Archive ouverte de l'ESSEC Business School
Not a member yet
1344 research outputs found
Sort by
Collective Matrix Completion
No full textInternational audienceMatrix completion aims to reconstruct a data matrix based on observations of a small number of its entries. Usually in matrix completion a single matrix is considered, which can be, for example, a rating matrix in recommendation system. However, in practical situations, data is often obtained from multiple sources which results in a collection of matrices rather than a single one. In this work, we consider the problem of collective matrix completion with multiple and heterogeneous matrices, which can be count, binary, continuous, etc. We first investigate the setting where, for each source, the matrix entries are sampled from an exponential family distribution. Then, we relax the assumption of exponential family distribution for the noise and we investigate the distribution-free case. In this setting, we do not assume any specific model for the observations. The estimation procedures are based on minimizing the sum of a goodness-of-fit term and the nuclear norm penalization of the whole collective matrix. We prove that the proposed estimators achieve fast rates of convergence under the two considered settings and we corroborate our results with numerical experiments
Évaluation de la gouvernance de l'information : une approche multifacette et multicritères
No full textNational audienceFace aux enjeux représentés par l'information et son rôle dans la prise de décision, les entreprises sont de plus en plus amenées à passer du management de cette information à sa gouvernance. L'objectif de cet article est de présenter une approche d'évaluation de cette gouvernance dont l'avantage est de prendre en compte les points de vue des différentes parties prenantes. Cette démarche est fondée sur une évaluation multifacette et multicritères, reposant sur une hiérarchie générique de critères d'évaluation, ensuite spécialisée par partie prenante (management, gouvernance de l'information, métiers). La hiérarchie est organisée selon les dimensions d'un système. La démarche est illustrée par un cas pratique
Sur l'inégalité des chances : le cas de l'Inde
No full textThe thesis consists of three empirical chapters on the extent of inequality of opportunity in India using data from the National Sample Survey. Inequality of opportunity is conceptualized as the unequal distribution in an outcome that is generated by factors beyond any individual responsibility. Chapter 1 estimates the index measures of inequality of opportunity in India during the time frame of 2004-12, for consumption expenditure, wage earning and education. Taking caste, sex, region of residence, parental education and occupation as our fatalistic circumstances, both the non-parametric and parametric index shows that more than one-fourth of the respective inequality in wage and education is due to unequal opportunities among the circumstances. Since the traditional approach of non-parametric and parametric estimation strategy can not point out the meaningful intertwining among the circumstances, we further adopt the recently introduced approach of the regression tree analysis that is able to draw an opportunity tree by hierarchical order of circumstances. The opportunity tree for contemporary India identifies parental education as one of the main circumstance variable behind the unequal opportunity in the country, especially for education. Chapter 2 focuses on a temporal perspective that explores equalization of opportunities among the different caste categories in India over the time span of 1983-2012, adopting the robust methodology of distributional dominance. The relatively upper caste groups are not only found to clearly dominate the historically disadvantageous caste categories, but the premium enjoyed by the former actually increases over time, as far as earning opportunity is concerned. Opportunity in consumption expenditure among the caste groups however have equalized to an impressive extent, especially during 2004-12. Chapter 3 estimates the degree of inequality of opportunity in access to elementary and post-elementary education among Indian children, using the dissimilarity index and the Human Opportunity Index. Indian children are found to enroll in elementary schools on time, irrespective of their circumstances. However children of lesser educated parents are always less likely to complete the eight years of elementary schooling on time, even in 2012, when elementary education is made free and compulsory by the Government, for all eligible children. The regional analysis reveals that access to basic educational opportunity is relatively worse for East and Central India.La thèse se compose de trois chapitres empiriques sur l'ampleur de l'inégalité des chances en Inde à l'aide des données de la National Sample Survey. L'inégalité des chances est conceptualisée comme une distribution inégale dans un résultat qui est généré par des facteurs qui dépassent toute responsabilité individuelle. Le chapitre 1 donne une estimation des indexées de l'inégalité des chances en Inde au cours de la période du 2004-12, pour les dépenses de consommation, les salaires et l'éducation. Si nous considérons la caste, le sexe, la région de résidence, l'éducation parentale et le métier comme circonstances fatalistes, l'indice non paramétrique et l'indice paramétrique montrent que plus du quart de l'inégalité respective en matière de salaire et d'éducation est due à l'inégalité des