Showing papers by "Orange S.A. published in 2016"

Random Forest for the Contextual Bandit Problem

Abstract: To address the contextual bandit problem, we propose an online random forest algorithm. The analysis of the proposed algorithm is based on the sample complexity needed to find the optimal decision stump. Then, the decision stumps are recursively stacked in a random collection of decision trees, BANDIT FOREST. We show that the proposed algorithm is optimal up to logarithmic factors. The dependence of the sample complexity upon the number of contextual variables is logarithmic. The computational cost of the proposed algorithm with respect to the time horizon is linear. These analytical results allow the proposed algorithm to be efficient in real applications, where the number of events to process is huge, and where we expect that some contextual variables, chosen from a large set, have potentially non-linear dependencies with the rewards. In the experiments done to illustrate the theoretical analysis, BANDIT FOREST obtain promising results in comparison with state-of-the-art algorithms.

Fault and Byzantine Tolerant Self-stabilizing Mobile Robots Gathering — Feasibility Study —

Abstract: Gathering is a fundamental coordination problem in cooperative mobile robotics. In short, given a set of robots with arbitrary initial locations and no initial agreement on a global coordinate system, gathering requires that all robots, following their algorithm, reach the exact same but not predetermined location. Gathering is particularly challenging in networks where robots are oblivious (i.e., stateless) and direct communication is replaced by observations on their respective locations. Interestingly any algorithm that solves gathering with oblivious robots is inherently self-stabilizing if no specific assumption is made on the initial distribution of the robots. In this paper, we significantly extend the studies of de-terministic gathering feasibility under different assumptions This manuscript considerably extends preliminary results presented as an extended abstract at the DISC 2006 conference [7]. The current version is under review at Distributed Computing Journal since February 2012 (in a previous form) and since 2014 in the current form. The most important results have been also presented in MAC 2010 organized in Ottawa from August 15th to 17th 2010 related to synchrony and faults (crash and Byzantine). Unlike prior work, we consider a larger set of scheduling strategies , such as bounded schedulers. In addition, we extend our study to the feasibility of probabilistic self-stabilizing gathering in both fault-free and fault-prone environments.

Distance maps: a robust illumination preprocessing for active appearance models

Abstract: Methods of deformable appearance models are useful for realistically modelling shapes and textures of visual objects for reconstruction. A first application can be the fine analysis of face gestures and expressions from videos, as deformable appearance models make it possible to automatically and robustly locate several points of interest in face images. That opens development prospects of technologies in many applications like video coding of faces for videophony, animation of synthetic faces, word visual recognition, expressions and emotions analysis, tracking and recognition of faces. However, these methods are not very robust to variations in the illumination conditions, which are expectable in non constrained conditions. This article describes a robust preprocessing method designed to enhance the performances of deformable models methods in the case of lighting variations. The proposed preprocessing is applied to the Active Appearance Models (AAM). More precisely, the contribution consists in replacing texture images (pixels) by distance maps as input of the deformable appearance models methods. The distance maps are images containing information about the distance between edges in the original object images, which enhance the robustness of the AAMs models against lighting variations.

5G on Board: How Many Antennas Do We Need on Connected Cars?

Abstract: Mobile networks will support increasing numbers of connected vehicles. Successive generations of mobile networks have reduced the cost of data rate, in terms of spectrum usage and power consumption at the base station, by increasingly exploiting the concept of channel state information at the transmitter. Unfortunately, beyond a limiting velocity (which depends on the carrier frequency), networks are no longer cost efficient, since such information is not usable. Recently, channel prediction techniques requiring several antennas on the car roof have been introduced to solve this problem. In this paper, for the first time, we determine the most cost efficient configurations, in terms of numbers of antennas on the car roof and carrier frequency, in various scenarios (highway and dense urban). Our studies show that with a simple prediction technique based on predictor antennas, the network can use twice less spectrum and around 20 dB less power, for cars with 3 antennas on their tops than for cars with the same number of antennas and not using prediction.

