Connected and autonomous vehicles can be highly affected by sudden changes in the provided mobile network coverage. Driving safety increases with a high Signal-to-Noise Ratio (SNR) that enhances the Quality of Service (QoS). Therefore, monitoring the QoS for geolocated vehicular mobility from mobile phones is required to detect position and timestamp of QoS changes, and alert the operators to improve the QoS. However, mobility data sharing raises privacy concerns, which, in turn, limits accessibility to the data. This article proposes an adapted Generative Adversarial Network (GAN) model combined with OverPy tool to generate SNR data associated with real vehicle locations, to enrich the dataset while guaranteeing the privacy of users. The dataset used in this paper to train the model is real opt-in data collected from the Signal Map mobile application from volunteers in the region of Nice (France). The training process of the model is iterative and involves updating the parameters of the generator and discriminator in a way that improves the overall performance of the model. The preliminary evaluation results show that the model generates high-quality traffic mobility data correlated to SNR parameter. Specifically, the synthetic generated data have statistical properties similar to real data reflecting real mobility statistics, and therefore protecting the individual’s privacy.

Recently, convolutional auto-encoders (CAE) were introduced for image coding. They achieved performance improvements over the state-of-the-art JPEG2000 method. However, these performances were obtained using massive CAEs featuring a large number of parameters and whose training required heavy computational power.In this paper, we address the problem of lossy image compression using a CAE with a small memory footprint and low computational power usage.In this work, we propose a constrained approach and a new structured sparse learning method. We design an algorithm and test it on three constraints: the classical ℓ1constraint, the ℓ1,∞and the new ℓ1,1constraint. Experimental results show that the ℓ1,1constraint provides the best structured sparsity, resulting in a high reduction of memory ( 82 %) and computational cost reduction (25 %), with similar rate-distortion performance as with dense networks.

Annotation of biomedical databases by clinicians is a very difficult, sometimes imprecise, and time consuming task. An alternative is to ask the clinician expert for the annotations they are the most confident in, which results in a semi-supervised classification problem. In this paper, we present a new approach to solve semi-supervised classification tasks for biomedical applications, involving a supervised autoencoder network. We train the Semi-Supervised AutoEncoder (SSAE) on labelled data using a double descent algorithm. Then, we classify unlabelled samples using the learned network thanks to a softmax classifier applied to the latent space which provides a classification confidence score for each class. Experiments show that the SSAE outperforms Label Propagation and Spreading and the Fully Connected Neural Network both on a synthetic dataset and on four real-world biological datasets.

This paper concerns the supervised generative non parametric autoencoder. Classical methods are based on variational non supervised autoencoders (VAE). Variational autoencoders encourage the latent space to fit a prior distribution, like a Gaussian. However, they tend to draw stronger assumptions for the data, often leading to higher asymptotic bias when the model is wrong.In this paper, we relax the parametric distribution assumption in the latent space and we propose to learn a non-parametric data distribution of the clusters in the latent space. The network encourages the latent space to fit a distribution learned with the labels instead of the parametric prior assumptions. We have built a network architecture that uses the labels to compute the latent space. Thus we define a global criterion combining classification and reconstruction loss. In addition, we have proposed a ℓ1,1regularization which has the advantage of sparsifying the network and improving the clustering. Finally we propose a tailored algorithm to minimize the criterion with constraint. We demonstrate the effectiveness of our method using the popular image dataset MNIST and two biological datasets.

Deep neural networks (DNN) have been applied recently to different domains and perform better than classical state-of-the-art methods. However the high level of performances of DNNs is most often obtained with networks containing millions of parameters and for which training requires substantial computational power. To deal with this computational issue proximal regularization methods have been proposed in the literature but they are time consuming. In this paper, we propose instead a constrained approach. We provide the general framework for this new projection gradient method. Our algorithm iterates a gradient step and a projection on convex constraints. We studied algorithms for different constraints: the classical ℓ1unstructured constraint and structured constraints such as the £2,1 constraint (Group LASSO). We propose a new ℓ1,1structured constraint for which we provide a new projection algorithm. Finally, we used the recent “Lottery optimizer” replacing the threshold by our ℓ1,1projection. We demonstrate the effectiveness of this method with three popular datasets (MNIST, Fashion MNIST and CIFAR). Experiments with these datasets show that our projection method using this new ℓ1,1structured constraint provides the best decrease in memory and computational power.

The automatic supervision of IT systems is a current challenge at Orange. Given the size and complexity reached by its IT operations, the number of sensors needed to obtain measurements over time, used to infer normal and abnormal behaviors, has increased dramatically making traditional expert-based supervision methods slow or prone to errors. In this paper, we propose a fast and stable method called UnSupervised Anomaly Detection for multivariate time series (USAD) based on adversely trained autoencoders. Its autoencoder architecture makes it capable of learning in an unsupervised way. The use of adversarial training and its architecture allows it to isolate anomalies while providing fast training. We study the properties of our methods through experiments on five public datasets, thus demonstrating its robustness, training speed and high anomaly detection performance. Through a feasibility study using Orange's proprietary data we have been able to validate Orange's requirements on scalability, stability, robustness, training speed and high performance.

RTP and Skype applications generate a growing part of the Internet traffic, especially for users on ADSL because of flatrate tarification. In this study, we focus on these two main Voice over IP traffic protocols to characterize the utilization, the performance and the evolution of VoIP systems over six months. We start by identifying the traffic generated by Skype and other classical RTP applications. We thus characterize the traffic of these applications and identify some statistical behaviors. We then analyze in more details and compare the performance of RTP and Skype applications in order to evaluate their influence over the traffic. From ISPs point of view, recognizing these real time applications and analyzing their characteristics and performance is a determining factor for quality of service and troubleshouting. Our analysis is based on ADSL traffic captured at TCP/UDP level on a Broadband Access Server comprising thousands of users. Thus, using a flow-based analysis, we characterize and compare the VoIP traffic and users, and we draw interesting results on traffic patterns and inter-arrivals, connectivity between users, location of sources and performance limitations.