Papers by Guenael Cabanes
Apprentissage non-supervisé relationnel dans l'espace des coordonnées barycentriques
EGC eBooks, 2019
Détection de changement dans les profils en ligne d'utilisateurs
EGC eBooks, 2019
HAL (Le Centre pour la Communication Scientifique Directe), Jul 8, 2014
In general, the change in the local strain field or global stiffness caused by damage in a struct... more In general, the change in the local strain field or global stiffness caused by damage in a structure is very small and the strain field tends to homogenize very quickly in the field close to the defect. Moreover, other environmental effects can fade the slight changes in the strain field. Only by comparing the response of the structure at several points some information about damage may be unveiled. By means of pattern recognition techniques based on the strain field, this task can be achieved. This is the basis of the strain measurements data-driven models. The main limitation of the strain field pattern recognition techniques lies in the susceptibility of the strain field to change depending on the load conditions. In the case of dynamic loads, this may reflect even a greater limitation. Robust automated techniques are required to manage these limitations. In first instance, automatic clustering techniques are needed so that data can be classified according to the load conditions and secondly, a dimensional reduction technique is needed in order to obtain patterns that often underlie from data. Within the context of this paper, a combination of Local Density-based Simultaneous Two-Level (DS2L-SOM) Clustering based on Self-Organizing Maps (SOM) and Principal Components Analysis (PCA) is proposed in order to firstly, classify load conditions and secondly, perform strain field pattern recognition. The clustering technique is the basis for an Optimal Baseline Selection. An experimental validation of the technique is discussed in this paper, comparing damages of different sizes and positions in an aluminum beam, under a set of combined loads under dynamic conditions. Strains were measured at several points by using Fiber Bragg Gratings.
A New Clustering Algorithm for Dynamic Data
Lecture Notes in Computer Science, 2016
In this paper, we propose an algorithm for the discovery and the monitoring of clusters in dynami... more In this paper, we propose an algorithm for the discovery and the monitoring of clusters in dynamic datasets. The proposed method is based on a Growing Neural Gas and learns simultaneously the prototypes and their segmentation using and estimation of the local density of data to detect the boundaries between clusters. The quality of our algorithm is evaluated on a set of artificial datasets presenting a set of static and dynamic cluster structures.
HAL (Le Centre pour la Communication Scientifique Directe), Jul 8, 2014
The global trends in the construction of modern structures require the integration of sensors tog... more The global trends in the construction of modern structures require the integration of sensors together with data recording and analysis modules so that their integrity can be continuously monitored for safe-life, economic and ecological reasons. This process of measuring and analysing the data from a distributed sensor network all over a structural system in order to quantify its condition is known as structural health monitoring (SHM). Guided ultrasonic wave-based techniques are increasingly being adapted and used in several SHM systems which benefit from built-in transduction, large inspection ranges, and high sensitivity to small flaws. However, for reliable health monitoring, much information regarding the innate characteristics of the sources and their propagation is essential. Moreover, any SHM system which is expected to transition to field operation must take into account the influence of environmental and operational changes which cause modifications in the stiffness and damping of the structure and consequently modify its dynamic behaviour. On that account, special attention is paid in this paper to the development of an efficient SHM methodology where robust signal processing and pattern recognition techniques are integrated for the correct interpretation of complex ultrasonic waves within the context of damage detection and identification. The methodology is based on an acousto-ultrasonics technique where the discrete wavelet transform is evaluated for feature extraction and selection, linear principal component analysis for data-driven modelling and selforganizing maps for a two-level clustering under the principle of local density. At the end, the methodology is experimentally demonstrated and results show that all the damages were detectable and identifiable.