chances. L'approche traditionnelle de la stratégie d'estimation paramétrique et paramétrique ne permettant pas de mettre en évidence une imbrication significative des circonstances, nous adoptons en outre l'approche la plus récente de l'analyse par arbre de régression, qui permet de tracer un arbre d'opportunité par ordre hiérarchique de circonstances. L'arbre des opportunités de l'Inde contemporaine identifié l'éducation parentale comme l'une des principales variables de l'inégalité des chances, en particulier dans le domaine de l'éducation. Le chapitre 2 se concentre sur une perspective temporelle qui explore l'égalisation des chances entre les différentes catégories de castes en Inde sur la période 1983-2012, en adoptant la méthodologie robuste de la dominance distributionnelle. Les groupes de castes relativement supérieurs non seulement dominent clairement les catégories de castes historiquement désavantageuses, mais la prime dont bénéficient les premiers augmente en fait avec le temps, en ce qui concerne les possibilités de revenus. Les opportunités de consommation des groupes de castes se sont toutefois égalisées de manière impressionnante, en particulier au cours de la 2004-12. Le chapitre 3 évalue le degré d'inégalité des chances d'accès aux études élémentaires et post-élémentaires chez les enfants indiens, à l'aide de l'indice de dissimilarité et de l'indice de chances humaines. On constate que les enfants indiens s'inscrivent à l'école primaire à temps, quelle que soit leur situation. Cependant, les enfants de parents moins scolarisés ont toujours moins de chances de terminer les huit années d'enseignement primaire à temps, même en 2012, année où l'enseignement primaire est rendu gratuit et obligatoire pour tous les enfants éligibles par le gouvernement. L'analyse régionale révèle que l'accès aux possibilités d'éducation de base est relativement moins bon pour l'Inde orientale et centrale
Comparing Lead Users to Emergent‐Nature Consumers as Sources of Innovation at Early Stages of New Product Development
No full textInternational audienceThe large potential of lead users (LUs) in developing innovative and radically new product concepts is well established in the literature. However, in a widely acknowledged study, Hoffman et al. introduced the new concept of emergent‐nature consumers (ENCs) and showed the superiority of this group of individuals over LUs in developing new product concepts. Consequently, they postulated ENCs to be "the right consumers" to be integrated in new product development processes. In this article, we critically reflect and build on Hoffman et al.'s study and further investigate the promise of the ENC concept as compared to the LU concept. In a pilot study, we replicated the study by Hoffman and colleagues: We conducted a crowdsourcing competition and asked for concepts for new services; those concepts had been generated collectively by the participants and had been assessed by a consumer crowd. In the main study, the participants of a crowdsourcing competition submitted individually generated concepts for products, which were evaluated by industry experts. Across both studies, using different empirical methods in two different contexts, and in contrast to Hoffman et al.'s work, we find support for LUs outperforming ENCs (as well as average users) in generating the commercially most promising concepts. Thus, our insights reinforce the existing user innovation literature and the notion of LUs being the primary source of new product/service concepts
Pledges as a Social Influence Device: Experimental Evidence
No full textThis paper reports the results from a two-person "pledge and give" experiment. Each persons endowment is private information available only to him. In the first stage, each agent informs the other about the amount he intends to give, or makes a pledge. In the second stage, each agent makes a contribution to the joint donation. A simple theoretical model shows that in this game the equilibrium pledge function is linear in the endowment of each agent. Furthermore, if agents have a strong taste for conformity, the optimal gift is positively related to one’s own endowment and to the pledge of his partner. Data from the lab experiment show that, indeed, subjects pledge approximately 60% of their endowment. Also, pledges have an important social influence role: an agent will increase his donation by 20 cents on average if his partner pledges one more euro
An exact algorithm for the inventory routing problem with logistic ratio
No full textInternational audienc
Community energy and the virtues of mistrust and distrust: Lessons from Brighton and Hove energy cooperatives
No full textInternational audienc
Breakthrough recognition: Bias against novelty and competition for attention
No full textInternational audienc
Optimal graphon estimation in cut distance
No full textInternational audienceConsider the twin problems of estimating the connection probability matrix of an inhomogeneous random graph and the graphon of a W-random graph. We establish the minimax estimation rates with respect to the cut metric for classes of block constant matrices and step function graphons. Surprisingly, our results imply that, from the minimax point of view, the raw data, that is, the adjacency matrix of the observed graph, is already optimal and more involved procedures cannot improve the convergence rates for this metric. This phenomenon contrasts with optimal rates of convergence with respect to other classical distances for graphons such as the l 1 or l 2 metrics