Joint Optimization of User Association and User Satisfaction in Heterogeneous Cellular Networks

Abstract: In this paper, we address the joint problem of user association and resource allocation in wireless heterogeneous networks. Therefore, we formulate an optimization approach considering two objectives, namely, maximizing the number of served User Equipments (UEs) and maximizing the sum of the UE utilities. Precisely, the aim is to associate UEs with the optimal Radio Access Technology (RAT) and to allocate to these UEs the optimal Resource Units (RUs) based on their requested services and contracts. Our problem is challenging because it is mixed integer non-linear optimization. To tackle this difficulty, we provide a Mixed Integer Linear Programming (MILP) re-formulation of the problem that makes it computationally tractable. Various preferences for user association and resource allocation are conducted by tuning: on the one hand, the weights associated with different services and contracts, on the other hand, the weights associated with the considered two objectives. The optimal solution of the MILP problem is computed for a realistic network scenario and compared with legacy solution. Extensive simulation results show that the proposed optimization approach improves the overall network performance while considering the UE requested service and contract: it outperforms legacy solutions in terms of user satisfaction. Moreover, it provides an efficient distribution of UEs on the different RATs.

Web Chat Conversations from Contact Centers: a Descriptive Study.

Abstract: In this article we propose a descriptive study of a chat conversations corpus from an assistance contact center. Conversations are described from several view points, including interaction analysis, language deviation analysis and typographic expressivity marks analysis. We provide in particular a detailed analysis of language deviations that are encountered in our corpus of 230 conversations, corresponding to 6879 messages and 76839 words. These deviations may be challenging for further syntactic and semantic parsing. Analysis is performed with a distinction between Customer messages and Agent messages. On the overall only 4% of the observed words are misspelled but 26% of the messages contain at least one erroneous word (rising to 40% when focused on Customer messages). Transcriptions of telephone conversations from an assistance call center are also studied, allowing comparisons between these two interaction modes to be drawn. The study reveals significant differences in terms of conversation flow, with an increased efficiency for chat conversations in spite of longer temporal span.

Coverage and performance of stratospheric balloons wireless networks

Abstract: In this paper we analyze the performance of wireless networks by using stratospheric balloons equipped by base stations (BS). We show that such a solution can be interesting in terms of performance and coverage. Considering a high altitude wireless network, we first establish an analytical expression of the Signal to Interference plus Noise Ratio (SINR) received by a user equipment (UE) located on the ground. We establish the cumulative distribution function (CDF) of the SINR, since it allows an accurate evaluation of the performance and the quality of service (QoS). We compare different deployment scenarios of high altitude balloons. This allows us to analyze and to quantify the impact of the use of balloons, in terms of coverage and performance for low dense population wide areas.

An Inverse Problem Approach for Content Popularity Estimation

Abstract: The Internet increasingly focuses on content, as exemplified by the now popular Information Centric Networking paradigm. This means, in particular, that estimating content popularities becomes essential to manage and distribute content pieces efficiently. In this paper, we show how to properly estimate content popularities from a traffic trace. Specifically, we consider the problem of the popularity inference in order to tune content-level performance models, e.g. caching models. In this context, special care must be taken due to the fact that an observer measures only the flow of requests, which differs from the model parameters, though both quantities are related by the model assumptions. Current studies, however, ignore this difference and use the observed data as model parameters. In this paper, we highlight the inverse problem that consists in determining parameters so that the flow of requests is properly predicted by the model. We then show how such an inverse problem can be solved using Maximum Likelihood Estimation. Based on two large traces from the Orange network and two synthetic datasets, we eventually quantify the importance of this inversion step for the performance evaluation accuracy.

Joint Optimization of User Association and User Satisfaction in Heterogeneous Cellular Networks.

Abstract: In this paper, we address the joint problem of user association and resource allocation in wireless heterogeneous networks. Therefore, we formulate an optimization approach considering two objectives, namely, maximizing the number of served User Equipments (UEs) and maximizing the sum of the UE utilities. Precisely, the aim is to associate UEs with the optimal Radio Access Technology (RAT) and to allocate to these UEs the optimal Resource Units (RUs) based on their requested services and contracts. Our problem is challenging because it is mixed integer non-linear optimization. To tackle this difficulty, we provide a Mixed Integer Linear Programming (MILP) re-formulation of the problem that makes it computationally tractable. Various preferences for user association and resource allocation are conducted by tuning: on the one hand, the weights associated with different services and contracts, on the other hand, the weights associated with the considered two objectives. The optimal solution of the MILP problem is computed for a realistic network scenario and compared with legacy solution. Extensive simulation results show that the proposed optimization approach improves the overall network performance while considering the UE requested service and contract: it outperforms legacy solutions in terms of user satisfaction. Moreover, it provides an efficient distribution of UEs on the different RATs.

