Search

Showing total 4 results
Start Over Topic machine learning Publication Year Range Last 10 years Language french Database Academic Search Index
4 results

1. Médiations interpersonnelles et évolutivité de systèmes à base d'IA: Un exemple à partir de l'appropriation de chatbots en entreprise.

Author

Gentiletti, M. Gras, Fréjus, M., Bourmaud, G., and Decortis, F.

Subjects
*ARTIFICIAL intelligence, *SOCIOTECHNICAL systems, *CHATBOTS, *MACHINE learning, *TELECOMMUNICATION systems
Abstract

During the 2010's, The new communication networks, the computing power improvement, the Big data and machine learning have seen a sharp increase recently and have created the conditions favourable conducive to Artificial Intelligence. Among these AI-based devices, the rise of conversational agents has been growing since 2010. The massive deployment of chatbots in companies testifies to this craze and raises questions about the relationship between humans and this new type of machine that would be able to dialogue 'like a human'. In this paper, we stand on a study carried out in 2019, at the request of the Artificial Intelligence division of EDF R&D, aimed at characterizing the impact of chatbots on the group's (historical) professions. We will see how an approach in terms of mediations comes to replace human activity at the heart of the sociotechnical system and questions, in particular, the role of interpersonal mediations and the impact of not taking them into account on the learning and performance chatbots. [ABSTRACT FROM AUTHOR]

Published
2020

2. Le vandalisme dans l'information géographique volontaire: Du concept à la détection non supervisée d'anomalie.

Author

Quy Thy Truong, Touya, Guillaume, and de Runz, Cyril

Subjects
*OUTLIER detection, *VANDALISM, *MACHINE learning, *CASE studies, *CORPORA
Abstract

Since vandalism is a serious matter for the quality of Volunteered Geographic Information, this paper aims at exploring machine learning techniques that enable its detection. First, a focus on the various definitions of vandalism highlights the complexity of this concept. This focus comprises a case study on proven vandals in OpenStreetMap (OSM). Second, we present an experimental vandalism detection on OSM data using a clustering-based outlier detection. The analysis of initial results leads to a discussion about the construction of an OSM vandalism corpus that would be useful in a supervised learning context. [ABSTRACT FROM AUTHOR]

Published
2019
Full Text
View/download PDF

3. Une approche pour la sélection de variables stables : application à l'encodage des diagnostics secondaires.

Author

Chahbandarian, Ghazar, Bricon-Souf, Nathalie, Megdiche, Imen, Bastide, Rémi, and Steinbach, Jean-Christophe

Subjects
*MACHINE learning, *FORECASTING, *MEDICAL informatics, *MEDICAL care, *HOSPITALS
Abstract

In this paper, we focus on applying feature selection in the context of secondary diagnoses prediction starting from medico-economic data sources. The results of the prediction is used as guidelines for encoding secondary diagnoses which is a sensitive task in the hospitals requiring a lot of attention in order to be achieved properly. We propose a practical approach to select stable and relevant features from imbalanced datasets. The stability of features is obtained through the convergence of several FS methods to a fair number of features without being impacted by the sampled dataset. The quality of featured shall be deducted from the quality prediction of machine learning algorithms on the selected features. We evaluate the proposed approach on the PMSI database of the CHIC-CM hospital. Our results are quite interesting and opening discussions for these specific health care data supports. [ABSTRACT FROM AUTHOR]

Published
2018
Full Text
View/download PDF

4. L'usage des entropies est-il justifie en apprentissage a partir des donnees?

Author

ZIGHED, DJAMEL A.

Subjects
*ENTROPY (Information theory), *STRUCTURAL optimization, *ALGORITHMS, *MACHINE learning, *MATHEMATICAL models
Abstract

Many machine learning algorithms use entropy measures as a criterion of construction that they seek to optimize. Among the most applied measures, Shannon's entropy is certainly the most known. However, in the real world applications, the use of the entropy measure turns out to be totally inadequate both in theory and in practice. Indeed, many hypothesis are in fact implicitly assumed whereas they are unfounded, therefore unjustified. In this paper, we will try to identify those hypothesis and we will demonstrate that they are unsuitable in machine learning with real data. Then, we will introduce, intuitively, a set of new prosperities that should be required for measures that are supposed to lead to efficients algorithms. [ABSTRACT FROM AUTHOR]

Published
2014