Accueil > Pages Perso > Frédéric Lavigne

Frédéric Lavigne

PR -  UNS

My research interests are in the areas of prediction and language comprehension. They focus on the dynamics of semantic and combinatory learning and processing in memory (experimental approach) and the underlying synaptic and neuronal mechanisms in the cerebral cortex (modeling approach).

CV

Frédéric Lavigne

Researcher/Professor in Cognitive Science.

Laboratoire BCL (Bases, corpus, Langage) - UMR 7320 CNRS - Université de Nice - Sophia Antipolis, France. Equipe Langage et Cognition

Research Background :

Post Doc positions :

  • Laboratoire de Neurophysiologie de la Perception et de l’Action - CNRS - Collège de France, Paris, France.
  • Laboratorium voor Experimentele Psychologie - Katholieke Universiteit, Leuven, Belgium.

Ph.D. in Cognitive Sciences : Laboratoire Language, Cognition, Ergonomie. CNRS - Ecole Normale Supérieure, Paris, France.

Advanced Master Degree in Cognitive Sciences : Université Pierre et Marie Curie and EHESS, Paris, France.

Master Degree in Neuroscience and Psychophysiology : Université Pierre et Marie Curie, Paris, France.

Dernières publicationsHAL

pour l'idHal "frederic-lavigne" :

