Daniel Defays

Université de Liège, Psychology, Professeur de Faculté

Followers

Following

Public Views

Daniel Defays est docteur en sciences et, depuis 1991, professeur de Faculté à l’université de Liège où il donne un cours d’introduction à l’IA à des étudiants en sciences humaines. Il s’intéresse à l’Intelligence Artificielle depuis plus de 40 ans et a déjà publié un premier livre sur le sujet en 1987 : « L’esprit en friche : les foisonnements de l’intelligence artificielle », après un séjour d’un an passé dans un laboratoire d’IA aux Etats-Unis. Depuis, il a écrit des nombreux articles scientifiques dans divers ouvrages et revues, et un livre en 2020 sur les rapports qui existent entre la musique, les images et les saveurs. Sa dernière publication, parue en avril 2023 dans « Cognitive Science », traite de la perception de la musique au moyen d’un réseau de neurones artificiels.

less

Sebastian Trump

Hochschule für Musik Nürnberg

Lancaster University

University of California, San Diego

International Journal of Scientific Research in Computer Science, Engineering and Information Technology IJSRCSEIT

Techno Scienceacademy

Catherine Reed

Jaeyong Kang

Gwangju Institute of Science and Technology

Sebastian Martinez valencia

Interests

Uploads

Papers by Daniel Defays

From melodic note sequences to pitches using word2vec

arXiv, 2024

Applying the word2vec technique, commonly used in language modeling, to melodies, where notes are... more Applying the word2vec technique, commonly used in language modeling, to melodies, where notes are treated as words in sentences, enables the capture of pitch information. This study examines two datasets: 20 children's songs and an excerpt from a Bach sonata. The semantic space for defining the embeddings is of very small dimension, specifically 2. Notes are predicted based on the 2, 3 or 4 preceding notes that establish the context. A multivariate analysis of the results shows that the semantic vectors representing the notes have a multiple correlation coefficient of approximately 0.80 with their pitches.

Download

Contribution à l'étude du potentiel évoqué moyen visuel (PEMV); caractère non-gaussien de l'activité électroencéphalographique spontanée (AS); non-indépendance du bruit (AS) et du signal (PE); non-stationnarité éventuelle du PE

Acta Neurologica Belgica, 1978

peer reviewe

Reliability of Mean Length of Utterance as a Function of Sample Size in Early Language Development

Journal of Genetic Psychology, Dec 1, 1978

Numbo: a study in cognition and recognition

Basic Books, Inc, Feb 28, 1995

L’intelligence artificielle et ses métaphores

A short note on a method of seriation

British Journal of Mathematical and Statistical Psychology, May 1, 1978

The classical problem of the sequencing of objects along a continuum may be solved in a rather el... more

An efficient algorithm for a complete link method

The Computer Journal, Apr 1, 1977

Download

A comparison of three linear rules for classification

Applied Stochastic Models and Data Analysis, 1986

This paper compares three different linear procedures for classification: the normal one, the can... more This paper compares three different linear procedures for classification: the normal one, the canonical one and a distribution‐free one recently described by Heuchenne. The study is mainly conducted using a simulation which makes it possible to compute the probabilities of correct allocation of the three methods in 3888 different cases. The normal rule looks slightly better than Heuchenne's, which looks clearly better than the canonical one. Finally, inference on Heuchenne's method is examined and conditions under which this method is optimal are given.

Appariement de matrices de dissimilarités

Monde des Util. Anal. Données, 2018

Dans cet article, nous comparons différentes méthodes qui permettent de trouver dans un ensemble ... more Dans cet article, nous comparons différentes méthodes qui permettent de trouver dans un ensemble fini S, sur lequel on a défini une dissimilarité, un sous-ensemble d'éléments qui s'apparient au mieux aux éléments d'un ensemble cible C également muni d'une dissimilarité. La qualité de l'appariement est mesurée par la distance observée entre les matrices de dissimilarités définies sur les sous-ensembles de S et sur C. Différents algorithmes sont présentés et les performances sont commentées. Mots clefs. Distances-Espaces métriques-Appariement-Isométrie-Reconnaissance de formes Summary. In this paper, we compare different methods to map in a optimal way a subset of a finite set Son which a dissimilarity is defined-on a target set C, on which another dissimilarity is defined. The quality of the mapping is assessed by comparing the dissimilarities of the elements of the target C with the corresponding dissimilarities in S. The closer they are, the better the fit is. Different algorithms are presented and performances are compared.

