Faiez Gargouri

University of Sfax | US · Higher Institute of Computer Sceince and Multimedia of Sfax (ISIM)

Artificial neural networks (ANNs) are being widely used in supervised machine learning to analyze signals or images for many applications. Using an annotated learning database, one of the main challenges is to optimize the network weights. A lot of work on solving optimization problems or improving optimization methods in machine learning has been proposed successively such as gradient-based method, Newton-type method, meta-heuristic method. For the sake of efficiency, regularization is generally used. When non-smooth regularizers are used especially to promote sparse networks, such as the ℓ1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ell _1$$\end{document} norm, this optimization becomes challenging due to non-differentiability issues of the target criterion. In this paper, we propose an MCMC-based optimization scheme formulated in a Bayesian framework. The proposed scheme solves the above-mentioned sparse optimization problem using an efficient sampling scheme and Hamiltonian dynamics. The designed optimizer is conducted on four (4) datasets, and the results are verified by a comparative study with two CNNs. Promising results show the usefulness of the proposed method to allow ANNs, even with low complexity levels, reaching high accuracy rates of up to 94%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$94\%$$\end{document}. The proposed method is also faster and more robust concerning overfitting issues. More importantly, the training step of the proposed method is much faster than all competing algorithms.

The massive use of ontologies generates a large amount of semantic data. To facilitate their management, persistent solutions for storing and querying these semantic data loads have been proposed. This gave rise to a new type of databases, called ontology-based databases (OBDB). In recent years, the need for data and real-time services has increased significantly in a large number of applications. However, the OBDB does not implement any mechanism to address real-time applications which are characterized, not only by handling large amounts of data, but also by temporal constraints, to which can be submitted data and treatments. As well, geographically extended applications, requiring using real-time databases that manage data and distributed processing are increasingly needed.These applications are managed by Distributed Real-Time DataBase Management System (DRTDBMS). Like any system, the DRTDBMS, often go through overload phases, due to the unpredictable arrival of transactions submitted by users. In order to better manage Quality of Service (QoS) in these systems by facing instability periods, approaches based on Distributed Feedback Control Scheduling (DFCS) were proposed. These approaches does not address the use of ontological data. In this paper, we propose an approach aiming to enhance QoS in DRTDBMS based on data replication. It consists in extending the DFCS architecture by the manipulation of ontological data as well as handling the execution of accessing transactions. In the extension we propose, we study the applicability of different data replication policies. The proposed architecture is then called Replication-Based-Distributed Feedback Control Scheduling Architecture for Real-Time Ontology (Replication-Based-DFCS-RTO). We also show the contribution provided by our approach through simulation results.

Dealing with imperfect temporal data entries in the context of Collective and Personal Memory applications is an imperative matter. Data are structured semantically using an ontology called “Collective Memo Onto”. In this paper, we propose an approach that handles temporal data imperfections in OWL 2. We reduce to four types of imperfection defined in our typology of temporal data imperfections which are imprecision, uncertainty, simultaneously uncertainty and imprecision and conflict. The approach consists of representing imperfect quantitative and qualitative time intervals and time points by extending the 4D-fluents approach and defining new components, as well as reasoning about the handled data by extending the Allen’s Interval algebra. Based on both extensions, we propose an OWL 2 ontology named “TimeOntoImperfection”. The proposed qualitative temporal relations are inferred via a set of 924 SWRL rules. We validate our work by implementing a prototype based on the proposed ontology and we apply it in the context of the Collective Memory Temporal Data.

Efficient and accurate early prediction of Alzheimer's disease (AD) based on the neuroimaging data has attracted interest from many researchers to prevent its progression. Deep learning networks have demonstrated an optimal ability to analyse large-scale multimodal neuroimaging for AD classification. The most widely used architecture of deep learning is the Convolution neural networks (CNN) that have shown great potential in AD detection. However CNN does not capture long range dependencies within the input image and does not ensure a good global feature extraction. Furthermore, increasing the receptive field of CNN by increasing the kernels sizes can cause a feature granularity loss. Another limitation is that CNN lacks a weighing mechanism of image features; the network doesn’t focus on the relevant features within the image. Recently,vision transformer have shown an outstanding performance over the CNN and overcomes its main limitations. The vision transformer relies on the self-attention layers. The main drawbacks of this new technique is that it requires a huge amount of training data. In this paper, we combined the main strengths of these two architectures for AD classification. We proposed a new method based on the combination of the Cross ViT and Wide Residual Squeeze-and-Excitation Network. We acquired MRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Open Access Series of Imaging Studies (OASIS). We also proposed a new data augmentation based on the self attention progressive generative adversarial neural network to overcome the limitation of the data. Our proposed method achieved 99% classification accuracy and outperforms CNN models.

Personalised medicine is a new approach that ensure a tolerant and optimal diagnosis for the patient basing on his own data and its profile information such as life style, medical history, genetic data, behaviours, and his environment. These data is vital to predict the potential disease progression. Extracting insights from these heterogeneous data is a challenging task. Brain disorders such as neurodegenerative diseases detection and prediction is still an open challenge for research. The early prediction of these diseases is the key solution to prevent their progression. Deep learning methods has shown an outstanding performance on the brain diseases diagnosis such as the Alzheimer’s disease (AD). In this paper we present two contributions. Firstly, we adopt a conditional generative adversarial network for data augmentation based on two datasets the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Open Access Series of Imaging Studies (OASIS) as a solution for the shortage of health information. Additionally, we present a new method for an early detection of the Alzheimer’s disease based on the combination of the Densenet 201 network and the Squeeze and Excitation network. Furthermore, we compare our approach with the traditional Densenet 201, Squeeze and Excitation network and other networks. We figure out that our approach yields best results over these networks. We reach an accuracy of 98%.