titre
Neuronal mechanisms for sequential activation of memory items: dynamics and reliability
auteur
Elif Köksal-Ersöz, Carlos Aguilar, Pascal Chossat, Martin Krupa, Frédéric Lavigne
article
2019
annee_publi
2019
typdoc
Pré-publication, Document de travail
Accès au bibtex
https://arxiv.org/pdf/1904.12133 BibTex
titre
Organization in Working Memory is Driven by the Compressibility of Information
auteur
Laura Lazartigues, Frédéric Lavigne, Carlos Aguilar, Fabien Mathy
article
60th Annual Meeting of the Psychonomic Society, Nov 2019, Montréal, Canada
annee_publi
2019
resume
Working memory (WM) is known to be limited in capacity, but mechanisms based on compression of information could contribute to the storage process. It has been shown that chunking governed by compression could be one of these mechanisms. The present study investigated how chunks can be formed using patterns to be discovered on the spot, that is, without these chunks being already formed in long-term memory. We predicted that a compact representation could leave room in WM at the expense of the quality. Our method was based on a compressibility metric that allowed prediction of memory errors linked to a lossy compression process. Our result showed that RTs and errors depended on compressibility, and those measures might be interpretable in terms of over-compressibility. We discuss the results to conclude that the present study offers a comprehension of WM capacity which cannot be accounted easily by shared-allocation models or discrete-slots models.
typdoc
Poster
Accès au bibtex
BibTex
titre
Statistical learning of adjacent and non-adjacent pairs in non-linguistic short sequences
auteur
Laura Lazartigues, Fabien Mathy, Arnaud Rey, Joël Fagot, Frédéric Lavigne
article
21st Conference of the European Society for Cognitive Psychology (ESCOP), Sep 2019, Tenerife, Spain
annee_publi
2019
resume
The ability to learn adjacent and non-adjacent pairs is central in language processing. However, current evidence indicates that adjacent and non-adjacent pairs are not equally learnable. The present study investigated the role of transitional probabilities during the learning of adjacent and non-adjacent pairs appearing in non-linguistic short sequences. Participants were exposed to four sequences of three stimuli ABC repeated randomly during the experiment, with each stimulus corresponding to a given position of a dot on a touchscreen. In the first experiment the transition BC of the triplet ABC was predictable while the first transition AB was unpredictable. The second experiment required the learning of the fully predictable non-adjacent pair AC while the transitions AB and BC were unpredictable. The results showed that participants learned adjacent pairs and had greater difficulty to learn the non-adjacent pairs. These data provide additional constraints for modeling statistical learning mechanisms.
typdoc
Poster
Accès au bibtex
BibTex
titre
Statistical learning of first and second order transitional probabilities
auteur
Laura Lazartigues, Fabien Mathy, Arnaud Rey, Joël Fagot, Frédéric Lavigne
article
Interdisciplinary Advances in Statistical Learning, Jun 2019, San Sebastian, Spain
annee_publi
2019
resume
The present study manipulates first and second order transitional probabilities during the statistical learning of short sequences. Participants were exposed to four sequences of three stimuli (ABC) repeated during the task, with each stimulus corresponding to the position of a red dot on a touchscreen. Participants were required to touch the dots as quickly as possible and response times were recorded between the first two stimuli (Transition Time 1 or TT1) and between the last two stimuli (TT2). In the first experiment the transition AB of a triplet ABC was fully predictable (p(B|A) = 1) while the second transition BC was unpredictable (p(C|B) = .5). The second experiment was a serial version of the exclusive-or (XOR), all first order transitional probabilities were equally unpredictable (p(B|A) = .5, p(C|B) = .5), while the combination of the first two stimuli fully predicted the last stimulus (p(C|AB) = 1). Results showed that participants were able to learn both type of transitional probabilities. The different evolution patterns of TT1 and TT2 and their implications in term of statistical learning mechanisms are discussed.
typdoc
Poster
Accès au bibtex
BibTex
titre
Linguistic processes do not beat visuo-motor constraints, but they modulate where the eyes move regardless of word boundaries: Evidence against word-based eye-movement control during reading.
auteur
Claire Albrengues, Frédéric Lavigne, Carlos Aguilar, Eric Castet, Françoise Vitu
article
2019
annee_publi
2019
typdoc
Pré-publication, Document de travail
Accès au texte intégral et bibtex
https://hal.archives-ouvertes.fr/hal-02081090/file/Albrengues-et-al_Submitted-to-PlosOne.pdf BibTex
titre
Linguistic processes do not beat visuo-motor constraints, but they modulate where the eyes move regardless of word boundaries: Evidence against top-down word-based eye-movement control during reading
auteur
Claire Albrengues, Frédéric Lavigne, Carlos Aguilar, Eric Castet, Françoise Vitu
article
PLoS ONE, Public Library of Science, 2019, 14 (7), pp.e0219666. ⟨10.1371/journal.pone.0219666⟩
annee_publi
2019
resume
Where readers move their eyes, while proceeding forward along lines of text, has long been assumed to be determined in a top-down word-based manner. According to this classical view, readers of alphabetic languages would invariably program their saccades towards the center of peripheral target words, as selected based on the (expected) needs of ongoing (word-identification) processing, and the variability in within-word landing positions would exclusively result from systematic and random errors. Here we put this predominant hypothesis to a strong test by estimating the respective influences of language-related variables (word frequency and word predictability) and lower-level visuo-motor factors (word length and saccadic launch-site distance to the beginning of words) on both word-skipping likelihood and within-word landing positions. Our eye-movement data were collected while forty participants read 316 pairs of sentences, that differed only by one word, the prime; this was either semantically related or unrelated to a following test word of variable frequency and length. We found that low-level visuo-motor variables largely predominated in determining which word would be fixated next, and where in a word the eye would land. In comparison, language-related variables only had tiny influences. Yet, linguistic variables affected both the likelihood of word skipping and within-word initial landing positions, all depending on the words’ length and how far on average the eye landed from the word boundaries, but pending the word could benefit from peripheral preview. These findings provide a strong case against the predominant word-based account of eye-movement guidance during reading, by showing that saccades are primarily driven by low-level visuo-motor processes, regardless of word boundaries, while being overall subject to subtle, one-off, language-based modulations. Our results also suggest that overall distributions of saccades’ landing positions, instead of truncated within-word landing-site distributions, should be used for a better understanding of eye-movement guidance during reading.
