Episciences.org
Not a member yet
6707 research outputs found
Sort by
Heterodoxy Needs Institutional Backing
No full textA general aversion to new ideas, psychological factors, and foremost, institutional conditions shape the challenging position of heterodox economics. This institutional framework is coined by a strong orientation towards publication metrics and influences young scholars to conformity. We propose two ideas to improve the conditions for heterodox research. First, to introduce competition between journals for the scientific papers they want to have the most. Second, to establish a qualified random selection of papers to equalize the chances of publishing
Espaces odorants et espaces olfactifs
No full textAttester le partage d'une expérience olfactive par plusieurs individus est une gageure, du fait d'obstacles théoriques et méthodologiques souvent présentés comme irréductibles. Après une brève discussion de la réalité de cette irréductibilité, nous essayons de surmonter certains de ces obstacles en distinguant espaces odorants et espaces olfactifs. Si un espace odorant peut-être objectivé et donc partagé, il n'en va pas de même d'un espace olfactif qui relève d'une expérience subjective. Cependant, l'effet invasif des molécules odorantes, plus spécifiquement celles qui provoquent des « mauvaises odeurs », est de nature à atténuer cette subjectivité et, du même coup, à faciliter le partage d'un espace olfactif. Notre argumentation prend appui sur des enquêtes ethnographiques menées au Brésil, en Chine et en Inde et sur une recherche menée en géographie sur la spatialisation des nuisances olfactives
A denotationally-based program logic for higher-order store
No full textSeparation logic is used to reason locally about stateful programs. State ofthe art program logics for higher-order store are usually built on top ofuntyped operational semantics, in part because traditional denotational methodshave struggled to simultaneously account for general references and parametricpolymorphism. The recent discovery of simple denotational semantics for generalreferences and polymorphism in synthetic guarded domain theory has enabled usto develop TULIP, a higher-order separation logic over the typed equationaltheory of higher-order store for a monadic version of System F{mu,ref}. TheTulip logic differs from operationally-based program logics in two ways:predicates range over the meanings of typed terms rather than over the raw codeof untyped terms, and they are automatically invariant under the equationalcongruence of higher-order store, which applies even underneath a binder. As aresult, "pure" proof steps that conventionally require focusing the Hoaretriple on an operational redex are replaced by a simple equational rewrite inTulip. We have evaluated Tulip against standard examples involving linked listsin the heap, comparing our abstract equational reasoning with more familiaroperational-style reasoning. Our main result is the soundness of Tulip, whichwe establish by constructing a BI-hyperdoctrine over the denotational semanticsof F{mu,ref} in an impredicative version of synthetic guarded domain theory.Comment: To appear in proceedings of MFPS 202
Antisquares and Critical Exponents
No full textThe (bitwise) complement of a binary word is obtained bychanging each in to and vice versa. An is anonempty word of the form . In this paper, we study infinitebinary words that do not contain arbitrarily large antisquares. For example, weshow that the repetition threshold for the language of infinite binary wordscontaining exactly two distinct antisquares is . We also studyrepetition thresholds for related classes, where "two" in the previous sentenceis replaced by a larger number. We say a binary word is if the only antisquares it containsare and . We characterize the minimal antisquares, that is, thosewords that are antisquares but all proper factors are good. We determine thegrowth rate of the number of good words of length and determine therepetition threshold between polynomial and exponential growth for the numberof good words
Le projet OPTIMICE : une optimisation de la qualité des traductions de métadonnées par la collaboration entre acteurs du monde scientifique et traduction automatique
No full textThe OPTIMICE project, which stands for optimising machine translation of metadata and its integration into the editorial chain, aims at devising a method – transferrable to other journals and disciplinary fields – that combines neural machine translation (DeepL) and human post-editing to improve the quality of article metadata (abstracts, keywords, etc.) from French to English in the editorial process of journals. A team of translation researchers who are also translators worked on four journals edited by the Presses universitaires de Rennes (PUR), in collaboration with the editorial comittees and the MSHB (Maison des sciences de l’Homme en Bretagne). The translation of the paper metadata by their authors and by machine translation was comparatively assessed. A survey on translation practices among researchers in HSS was led, and recommendations for writing and translating metadata were formulated for the organized integration of the methodology within the editorial process.Le projet OPTIMICE, pour optimisation de la traduction automatique des métadonnées et de son insertion dans la chaîne éditoriale, vise à concevoir une méthode, transférable à d'autres revues et domaines disciplinaires, en combinant la traduction automatique neuronale (DeepL) et la post-édition humaine pour améliorer la qualité des métadonnées des articles (résumés, mots-clés, etc.) du français vers l'anglais dans le processus éditorial des revues. Une équipe de traducteurs-traductologues a travaillé sur quatre revues éditées par les Presses universitaires de Rennes (PUR), en collaboration avec les comités éditoriaux et la MSHB (Maison des sciences de l'Homme en Bretagne). La traduction des métadonnées d'articles par leurs auteurs et par la traduction automatique a été évaluée comparativement. Une enquête sur les pratiques de traduction des chercheurs en SHS a été menée, et des recommandations de rédaction et de traduction des métadonnées ont été formulées pour l'insertion concertée de la méthodologie au sein du processus éditorial. Mots-clés TAN (traduction automatique neuronale); post-édition; SHS; métadonnées; revue
Boosting Simple Learners
