http://dspace.univ-mascara.dz:8080/jspui/handle/123456789/15 2026-05-20T05:02:48Z http://dspace.univ-mascara.dz:8080/jspui/handle/123456789/1441 Titre: Design of Bio-Inspired Metaheuristics for Medical Image Classification Auteur(s): KHALDI, Brahim Résumé: Breast cancer diagnosis relies fundamentally on histopathological examination of tissue samples; however, manual microscopic evaluation by pathologists is subjective, time-consuming, and prone to inter-observer variability. While deep learning has demonstrated remarkable success in medical image analysis, conventional single-architecture models exhibit inherent limitations. Convolutional neural networks excel at local feature extraction but fail to capture global contextual relationships, whereas Vision Transformers effectively model long-range dependencies but may be inefficient in representing fine-grained local details that are critical for histopathological interpretation. Moreover, optimizing deep learning models through hyperparameter tuning and feature selection remains computationally expensive and often suboptimal when relying solely on gradient-based optimization techniques. This thesis proposes HNet, a novel hybrid deep learning architecture that strategically integrates three complementary neural paradigms for histopathological image classification. HNet combines EfficientNet for efficient local feature extraction from tissue morphology, an Advanced Vision Transformer (AVT) for capturing global contextual relationships and long-range tissue patterns, and Capsule Networks (CapsNet) for explicitly modeling spatial hierarchies and part-whole relationships inherent in tissue organization. To address the challenges of high-dimensional feature spaces and suboptimal parameter tuning, a Genetic Algorithm (GA)-based feature selection mechanism is incorporated as a critical preprocessing step between the concatenated EfficientNet-AVT feature representations and the CapsNet input. This metaheuristic-driven optimization enables automatic identification of the most discriminative features while reducing dimensionality and computational complexity. Comprehensive experimental validation is conducted on the BreakHis dataset, encompassing binary (benign vs. malignant) breast cancer classification tasks. Detailed ablation studies quantify the individual contributions of each architectural component as well as the impact of GA-based feature selection on overall performance. Comparative evaluation against recent state-of-the-art methods demonstrates that the proposed HNet architecture achieves superior classification accuracy, sensitivity, specificity, and F1-score, establishing a new benchmark for breast cancer histopathology classification. The integration of metaheuristic-driven optimization with hybrid deep learning significantly enhances model robustness, generalization to unseen data, and computational efficiency compared to non- optimized approaches. Beyond empirical performance gains, this work demonstrates how hybrid metaheuristic, deep learning frameworks can be effectively integrated into clinical workflows by addressing real-world deployment constraints, including inference latency, computational resource utilization, and model interpretability. The proposed methodology provides a generalizable template for applying metaheuristic-optimized 5 hybrid deep learning to a wide range of medical image classification tasks, offering substantial potential for advancing automated computer-aided diagnosis systems and digital pathology. 2026-05-06T00:00:00Z http://dspace.univ-mascara.dz:8080/jspui/handle/123456789/1432 Titre: Algorithmics and Data Structures 1 : Lecture notes and solved exercises Auteur(s): ZAGANE, Mohammed 2026-05-03T00:00:00Z http://dspace.univ-mascara.dz:8080/jspui/handle/123456789/1429 Titre: Improving IoT Network Security Using Deep Learning and Hybrid Feature Selection Techniques Auteur(s): Khatemi, Fouad Résumé: The rapid expansion of the Internet of Things (IoT) has introduced significant security challenges due to heterogeneous devices, high-dimensional data, and evolving cyberattacks. Traditional intrusion detection systems (IDSs) often struggle to achieve high accuracy while maintaining computational efficiency in such environments. This thesis proposes an enhanced deep learning-based IDS that integrates a dual statistical feature selection framework to improve detection performance and reduce computational complexity. The proposed approach combines Pearson correlation and Mutual Information in a two-stage feature selection process to identify the most relevant features from IoT network traffic. This framework is integrated with a fully connected deep neural network designed to learn complex intrusion patterns. Two IDS architectures are evaluated using the NSL-KDD and RT-IoT2022 datasets. Experimental results show that the proposed system achieves high detection accuracy, exceeding 99% in most scenarios, while reducing training time. Comparative analysis with existing machine learning and deep learning approaches confirms the effectiveness and robustness of the proposed solution for IoT intrusion detection. 2026-04-27T00:00:00Z http://dspace.univ-mascara.dz:8080/jspui/handle/123456789/1414 Titre: Segmentation automatique des régions d'intérêt dans des images réelles Auteur(s): BOUDAIEB, Ahmed Résumé: Cette thèse porte sur la segmentation automatique des régions d’intérêt dans des images réelles, avec une application spécifique à la détection des lésions cutanées. Nous proposons une architecture de segmentation basée sur U-Net améliorée par un encodeur ResNet50V2, une stratégie d’augmentation avancée et une fonction de perte hybride Dice + BCE. Le modèle a été évalué sur le dataset PH2 puis testé sur un jeu de données externe (ISIC 2016) afin d’analyser sa robustesse et sa capacité de généralisation. Les résultats obtenus montrent des performances élevées en segmentation, surpassant plusieurs méthodes traditionnelles et modèles profonds de référence. Les analyses menées, incluant une étude d’ablation et une comparaison complète avec l’état de l’art, confirment l’efficacité du modèle proposé pour la segmentation précise des lésions cutanées dans des conditions réelles. 2026-03-24T00:00:00Z