Centre d'études et de recherche en informatique et communications

Clustering is the partitioning of a set of objects into groups (clusters) so that objects within a group are more similar to each others than objects in different groups. Most of the clustering algorithms depend on some assumptions in order to define the subgroups present in a data set. As a consequence, the resulting clustering scheme requires some sort of evaluation as regards its validity.The evaluation procedure has to tackle difficult problems such as the quality of clusters, the degree with which a clustering scheme fits a specific data set and the optimal number of clusters in a partitioning. In the literature, a wide variety of indices have been proposed to find the optimal number of clusters in a partitioning of a data set during the clustering process. However, for most of indices proposed in the literature, programs are unavailable to test these indices and compare them.The R package NbClust has been developed for that purpose. It provides 30 indices which determine the number of clusters in a data set and it offers also the best clustering scheme from different results to the user. In addition, it provides a function to perform k-means and hierarchical clustering with different distance measures and aggregation methods. Any combination of validation indices and clustering methods can be requested in a single function call. This enables the user to simultaneously evaluate several clustering schemes while varying the number of clusters, to help determining the most appropriate number of clusters for the data set of interest.

Data imbalance is frequently encountered in biomedical applications. Resampling techniques can be used in binary classification to tackle this issue. However such solutions are not desired when the number of samples in the small class is limited. Moreover the use of inadequate performance metrics, such as accuracy, lead to poor generalization results because the classifiers tend to predict the largest size class. One of the good approaches to deal with this issue is to optimize performance metrics that are designed to handle data imbalance. Matthews Correlation Coefficient (MCC) is widely used in Bioinformatics as a performance metric. We are interested in developing a new classifier based on the MCC metric to handle imbalanced data. We derive an optimal Bayes classifier for the MCC metric using an approach based on Frechet derivative. We show that the proposed algorithm has the nice theoretical property of consistency. Using simulated data, we verify the correctness of our optimality result by searching in the space of all possible binary classifiers. The proposed classifier is evaluated on 64 datasets from a wide range data imbalance. We compare both classification performance and CPU efficiency for three classifiers: 1) the proposed algorithm (MCC-classifier), the Bayes classifier with a default threshold (MCC-base) and imbalanced SVM (SVM-imba). The experimental evaluation shows that MCC-classifier has a close performance to SVM-imba while being simpler and more efficient.

Dossier hors série Les Mathématiques sociales, Pour La Science, Juillet 1999, 36-44

432 pages

3ème édition révisée

Blockchain is a technology making the shared registry concept from distributed systems a reality for a number of application domains, from the cryptocurrency one to potentially any industrial system requiring decentralized, robust, trusted, and automated decision making in a multi-stakeholder situation. Nevertheless, the actual advantages in using blockchain instead of any other traditional solution (such as centralized databases) are not completely understood to date, or at least there is a strong need for a vademecum guiding designers toward the right decision about when to adopt blockchain or not, which kind of blockchain better meets use-case requirements, and how to use it. In this paper, we aim at providing the community with such a vademecum, while giving a general presentation of blockchain that goes beyond its usage in Bitcoin and surveying a selection of the vast literature that emerged in the last few years. We draw the key requirements and their evolution when passing from permissionless to permissioned blockchains, presenting the differences between proposed and experimented consensus mechanisms, and describing existing blockchain platforms.

Real-scale Semantic Web applications, such as Knowledge Portals and E-Marketplaces, require the management of large volumes of metadata, i.e., information describing the available Web content and services. Better knowledge about their meaning, usage, accessibility or quality will considerably facilitate an automated processing of Web resources. The Resource Description Framework (RDF) enables the creation and exchange of metadata as normal Web data. Although voluminous RDF descriptions are already appearing, sufficiently expressive declarative languages for querying both RDF descriptions and schemas are still missing. In this paper, we propose a new RDF query language called RQL. It is a typed functional language (a la OQL) and relies on a formal model for directed labeled graphs permitting the interpretation of superimposed resource descriptions by means of one or more RDF schemas. RQL adapts the functionality of semistructured/XML query languages to the peculiarities of RDF but, foremost, it enables to uniformly query both resource descriptions and schemas. We illustrate the RQL syntax, semantics and typing system by means of a set of example queries and report on the performance of our persistent RDF Store employed by the RQL interpreter.

This paper introduces WILDCAT, a deep learning method which jointly aims at aligning image regions for gaining spatial invariance and learning strongly localized features. Our model is trained using only global image labels and is devoted to three main visual recognition tasks: image classification, weakly supervised object localization and semantic segmentation. WILDCAT extends state-of-the-art Convolutional Neural Networks at three main levels: the use of Fully Convolutional Networks for maintaining spatial resolution, the explicit design in the network of local features related to different class modalities, and a new way to pool these features to provide a global image prediction required for weakly supervised training. Extensive experiments show that our model significantly outperforms state-of-the-art methods.