typdoc
Article dans une revue
Accès au texte intégral et bibtex
https://hal.archives-ouvertes.fr/hal-02360828/file/journal.pone.0219666.pdf BibTex
titre
Explaining variation in wh-position in child French: A statistical analysis of new seminaturalistic data
auteur
Katerina Palasis, Richard Faure, Frédéric Lavigne
article
Language Acquisition, Taylor & Francis (Routledge), 2019, 26 (2), pp.210-234. ⟨10.1080/10489223.2018.1513004⟩
annee_publi
2019
resume
The two possible positions for wh-words (i.e., in situ or preposed) represent a long-standing area of research in French. The present study reports on statistical analyses of a new seminaturalistic corpus of child L1 French. The distribution of the wh-words is examined in relation to a new verb tripartition: Free be forms, the Fixed be form c’est ‘it is’, and Other Verbs. Results indicate that a discriminating variable is verb form (i.e., Free vs. Fixed), regardless of verb type (i.e., be vs. Other Verbs), and that there is a correlation between the wh-in-situ position and the Fixed be form. The Fixed be form is thus identified as the component that leads to wh-in-situ utterances, in contrast to other languages such as English. Overuse of the Fixed be form in child speech could also account for the predominance of wh-in-situ in child object questions compared to adjunct questions and child wh-questions in general compared to adult questions
typdoc
Article dans une revue
Accès au bibtex
BibTex
titre
Latching dynamics in neural networks with synaptic depression
auteur
Carlos Aguilar, Pascal Chossat, Maciej Krupa, Frédéric Lavigne
article
PLoS ONE, Public Library of Science, 2017, 12 (8), ⟨10.1371/journal.pone.0183710⟩
annee_publi
2017
resume
Prediction is the ability of the brain to quickly activate a target concept in response to arelated stimulus (prime). Experiments point to the existence of an overlap between the populationsof the neurons coding for different stimuli, and other experiments show that prime-targetrelations arise in the process of long term memory formation. The classical modellingparadigm is that long term memories correspond to stable steady states of a Hopfield networkwith Hebbian connectivity. Experiments show that short term synaptic depressionplays an important role in the processing of memories. This leads naturally to a computationalmodel of priming, called latching dynamics; a stable state (prime) can become unstableand the system may converge to another transiently stable steady state (target).Hopfield network models of latching dynamics have been studied by means of numericalsimulation, however the conditions for the existence of this dynamics have not been elucidated.In this work we use a combination of analytic and numerical approaches to confirmthat latching dynamics can exist in the context of a symmetric Hebbian learning rule, howeverlacks robustness and imposes a number of biologically unrealistic restrictions on themodel. In particular our work shows that the symmetry of the Hebbian rule is not an obstructionto the existence of latching dynamics, however fine tuning of the parameters of themodel is needed.
typdoc
Article dans une revue
Accès au bibtex
https://arxiv.org/pdf/1611.03645 BibTex
titre
Beyond transitional probabilities: learning XOR in non-human primates
auteur
Arnaud Rey, Frédéric Lavigne, Fabien Mathy, Joël Fagot
article
Fifth Implicit Learning Seminar, Lancaster University, Jun 2016, Lancaster, United Kingdom
annee_publi
2016
resume
Paper presented at the Fifth Implicit Learning Seminar, Lancaster, UK.
typdoc
Communication dans un congrès
Accès au bibtex
BibTex
titre
Machine Learning under the light of Phraseology expertise: use case of presidential speeches, De Gaulle -Hollande (1958-2016)
auteur
Mélanie Ducoffe, Damon Mayaffre, Frédéric Precioso, Frédéric Lavigne, Laurent Vanni, A Tre-Hardy
article
JADT 2016 - Statistical Analysis of Textual Data, Damon Mayaffre; Céline Poudat; Laurent Vanni; Véronique Magri; Peter Follette; Caroline Daire, Jun 2016, Nice, France. pp.157-168
annee_publi
2016
resume
Author identification and text genesis have always been a hot topic for the statistical analysis of textual data community. Recent advances in machine learning have seen the emergence of machines competing state-of-the-art computational linguistic methods on specific natural language processing tasks (part-of-speech tagging, chunking and parsing, etc). In particular, Deep Linguistic Architectures are based on the knowledge of language speci-ficities such as grammar or semantic structure. These models are considered as the most competitive thanks to their assumed ability to capture syntax. However if those methods have proven their efficiency, their underlying mechanisms, both from a theoretical and an empirical analysis point of view, remains hard both to explicit and to maintain stable, which restricts their area of applications. Our work is enlightening mechanisms involved in deep architectures when applied to Natural Language Processing (NLP) tasks. The Query-By-Dropout-Committee (QBDC) algorithm is an active learning technique we have designed for deep architectures: it selects iteratively the most relevant samples to be added to the training set so that the model is improved the most when built from the new training set. However in this article, we do not go into details of the QBDC algorithm-as it has already been studied in the original QBDC article-but we rather confront the relevance of the sentences chosen by our active strategy to state of the art phraseology techniques. We have thus conducted experiments on the presidential discourses from presidents C. De Gaulle, N. Sarkozy and F. Hollande in order to exhibit the interest of our active deep learning method in terms of discourse author identification and to analyze the extracted linguistic patterns by our artificial approach compared to standard phraseology techniques.
typdoc
Communication dans un congrès
Accès au texte intégral et bibtex
https://hal.archives-ouvertes.fr/hal-01343209/file/JADT2016_Ducoffe_et_al.pdf BibTex
  • + de résultats dans la Collection HAL du laboratoire BCL
  • Voir l'ensemble des résultats sur la plateforme HAL