Multi-view Clustering using Barycentric Coordinate Representation
2023 International Joint Conference on Neural Networks (IJCNN)
Understanding societies of individuals is a challenging task. In several ant species individuals ... more Understanding societies of individuals is a challenging task. In several ant species individuals seem to have the same physical characteristics and, in principle, could assume any role the social environment requires. However, most ants’ societies present a caste organization in their colonies with particular roles. For the biologist understanding the complex dynamics ruling a colony is hard due to the difficulty of collecting and classifying long term ant activities in the field. The relocation phenomena which generally correspond to perturbation phases suffered by ant colonies are critical moments in the colony history and require rapid and effective response from workers in order to ensure the survival of the colony and its setup in a new safe nest. How the colonial group deals with this process? Is there a particular organization of the workers during the relocation periods? Here, we applied an experimentally induced protocol in the laboratory combined with behavioral observatio...
Développement social et variabilité inter-sexe : dynamique développementale des différences sexuées entre 2 et 6 ans
Développement et variabilités
National audienc
Incorporating Neighborhood Information During NMF Learning
Communications in Computer and Information Science, 2021
Aust. J. Intell. Inf. Process. Syst., 2019
Research on Multi-View Clustering (MVC) has become more and more attractive thanks to the richnes... more Research on Multi-View Clustering (MVC) has become more and more attractive thanks to the richness of its application in several fields. In this paper, we present a novel framework for MVC, inspired by the Optimal Transport (OT) theory to learn the clusters in each view and compute an ensemble clustering of all the views in order to form a consensus cluster. We propose two algorithms for this purpose: a Consensus Projection Approach (CPA) and a Consensus with New Representation (CNR), both based on the entropy regularized Wasserstein distance and the Wasserstein barycenters between the data distributions in each view. To validate the approaches, extensive experiments were conducted on multiple data-sets.
Aust. J. Intell. Inf. Process. Syst., 2019
Hierarchical structures are known since decades for their outstanding properties that make them i... more Hierarchical structures are known since decades for their outstanding properties that make them ideal for representing data and has been suggested as a particularly important method for organizing concepts. This paper fills a big gap between Multilayer Nonnegative Matrix Factorization and hierarchical structures. We prove that this prototype based model is a deep architecture. We prove mathematically and by experiments that each layer depends on the preceding layers, even being trivial it doesn’t exist any stated proof of this. We conclude that different layers in Multilayer Nonnegative Matrix Factorization are not only dependant but also the order of construction is prominent. In other words, Multilayer NMF is indeed a hierarchical dimensionality reduction and clustering method. It involves learning multiple levels of representation, corresponding to different levels of abstractions.
Subspace Guided Collaborative Clustering Based on Optimal Transport
Sex differences in social play: developmental trends in early childhood
Sex differences in play represent one of the largest non-reproductive physical or psychological s... more Sex differences in play represent one of the largest non-reproductive physical or psychological sex differences and have been widely observed across taxa and cultures. Several aspects of play – including toys and activities, playmates’ sex, and play styles – differ between males and females in various animal species, including humans. In early childhood, boys’ and girls’ play styles are characterized by different behaviors and patterns of social interaction. Girls’ play is assumed to be more social, cooperative, structured and adult oriented than boys’ play. However, there is a notable lack of developmental studies in order to track whether and how sex differences change over time. The present study focuses on age and sex differences in social play by 2- to 6- year old children, when most children begin to experience peer social interactions. Crosssectional observations of social participation of 164 children from 16 classrooms in three French nursery schools were made during outdoo...
Unsupervised machine learning approa ches involving several clustering algorithms working togethe... more Unsupervised machine learning approa ches involving several clustering algorithms working together to tackle difficult data sets are a recent area of research with a large number of applications such as clustering of distributed data, multi-expert clustering, multi-scale clustering analysis or multi-view clustering. Most of these frameworks can be regrouped under the umbrella of collaborative clustering, the aim of which is to reveal the common underlying structures found by the different algorithms while analyzing the data. Within this context, the purpose of this article is to propose a collaborative framework lifting the limitations of many of the previously proposed methods: Our proposed collaborative learning method makes possible for a wide range of clustering algorithms from different families to work together based solely on their clustering solutions, thus lifting previous limitation requiring identical prototypes between the different collaborators. Our proposed framework ...