Artifacts, such as software systems, pervade organizations and society. In the field of information systems (IS) they form the core of research. The evaluation of IS artifacts thus represents a major issue. Although IS research paradigms are increasingly intertwined, building and evaluating artifacts has traditionally been the purview of design science research (DSR). DSR in IS has not reached maturity yet. This is particularly true of artifact evaluation. This paper investigates the “what” and the “how” of IS artifact evaluation: what are the objects and criteria of evaluation, the methods for evaluating the criteria, and the relationships between the “what” and the “how” of evaluation? To answer these questions, we develop a taxonomy of evaluation methods for IS artifacts. With this taxonomy, we analyze IS artifact evaluation practice, as reflected by ten years of DSR publications in the basket of journals of the Association for Information Systems (AIS). This research brings to light important relationships between the dimensions of IS artifact evaluation, and identifies seven typical evaluation patterns: demonstration; simulation- and metric-based benchmarking of artifacts; practice-based evaluation of effectiveness; simulation- and metric-based absolute evaluation of artifacts; practice-based evaluation of usefulness or ease of use; laboratory, student-based evaluation of usefulness; and algorithmic complexity analysis. This study also reveals a focus of artifact evaluation practice on a few criteria. Beyond immediate usefulness, IS researchers are urged to investigate ways of evaluating the long-term organizational impact and the societal impact of artifacts.

Expectiles (EVaR) are a one-parameter family of coherent risk measures that have been recently suggested as an alternative to quantiles (VaR) and to expected shortfall (ES). In this work we review their known properties, we discuss their financial meaning, we compare them with VaR and ES and we study their asymptotic behaviour, refining some of the results in Bellini et al. [(2014). “Generalized Quantiles as Risk Measures.” Insurance: Mathematics and Economics, 54:41–48]. Moreover, we present a real-data example for the computation of expectiles by means of simple Garch(1,1) models and we assess the accuracy of the forecasts by means of a consistent loss function as suggested by Gneiting [(2011). “Making and Evaluating Point Forecast.” Journal of the American Statistical Association, 106 (494): 746–762]. Theoretical and numerical results indicate that expectiles are perfectly reasonable alternatives to VaR and ES risk measures.

MicroRNAs (miRNAs) and transcription factors control eukaryotic cell proliferation, differentiation, and metabolism through their specific gene regulatory networks. However, differently from transcription factors, our understanding of the processes regulated by miRNAs is currently limited. Here, we introduce gene network analysis as a new means for gaining insight into miRNA biology. A systematic analysis of all human miRNAs based on Co-expression Meta-analysis of miRNA Targets (CoMeTa) assigns high-resolution biological functions to miRNAs and provides a comprehensive, genome-scale analysis of human miRNA regulatory networks. Moreover, gene cotargeting analyses show that miRNAs synergistically regulate cohorts of genes that participate in similar processes. We experimentally validate the CoMeTa procedure through focusing on three poorly characterized miRNAs, miR-519d/190/340, which CoMeTa predicts to be associated with the TGFβ pathway. Using lung adenocarcinoma A549 cells as a model system, we show that miR-519d and miR-190 inhibit, while miR-340 enhances TGFβ signaling and its effects on cell proliferation, morphology, and scattering. Based on these findings, we formalize and propose co-expression analysis as a general paradigm for second-generation procedures to recognize bona fide targets and infer biological roles and network communities of miRNAs.

Major interest is currently given to the integration of clusters of virtualization servers, also referred to as `cloudlets' or `edge clouds', into the access network to allow higher performance and reliability in the access to mobile edge computing services. We tackle the edge cloud network design problem for mobile access networks. The model is such that the virtual machines (VMs) are associated with mobile users and are allocated to cloudlets. Designing an edge cloud network implies first determining where to install cloudlet facilities among the available sites, then assigning sets of access points, such as base stations to cloudlets, while supporting VM orchestration and considering partial user mobility information, as well as the satisfaction of service-level agreements. We present link-path formulations supported by heuristics to compute solutions in reasonable time. We qualify the advantage in considering mobility for both users and VMs as up to 20% less users not satisfied in their SLA with a little increase of opened facilities. We compare two VM mobility modes, bulk and live migration, as a function of mobile cloud service requirements, determining that a high preference should be given to live migration, while bulk migrations seem to be a feasible alternative on delay-stringent tiny-disk services, such as augmented reality support, and only with further relaxation on network constraints.

