliser. Notre approche prend en compte les avis des voisins indirects en étendant la relation de proximité entre les utilisateurs. Nos expérimentations ont démontré l'efficacité de cette approche qui peut être adaptée à bien d'autres types de recommandation. Enfin, nous proposons une technique d'optimisation du nombre d'objets à recommander en fonction du contexte. L'idée étant ici de chercher le nombre optimal d'objets à proposer à l'utilisateur, de telle sorte que la probabilité qu'il les choisisse tous soit la plus élevée.

National audience; It is today accepted that matrix factorization models allow a high quality of rating prediction in recommender systems.However, a major drawback of matrix factorization is its static nature that results in a progressive declining of the accuracy of the predictions after each factorization. This is due to the fact that the new obtained ratings are not taken into account until a new factorization is computed, which can not be done very often because of the high cost of matrix factorization.In this paper, aiming at improving the accuracy of recommender systems, we propose a cluster-based matrix factorization technique that enables online integration of new ratings. Thus, we significantly enhance the obtained predictions between two matrix factorizations. The experiments we did on a large dataset demonstrated the efficiency of our approach. Furthermore, we devise an experimental protocol that allows for identifying the optimal number of ratings to take into account for factorization. This contributes as a practical solution addressing the tradeoff between taking into account more ratings into factorization for better quality, versus missing fewer ratings during factorization.

8 pages; International audience; Tag recommendation is a major aspect of collaborative tagging systems. It aims to recommend tags to a user for a given item. In this paper we propose an adaptation of the search algorithms proposed in [14, 15, 1] to the tag recommendation problem. Our algorithm, called STRec, provides network-aware recommendations based on proximity measures computed on-the-fly in the network. STRec uses a bounded search to find good neighbors. On top of STRec, we apply a re-ranking scheme that improves the quality of the recommendations. We update the ranking according to the degree of association between the higher ranked tags and the lower ranked ones. This technique leads to better recommendations as we show in this paper and could be applicable on top of many recommender systems. The experiments we did on several datasets demonstrated the efficiency of our approach.

4 pages; International audience; Tag recommendation is a major aspect of collaborative tagging systems. It aims to recommend tags to a user for tagging an item. In this paper we present a part of our work in progress which is a novel improvement of recommendations by re-ranking the output of a tag recommender. We mine association rules between candidates tags in order to determine a more consistent list of tags to recommend. Our method is an add-on one which leads to better recommendations as we show in this paper. It is easily parallelizable and morever it may be applied to a lot of tag recommenders. The experiments we did on five datasets with two kinds of tag recommender demonstrated the efficiency of our method.

It is today accepted that matrix factorization models allow a high quality of rating prediction in recommender systems. However, a major drawback of matrix factorization is its static nature that results in a progressive declining of the accuracy of the predictions after each factorization. This is due to the fact that the new obtained ratings are not taken into account until a new factorization is computed, which can not be done very often because of the high cost of matrix factorization. In this paper, aiming at improving the accuracy of recommender systems, we propose a cluster-based matrix factorization technique that enables online integration of new ratings. Thus, we significantly enhance the obtained predictions between two matrix factorizations. We use finer-grained user biases by clustering similar items into groups, and allocating in these groups a bias to each user. The experiments we did on large datasets demonstrated the efficiency of our approach., 31 pages, 7 figures