Résumé. Le clustering collaboratif est un domaine émergeant du machine learning à fort potentiel ... more Résumé. Le clustering collaboratif est un domaine émergeant du machine learning à fort potentiel applicatif, ayant des similarités avec l’apprentissage par ensemble et l’apprentissage par transfert. Dans cette article, nous proposons une méthode permettant de combiner un framework collaboratif avec la structure des Cartes Topographiques (GTM) afin d’obtenir un algorithme permettant de l’apprentissage par transfert entre algorithmes travaillant sur des données similaires. Notre approche a été validée sur plusieurs jeux de données et a montré un fort potentiel.
Learning Useful Representations Through Stacked Self-Organizing Maps
2018 International Joint Conference on Neural Networks (IJCNN), 2018
In this work we explore an original strategy for building deep networks, based on stacking layers... more In this work we explore an original strategy for building deep networks, based on stacking layers of Self-Organizing Maps (SOM) with finite weights. We aim to show that our approach, with enough hidden variables, is capable to represents any “soft” distribution over the visible variables, where “soft” means that the distribution does not contain any probabilities of 1 or 0. The algorithm compute the model one layer at a time. The first layer receives the input observations and compute a probability of membership for each observation and each neuron. These probabilities become the input of the second layer, which compute a new set of probabilities, and so on. The number of neurons decrease in each layer after the first. The proposed algorithm is experimentally tested on artificial and real data-sets. The effect of the added hidden layers for the representation of data structure is experimentally demonstrated.
This paper deals with a clustering algorithm for histogram data based on a Self-Organizing Map (S... more This paper deals with a clustering algorithm for histogram data based on a Self-Organizing Map (SOM) learning. It combines a dimension reduction by SOM and the clustering of the data in a reduced space. Related to the kind of data, a suitable dissimilarity measure between distributions is introduced: the L_2 Wasserstein distance. Moreover, the number of clusters is not fixed in advance but it is automatically found according to a local data density estimation in the original space. Applications on synthetic and real data sets corroborate the proposed strategy.
Radio Frequency IDentification (RFID) is an advanced tracking technology that can be used to stud... more Radio Frequency IDentification (RFID) is an advanced tracking technology that can be used to study the spatial organization of individual's spatio-temporal activity. The aim of this work is firstly to build a new RFID-based autonomous system which can follow individuals' spatiotemporal activity, a tool not currently available. Secondly, we aim to develop new tools for automatic data mining. In this paper, we study how to transform these data to investigate the division of labor, the intra-colonial cooperation and conflict in an ant colony. We also develop a new unsupervised learning data mining method (DS2L-SOM: Density-based Simultaneous Two-Level -Self Organizing Map) to find homogeneous clusters (i.e., sets of individual which share a similar behavior). According to the experimental results, this method is very fast and efficient. It also allows a very useful visualization of the results.
Self-Organizing Maps, 2010
Unsupervised Learning for Analyzing the Dynamic Behavior of Online Banking Fraud
2013 IEEE 13th International Conference on Data Mining Workshops, 2013
ABSTRACT In many cases, databases are in constant evolution, new data is arriving continuously. D... more ABSTRACT In many cases, databases are in constant evolution, new data is arriving continuously. Data streams pose several unique problems that make obsolete the applications of classical data analysis methods. Indeed, these databases are constantly on-line, growing with the arrival of new data. In addition, the probability distribution associated with the data may change over time. In online banking, fraud is one of the major ethical issues. For this challenge, the main aims of the data mining approaches are, firstly, to identify the different types of credit card fraud, and, secondly, for the fraud detection. We propose in this paper a method of synthetic representation of the data structure for efficient storage of information, and a measure of dissimilarity between these representations for the detection of change in the stream structure, in order to detect different types of fraud during the a period of time. The proposed approach was validated on a real application for the on-line credit card fraud detection.
Uploads
Papers by Guenael Cabanes