Aging of Silica Optical Fiber in Static Fatigue Tests

Abstract: During their use, optical fibers are subject to harsh installation and environmental conditions. To evaluate more precisely the lifetime of an optical fiber, it is necessary to study the mechanical behavior of optical fibers under extreme conditions, in particular under mechanical and thermal stress.This paper presents the results of new silica optical fibers aged in hot water between 20°C and 70°C and subjected to mechanical static bending stresses from 3 GPa to 3.5 GPa. Thermal dependence of the time to failure was observed. This dependence can be described by the Arrhenius model, where the activation energy is one of the main physical characteristic.

A Contextual Data Selection Tool for an Enhanced Business Process Analysis

Abstract: Process mining has an important place in business process (BP) analysis. It aims to analyze process events in order to discover the related BP model. However, these techniques are only based on process events and sometimes on business data, leaving aside a large set of data, namely the BP execution context. Existing studies have shown the benefits of considering the context in the BP analysis but they only suggest manual techniques to bind a BP with its context, which is not scalable and time consuming in real deployment environment. To address this issue, we propose a semi-automatic BP contextualization solution which takes into account the BP execution context in the BP analysis time. It uses semantic techniques to perform a matching of the BP model with contextual data and with business data, and then it obtains the value of these data during the BP execution. In this paper, we present a tool that implements this solution in order to enhance current analysis techniques which facilitates the monitoring and enables deep BP analysis. The proposed tool is validated by its application to a real life process, a "palettization" process.

A practical study of neural network-based image classification model trained with transfer learning method.

Abstract: This paper deals with algorithms for image classification, which aim to guess “what is on the picture” using human-readable labels or categories. A supervised learning approach with Convolutional Neural Networks (CNNs) is studied as an effective solution to different computer vision problems, including image classification. Main contribution of this paper is a set of practical guidelines to tackle the image classification problem using publicly available tools and typical hardware platforms.

Exploring Collections of Multimedia Archives Through Innovative Interfaces in the Context of Digital Humanities.

Abstract: STIK is a platform that gathers Speech, Texts and Images of Knowledge. It allows browsing and navigating through collections of multimedia, facilitating access to archives in the domain of Knowledge resources. STIK includes a back-end with a specific automatic metadata extraction pipeline, a frontend with innovative interfaces for navigating within a document and a specific implementation of a search engine with dedicated key-word search functionality. It gathers multimedia contents from Canal-U, a French institution that exploits audiovisual archives produced by Higher Education and Research, with various formats and various academic disciplines. STIK is a contribution to the emerging domain of Digital Humanities.

Non-intentional mobility of data sessions across mobile networks while roaming because of operators' mobile DNS architecture

Abstract: The Domain Name System (DNS) is used almost everywhere in mobile networks to determine the IP addresses of the network elements and thanks to the DNS architecture of the inter-operator IP backbone network, a network element of a public mobile network (PMN) can easily determine IP addresses of any other PMN. The analysis of some customers' complaints showed that it could be the cause of billing mistakes. Consequently we decided to go further into the reasons behind. The aim of the paper is to alert operators to the risk of data session mobility across mobile networks when inter-operator DNS procedures allow the new PMN to exchange with the old PMN and retrieve the context of the data session. It explains the associated risk of charging/billing problems for the roamers because of non-intentional inter-PMN mobility in case the home charging/billing architecture is not ready to deal with a PMN change during a data session. Finally, it introduces a few solutions to forbid/control inter-PMN mobility of mobile data sessions.

Personalizing Information Retrieval Using an Extension of a Dung Argumentation

Abstract: In this paper, we propose an architecture for personalizing information retrieval (IR), exploiting the interactions between the user and the social network. We use an extension of a Dung argumentation framework to show how the precision of the personalized information retrieval system could be improved. We use also social media and search history to define the user-profile which is represented by a restriction of a Description Logic.