Download

Exploration of Trans-Sensory Mappings

Art & perception, Mar 9, 2017

This paper explores the way music can be translated into a sequence of images. A general approach... more This paper explores the way music can be translated into a sequence of images. A general approach is described that associates various structures captured by different sense modalities. This approach is illustrated by examples from vision and audition. Music and sets of images are digitalized and represented as forms in multidimensional spaces. A mapping, which preserves the dissimilarities in each space, is then defined and an algorithm that can approximate the optimal fit is proposed. An example that maps a set of photos of different versions of the wire chair on various Mozart variations is given.

Fluid Concepts and Creative Analogies: A Review

Download

Statistics and Artificial Intelligence

Encyclopedia of Statistical Sciences, Aug 15, 2006

Tree representations of ternary relations

Journal of Mathematical Psychology, Apr 1, 1979

... and sufficient conditions on the dissimilarity index for different types of representations h... more

A Recurrent Connectionist Model of Melody Perception: An Exploration Using TRACX2

Cognitive Science, Apr 1, 2023

Are similar, or even identical, mechanisms used in the computational modeling of speech segmentat... more Are similar, or even identical, mechanisms used in the computational modeling of speech segmentation, serial image processing, and music processing? We address this question by exploring how TRACX2, a recognition‐based, recursive connectionist autoencoder model of chunking and sequence segmentation, which has successfully simulated speech and serial‐image processing, might be applied to elementary melody perception. The model, a three‐layer autoencoder that recognizes “chunks” of short sequences of intervals that have been frequently encountered on input, is trained on the tone intervals of melodically simple French children's songs. It dynamically incorporates the internal representations of these chunks into new input. Its internal representations cluster in a manner that is consistent with “human‐recognizable” melodic categories. TRACX2 is sensitive to both contour and proximity information in the musical chunks that it encounters in its input. It shows the “end‐of‐word” superiority effect demonstrated by Saffran et al. (1999) for short musical phrases. The overall findings suggest that the recursive autoassociative chunking mechanism, as implemented in TRACX2, may be a general segmentation and chunking mechanism, underlying not only word‐ and image‐chunking, but also elementary melody processing.

Apport de l'intelligence artificielle à la psychologie

A roof rack comprises a pair of interconnected side rails overlying and at their ends connected t... more A roof rack comprises a pair of interconnected side rails overlying and at their ends connected to a vehicle roof, and elongated caps of a flexible material overlying and enclosing a portion of each rail and interlocked therewith. Each rail is an elongated extrusion including a pair of parallel top bends along its length with spaced strips depending from the beads defining undercut corners. Out-turned flanges extend from the strips and terminate in upturned marginal ledges. The cap has a top wall that overlies the beads and a pair of side walls terminating in a pair of opposed internal supports respectively overlying the out-turned flanges and retained and interposed between the strips and marginal ledges. A pair of opposed internal anchor flanges on the side walls interlockingly engage the shoulders to secure then cap along its length to the rail.

Les Représentations en Arbres: Présentation, Critiques et Suggestions

Psychologica Belgica, 1979

A Recurrent Connectionist Model of Melody Perception: An Exploration Using TRACX2

Cognitive Science

[Intensity of spontaneous activity and mean visual evoked response during calculation of variable mean. New technique for measuring retinocortical time (author's transl)]

Acta neurologica Belgica

Various statistical discrimination indices reveal deflections regarded as significant before the ... more Various statistical discrimination indices reveal deflections regarded as significant before the 20th ms after stimulation (probably equivalent to Bickford responses or subcortical activity) and between the 20th and 23rd ms (early fractions of MVER). The independence of signal and noise is discussed. Considerations concerning the calculation of retinocortical time are developed.

Mapping of dissimilarity matrices