ISBN: 0-470-84766-2Hardcoveroctober 2002

A sleepy driver is arguably much more dangerous on the road than the one who\nis speeding as he is a victim of microsleeps. Automotive researchers and\nmanufacturers are trying to curb this problem with several technological\nsolutions that will avert such a crisis. This article focuses on the detection\nof such micro sleep and drowsiness using neural network based methodologies.\nOur previous work in this field involved using machine learning with\nmulti-layer perceptron to detect the same. In this paper, accuracy was\nincreased by utilizing facial landmarks which are detected by the camera and\nthat is passed to a Convolutional Neural Network (CNN) to classify drowsiness.\nThe achievement with this work is the capability to provide a lightweight\nalternative to heavier classification models with more than 88% for the\ncategory without glasses, more than 85% for the category night without glasses.\nOn average, more than 83% of accuracy was achieved in all categories. Moreover,\nas for model size, complexity and storage, there is a marked reduction in the\nnew proposed model in comparison to the benchmark model where the maximum size\nis 75 KB. The proposed CNN based model can be used to build a real-time driver\ndrowsiness detection system for embedded systems and Android devices with high\naccuracy and ease of use.\n

This techical note deals with interval state observer design for time-varying discrete-time systems. The problem of a similarity transformation computation which connects a (time-varying) matrix and its nonnegative representation is studied. Three solutions are proposed: for a generic time-varying system, a system with positive state, and for a particular class of periodical systems. Numerical simulations are provided to demonstrate advantages of the developed techniques.

Filter-bank multicarrier (FBMC) transmission system was proposed as an alternative approach to orthogonal frequency division multiplexing (OFDM) system since it has a higher spectral efficiency. One of the characteristics of FBMC is that the demodulated transmitted symbols are accompanied by interference terms caused by the neighboring transmitted data in time-frequency domain. The presence of this interference is an issue for some multiple-input multiple-output (MIMO) schemes and until today their combination with FBMC remains an open problem. We can cite, among these techniques, the Alamouti scheme and the maximum likelihood detection (MLD) with spatial multiplexing (SM). In this paper, we shall propose a new FBMC scheme and transmission strategy in order to avoid this interference term. This proposed scheme (called FFT-FBMC) transforms the FBMC system into an equivalent system formulated as OFDM regardless of some residual interference. Thus, any OFDM transmission technique can be performed straightforwardly to the proposed FBMC scheme with a corresponding complexity growth compared to the classical FBMC. First, we will develop the FFT-FBMC in the case of single-input single-output (SISO) configuration. Then, we extend its application to SM-MIMO configuration with MLD and Alamouti coding scheme. Simulation results show that FFT-FBMC can almost reach the OFDM performance, but it remains slightly outperformed by OFDM.

We consider the management of changes in a Web Warehouse of XML data. Our approach is change-centric in that it focuses on deltas, i.e., the changes themselves vs. other approaches based on snapshots or integrated representations. We study a logical representation of changes based on deltas and some particular identifiers, XIDs. We consider implementation issues and analyze the advantages and disadvantages of several storage policies. Based on some requirements (certain functionalities that we want to support efficiently), we motivate the choice of our particular storage policy. It is based on storing the last version, "completed" forward deltas and on an original identification scheme for XML nodes. We report briefly on the status of an implementation within the Xyleme project.

Deep neural networks represent, nowadays, the most effective machine learning technology in biomedical domain. In this domain, the different areas of interest concern the Omics (study of the genome—genomics—and proteins—transcriptomics, proteomics, and metabolomics), bioimaging (study of biological cell and tissue), medical imaging (study of the human organs by creating visual representations), BBMI (study of the brain and body machine interface) and public and medical health management (PmHM). This paper reviews the major deep learning concepts pertinent to such biomedical applications. Concise overviews are provided for the Omics and the BBMI. We end our analysis with a critical discussion, interpretation and relevant open challenges.

Abstract Partial least squares (PLS) refers to a set of iterative algorithms based on least squares that implement a broad spectrum of both explanatory and exploratory multivariate techniques, from regression to path modeling, and from principal component to multi‐block data analysis. This article focuses on PLS regression and PLS path modeling, which are PLS approaches to regularized regression and to predictive path modeling. The computational flows and the optimization criteria of these methods are reviewed in detail, as well as the tools for the assessment and interpretation of PLS models. The most recent developments and some of the most promising on going researches are enhanced. WIREs Comput Stat 2013, 5:1–19. doi: 10.1002/wics.1239 This article is categorized under: Statistical Learning and Exploratory Methods of the Data Sciences > Exploratory Data Analysis Statistical and Graphical Methods of Data Analysis > Multivariate Analysis Statistical Models > Linear Models Algorithms and Computational Methods > Least Squares

Can Augmented Reality (AR) techniques inform the design and implementation of a mobile multimedia guide for the museum setting? Drawing from our experience both on previous mobile museum guides projects and in AR technology, we present a fully functional prototype of an AR-enabled mobile multimedia museum guide, designed and implemented for the Museum of Fine Arts in Rennes, France. We report on the life cycle of the prototype and the methodology employed for the AR approach as well as on the selected mixed method evaluation process; finally, the first results emerging from quantitative evaluation are discussed, supported by evidence and findings from the qualitative part of the assessment process. We conclude with lessons learned during the full circle of conception, implementation, testing and assessment of the guide.