Portail HAL EHESS (École des hautes études en sciences sociales)
Not a member yet
103682 research outputs found
Sort by
Two antagonistic objectives for one multi-scale graph clustering framework
No full textInternational audienceIn the current state of knowledge, there is no consensus on an objective criterion for evaluating network communities as cohesive sets of nodes with the following two properties:PDC : Each community is Densely Connected;PWC : Communities are Weakly Connected to each other.This makes it difficult to conduct comparative studies between dozens of graph clustering methods proposed over more than 20 years. To fill this gap:■ We propose a graph clustering framework by faithfully formalizing PDC with precision and PWC with recall, which are two meaningful metrics, simple, well known and already widely used for many tasks in most sciences. The meaning of these metrics in the context of graph clustering is therefore easily interpretable by most users of real-world graphs.■ We show that for most graphs, these two metrics are antagonistic, i.e. there is no solution that simultaneously maximizes precision and recall. In other words, to select a clustering among the Pareto optimal solutions (clusterings such that no other clustering exist that both increases the precision and the recall) we must first make a subjective compromise, according to our needs between the two properties PDC and PWC .■ We then show how to use this framework to compare, even without 'ground truth', the performances of five hitherto incommensurable state-of-the-art clustering methods, as well as that of a new family of clustering methods inspired by our approach.Dans l'état actuel des connaissances, il n'existe pas de consensus sur un critère objectif permettant d'évaluer les communautés de réseaux comme des ensembles cohérents de nœuds présentant les deux propriétés suivantes :PDC : Chaque communauté est densément connectée ;PWC : Les communautés sont faiblement connectées entre elles.Cela complique la réalisation d'études comparatives entre les dizaines de méthodes de clustering de graphes proposées depuis plus de 20 ans. Pour combler cette lacune :■ Nous proposons un cadre de clustering de graphes en formalisant fidèlement PDC avec précision et PWC avec rappel, deux métriques significatives, simples, bien connues et déjà largement utilisées pour de nombreuses tâches dans la plupart des sciences. La signification de ces métriques dans le contexte du clustering de graphes est donc facilement interprétable par la plupart des utilisateurs de graphes réels.■ Nous montrons que pour la plupart des graphes, ces deux métriques sont antagonistes, c'est-à-dire qu'il n'existe pas de solution maximisant simultanément la précision et le rappel. En d'autres termes, pour sélectionner un clustering parmi les solutions optimales de Pareto (clusterings tels qu'il n'existe aucun autre clustering qui augmente à la fois la précision et le rappel), nous devons d'abord faire un compromis subjectif, selon nos besoins, entre les deux propriétés PDC et PWC.■ Nous montrons ensuite comment utiliser ce cadre pour comparer, même sans « vérité terrain », les performances de cinq méthodes de clustering de pointe jusqu'ici incommensurables, ainsi que celles d'une nouvelle famille de méthodes de clustering inspirées de notre approche
Pro securitate et valida custodia patrie et subdictorum dalphinalium. Le compte de la construction de la bâtie de Pont (1374-1379)
No full textInternational audienc
Mood and the malleability of moral reasoning: the impact of irrelevant factors on judicial decisions
No full textNational audienceEmotions are said to underlie moral decision-making. We detect intra-judge variation spanning three decades in 1.5 million judicial decisions driven by factors unrelated to case merits. U.S. immigration judges grant an additional 1.4 % points of asylum petitions–and U.S. district judges assign 0.6 % points fewer prison sentences and 5 % longer probation sentences—on the day after their city's NFL team won, relative to days after the team lost. Bad weather has the opposite effect of a team win. Unrepresented parties in asylum bear the brunt of NFL effects. The effect on district judges only appears for judges born in the same state as the current state of residence, providing clean evidence of extraneous influences on judge decision-making as opposed to lawyer or applicant behavior.Moving beyond OLS, we utilize models from machine learning to estimate the sentence length relative to the sentencing guideline. We find that while several appropriate features predict sentence length, such as details of the crime committed, other features seemingly unrelated, including daily temperature, sport game scores, and location of trial, are predictive as well. The predictive power of the unrelated events is derived from the permutation based variable importance score in random forests. We address recent criticism of the reliability of these scores with double residualization
Tax avoidance and commodity tax differentiation
No full textNational audienceWe examine the optimal combination of direct and indirect taxes in the presence of tax avoidance. Our findings indicate that linear commodity taxes should be included in the optimal tax mix, even when they are subject to avoidance and when the conditions of the Atkinson-Stiglitz theorem hold. This is because taxing consumption—despite the possibility of avoidance—enhances the ability to screen true income, whereas income taxation alone depends solely on reported income. Additionally, we show that when utility is weakly separable, tax rates should be positive and uniform across goods if the subutility function is homothetic, leading to linear Engel curves. However, when Engel curves are nonlinear, commodity taxes need not be uniform. Furthermore, the optimal taxation of luxuries versus necessities depends on the distribution of productivity levels
Ordonner, raconter, remémorer. Quelques remarques sur les chapiteaux de la crypte Saint-Girons d’Hagetmau (Landes)
No full textInternational audienc
Tropical Nature. Colonial and Post-Colonial Conservation in Africa and Asia
No full textInternational audienceAcross Africa and South-East Asia, the impulse to protect nature often dovetails with the domination of local people. From mass displacement to severe restrictions on land use and daily acts of violence, conservation work risks reproducing Eurocentric modes of colonialism and worsening the effects of the climate crisis. In this insightful and wide-ranging study of the colonial history of conservation, Tropical Nature seeks to provide a much-needed history of the Global South from its own perspective. Comparing case studies ranging from Ali Bongo’s Gabon, to the postcolonial African itinerary of the agronomist Arthur Bunting, this volume advances a “small-scale global history” that deciphers the relations binding human societies to the non-human world
Equilibrium with non-convex preferences: some insights
No full textWe study the existence of equilibrium when agents’ preferences may not beconvex. For some specific utility functions, we provide a necessary and sufficientcondition under which there exists an equilibrium. The standard approach cannot be directly applied to our examples because the demand correspondence of some agents is neither single-valued nor convex-valued
Beyond Present Bias: Exploring Temporal Smoothing Biases
No full textThe importance of the literature on present bias may have overlooked some other forms of temporal inconsistencies anchored in the smoothing properties of consumption. The article first clarifies how a list of axioms enable the obtention of time-dependent recursive utility functions. The properties of the latters are analysed in light of the well-known present bias but also through a temporal smoothing bias that emerges when two successives selves do no share the same aversion to fluctuations. It is shown that this new concept relates to a time-varying Morishima intertemporal elasticity of substitution. The second part of the article brings an axiomatic construction providing a parametric representation that is aimed at a careful account of future and present bias, how they relate to each other or to substitution mechanisms and is finally concerned with the testable implications of the current framework. The theory is finally applied by considering intertemporal choices with Markovian strategies and temporally consistent solutions. For some parameters configurations, there exists a multiplicity of Nash equilibria, and then an indeterminacy in the agent behaviour