No full textBoosting is a celebrated machine learning approach which is based on the ideaof combining weak and moderately inaccurate hypotheses to a strong and accurateone. We study boosting under the assumption that the weak hypotheses belong toa class of bounded capacity. This assumption is inspired by the commonconvention that weak hypotheses are "rules-of-thumbs" from an "easy-to-learnclass". (Schapire and Freund~'12, Shalev-Shwartz and Ben-David '14.) Formally,we assume the class of weak hypotheses has a bounded VC dimension. We focus ontwo main questions: (i) Oracle Complexity: How many weak hypotheses are neededto produce an accurate hypothesis? We design a novel boosting algorithm anddemonstrate that it circumvents a classical lower bound by Freund and Schapire('95, '12). Whereas the lower bound shows that weakhypotheses with -margin are sometimes necessary, our new methodrequires only weak hypothesis, provided that theybelong to a class of bounded VC dimension. Unlike previous boosting algorithmswhich aggregate the weak hypotheses by majority votes, the new boostingalgorithm uses more complex ("deeper") aggregation rules. We complement thisresult by showing that complex aggregation rules are in fact necessary tocircumvent the aforementioned lower bound. (ii) Expressivity: Which tasks canbe learned by boosting weak hypotheses from a bounded VC class? Can complexconcepts that are "far away" from the class be learned? Towards answering thefirst question we {introduce combinatorial-geometric parameters which captureexpressivity in boosting.} As a corollary we provide an affirmative answer tothe second question for well-studied classes, including half-spaces anddecision stumps. Along the way, we establish and exploit connections withDiscrepancy Theory.Comment: Journal versio
On the classification of sub-Riemannian structures on a 5D two-step nilpotent Lie group
No full textWe classify the left-invariant sub-Riemannian structures on the unique five-dimensional simply connected two-step nilpotent Lie group with two-dimensional commutator subgroup; this 5D group is the first twostep nilpotent Lie group beyond the three-and five-dimensional Heisenberg groups. Alongside, we also present a classification, up to automorphism, of the subspaces of the associated Lie algebra (together with a complete set of invariants)
Parsing as a lifting problem and the Chomsky-Sch\"utzenberger representation theorem
No full textWe begin by explaining how any context-free grammar encodes a functor ofoperads from a freely generated operad into a certain "operad of splicedwords". This motivates a more general notion of CFG over any category ,defined as a finite species equipped with a color denoting the start symboland a functor of operads into the operad of splicedarrows in . We show that many standard properties of CFGs can be formulatedwithin this framework, and that usual closure properties of CF languagesgeneralize to CF languages of arrows. We also discuss a dual fibrationalperspective on the functor via the notion of "displayed" operad,corresponding to a lax functor of operads . We then turn to the Chomsky-Sch\"utzenberger Representation Theorem. Wedescribe how a non-deterministic finite state automaton can be seen as acategory equipped with a pair of objects denoting initial and acceptingstates and a functor of categories satisfying the unique lifting offactorizations property and the finite fiber property. Then, we explain how toextend this notion of automaton to functors of operads, which generalize treeautomata, allowing us to lift an automaton over a category to an automaton overits operad of spliced arrows. We show that every CFG over a category can bepulled back along a ND finite state automaton over the same category, and hencethat CF languages are closed under intersection with regular languages. Thelast important ingredient is the identification of a left adjoint to the operad of spliced arrows functor, building the "contourcategory" of an operad. Using this, we generalize the C-S representationtheorem, proving that any context-free language of arrows over a category is the functorial image of the intersection of a -chromatic tree contourlanguage and a regular language.Comment: reformatted for publication in ENTICS, proceedings of MFPS 202
Classifying topoi in synthetic guarded domain theory
No full textSeveral different topoi have played an important role in the development andapplications of synthetic guarded domain theory (SGDT), a new kind of syntheticdomain theory that abstracts the concept of guarded recursion frequentlyemployed in the semantics of programming languages. In order to unify theaccounts of guarded recursion and coinduction, several authors have enrichedSGDT with multiple "clocks" parameterizing different time-streams, leading tomore complex and difficult to understand topos models. Until now these topoihave been understood very concretely qua categories of presheaves, and thelogico-geometrical question of what theories these topoi classify has remainedopen. We show that several important topos models of SGDT classify very simplegeometric theories, and that the passage to various forms of multi-clockguarded recursion can be rephrased more compositionally in terms of the lowerbagtopos construction of Vickers and variations thereon due to Johnstone. Wecontribute to the consolidation of SGDT by isolating the universal property ofmulti-clock guarded recursion as a modular construction that applies to anytopos model of single-clock guarded recursion.Comment: 38th International Conference on Mathematical Foundations of Programming Semantics (MFPS 2022
Searching for carriers of the diffuse interstellar bands across disciplines, using Natural Language Processing
No full textThe explosion of scientific publications overloads researchers with information. This is even more dramatic for interdisciplinary studies, where several fields need to be explored. A tool to help researchers overcome this is Natural Language Processing (NLP): a machine-learning (ML) technique that allows scientists to automatically synthesize information from many articles. As a practical example, we have used NLP to conduct an interdisciplinary search for compounds that could be carriers for Diffuse Interstellar Bands (DIBs), a long-standing open question in astrophysics. We have trained a NLP model on a corpus of 1.5 million cross-domain articles in open access, and fine-tuned this model with a corpus of astrophysical publications about DIBs. Our analysis points us toward several molecules, studied primarily in biology, having transitions at the wavelengths of several DIBs and composed of abundant interstellar atoms. Several of these molecules contain chromophores, small molecular groups responsible for the molecule's colour, could be promising candidate carriers. Identifying viable carriers demonstrates the value of using NLP to tackle open scientific questions, in an interdisciplinary manner