Publications Scientifiques
Permanent URI for this communityhttps://dspace.univ-boumerdes.dz/handle/123456789/10
Browse
7 results
Search Results
Item Review on deep learning optimization using knowledge and dataset distillation in medical imaging diagnostics(2025) Laribi, Nor-Elhouda; Gaceb, Djamel; Rezoug, Abdellah; Touazi, FaycalThe integration of deep learning-based artificial intelligence solutions in hospital environments introduces significant challenges, including data privacy restrictions, limited computational resources, and constraints related to the quality and simplicity of the models used. In this review, we highlight the recent advancements in knowledge distillation and dataset distillation as emerging solutions to these challenges in the field of medical imaging. These techniques offer practical benefits in clinical settings by enabling faster training, reduced model size, improved inference speed, and enhanced accuracy, while supporting privacy-preserving learning across decentralized systems and edge devices. Knowledge distillation transfers knowledge from a complex to a simple model, enabling efficient deployment without high loss in diagnostic performance. Dataset distillation, by contrast, focuses on synthesizing datasets that match the pretrained model on real data, reducing data storage requirements. Together, these methods improve learning efficiency, model accuracy, and resource optimization in hospital workflows. However, their integration into medical environments also presents limitations. Challenges such as pipeline complexity, scalability issues, and performance inconsistency across architectures or high-resolution tasks still persist. Overall, this review provides a comprehensive overview of potential and limitations of these two types of distillations in healthcare, offering insights into how these methods can support more scalable, accurate, and privacy-aware AI solutions for medical imaging.Item MOGA: Multi-Objective Genetic Algorithm to select Stacking Ensemble Learning for classification(Elsevier, 2025) Rezoug, Abdellah; Bader-el-den, MohamedStacking Ensemble Learning (SEL) has been effectively integrated with Multi-Objective Optimisation (MOO) heuristics for classification tasks across various domains, including finance, healthcare, and cybersecurity. This study aims to address the challenge of generalising SEL to a diverse set of classification cases. Thus, the Multi-Objective Genetic Algorithm (MOGA) framework is proposed, utilising a Genetic Algorithm (GA) to evolve a population of distinct SELs, each built from a varied subset of base models. The goal is to select the subset that composes the most effective SEL for a given classification task. MOGA is designed with two main objectives—maximising precision and recall—which helps to maintain independence from any specific classification case. In addition, incorporating models of varied types ensures adaptability and high performance in different situations. Comprehensive experimentation was conducted on 23 diverse datasets, where MOGA demonstrated high performance in nearly all datasets, outperforming other ensemble learning (EL) methods in 100% of the datasets in precision, 78% in recall, 69.5% in f1−score, and 78% in accuracy. A t-test analysis yielded results of p-value <0.05, indicating a statistically significant improvement in the accuracy of the MOGA over the base models. Moreover, the framework's application can be extended to regression tasks as wellItem Application of supervised machine learning methods on the multidimensional knapsack problem(Springer, 2021) Rezoug, Abdellah; Bader-el-den, Mohamed; Boughaci, DalilaMachine Learning (ML) has gained much importance in recent years as many of its effective applications are involved in different fields, healthcare, banking, trading, gaming, etc. Similarly, Combinatorial Optimisation (CO) keeps challenging researchers by new problems with more complex constraints. Merging both fields opens new horizons for development in many areas. This study investigates how effective is to solve CO problems by ML methods. The work considers the Multidimensional Knapsack Problem (MKP) as a study case, which is an np-hard CO problem well-known for its multiple applications. The proposed approach suggests to use solutions of small-size MKP to build models with different ML methods; then, to apply the obtained models on large-size MKP to predict their solutions. The features consist of scores calculated based on information about items while the labels consist of decision variables of optimal solutions calculated from applying CPLEX Solver on small-size MKP. Supervised ML methods build models that help to predict structures of large-size MKP solutions and build them accordingly. A comparison of five ML methods is conducted on standard data set. The experiments showed that the tested methods were able to reach encouraging results. In addition, the study proposes a Genetic Algorithm (GA) that exploits ML outputs essentially in initialisation operator and to repair unfeasible solutions. The algorithm denoted GaPR explores the ML solution neighbourhood as a way of intensification to approach optimal solutions. The carried out experiments indicated that the approach was effective and competitiveItem Optimisation of fuzzy logic quadrotor attitude controller–particle swarm, cuckoo search and BAT algorithms(Taylor & Francis, 2021) Zatout, Mohamed Siddiq; Rezoug, Amar; Rezoug, Abdellah; Baizid, Khalifa; Iqbal, JamshedBio-inspired optimisation algorithms have recently attracted much attention in the control community. Most of these algorithms mimic particular behaviours of some animal species in such a way that allows solving optimisation problems. The present paper aims at applying three metaheuristic methods for optimising fuzzy logic controllers used for quadrotor attitude stabilisation. The investigated methods are particle swarm optimisation (PSO), BAT algorithm and cuckoo search (CS). These methods are applied to find the best output distribution of singleton membership functions of the fuzzy controllers. The quadrotor control requires measured responses, therefore, three objective functions are considered: integral squared error, integral time-weighted absolute error and integral time-squared error. These metrics allow performance comparison of the controllers in terms of tracking errors and speed of convergence. The simulation results indicate that BAT algorithm demonstrated higher performance than both PSO and CS. Furthermore, BAT algorithm is capable of offering 50% less computation time than CS and 10% less time than PSO. In terms of fitness, BAT algorithm achieved an average of 5% better fitness than PSO and 15% better than CS. According to these results, the BAT-based fuzzy controller exhibits superior performance compared with other algorithms to stabilise the quadrotorItem Guided genetic algorithm for the multidimensional knapsack problem(Springer, 2017) Rezoug, Abdellah; Bader-El-Den, Mohamed; Boughaci, DalilaItem Memetic Algorithm for Solving the 0-1 Multidimensional Knapsack Problem(SPRINGER, 2015) Rezoug, Abdellah; Boughaci, Dalila; Badr-El-Den, MohamedIn this paper, we propose a memetic algorithm for the Multidimensional Knapsack Problem (MKP). First, we propose to combine a genetic algorithm with a stochastic local search (GA-SLS), then with a simulated annealing (GA-SA). The two proposed versions of our approach (GA-SLS and GA-SA) are implemented and evaluated on benchmarks to measure their performance. The experiments show that both GA-SLS and GA-SA are able to find competitive results compared to other well-known hybrid GA based approaches.Item Fuzzy logic controller for a pneumatic artificial muscle robot based on sliding mode control(2009) Rezoug, Abdellah; Meddahi, A.; Baizid, K.; Hamerlain, m.; Tadjine, M.Fuzzy Logic Control (FLC) has been successfully established in control systems engineering in the recent years, in other hand, Sliding Mode Control (SMC) is an active area in control research. The combination of this tow fields called Fuzzy Sliding Mode Control (FSMC) techniques to exploit the superior sides of these two controllers have drawn the attention of the scientific community. In this work, we proposed fuzzy logic controller based on the sliding mode theory to control the robot arm actuated by the pneumatics artificial muscles. Using bang-bang motion generation law, the objective of the control is the position and the velocity tracking by the robot. Simulations results demonstrate the feasibility and the advantages of our proposed research work.