Mohamed El Haziti’s research while affiliated with Mohammed V University and other places

Rapid urbanization poses several challenges, especially when faced with an uncontrolled urban development plan. Therefore, it often leads to anarchic occupation and expansion of cities, resulting in the phenomenon of urban sprawl (US). To support sustainable decision–making in urban planning and policy development, a more effective approach to addressing this issue through US simulation and prediction is essential. Despite the work published in the literature on the use of deep learning (DL) methods to simulate US indicators, almost no work has been published to assess what has already been done, the potential, the issues, and the challenges ahead. By synthesising existing research, we aim to assess the current landscape of the use of DL in modelling US. This article elucidates the complexities of US, focusing on its multifaceted challenges and implications. Through an examination of DL methodologies, we aim to highlight their effectiveness in capturing the complex spatial patterns and relationships associated with US. This work begins by demystifying US, highlighting its multifaceted challenges. In addition, the article examines the synergy between DL and conventional methods, highlighting the advantages and disadvantages. It emerges that the use of DL in the simulation and forecasting of US indicators is increasing, and its potential is very promising for guiding strategic decisions to control and mitigate this phenomenon. Of course, this is not without major challenges, both in terms of data and models and in terms of strategic city planning policies.

As the demand for efficient and lightweight models in image classification grows, knowledge distillation has emerged as a promising technique to transfer expertise from complex teacher models to simpler student models. However, the efficacy of knowledge distillation is intricately linked to the choice of datasets used during training. Datasets are pivotal in shaping a model’s learning process, influencing its ability to generalize and discriminate between diverse patterns. While considerable research has independently explored knowledge distillation and image classification, a comprehensive understanding of how different datasets impact knowledge distillation remains a critical gap. This study systematically investigates the impact of diverse datasets on knowledge distillation in image classification. By varying dataset characteristics such as size, domain specificity, and inherent biases, we aim to unravel the nuanced relationship between datasets and the efficacy of knowledge transfer. Our experiments employ a range of datasets to comprehensively explore their impact on the performance gains achieved through knowledge distillation. This study contributes valuable guidance for researchers and practitioners seeking to optimize image classification models through kno-featured applications. By elucidating the intricate interplay between dataset characteristics and knowledge distillation outcomes, our findings empower the community to make informed decisions when selecting datasets, ultimately advancing the field toward more robust and efficient model development.

This paper presents a comparative analysis of dif- ferent datasets’ influence on the effectiveness of response-based knowledge distillation in image classification. Through a series of experiments and evaluations, we explore the impact of dataset choice on the performance and transferability of distilled models. Our findings provide insights into optimizing dataset selection for response-based knowledge distillation, a critical aspect of model compression and transfer learning.

Urban sprawl, a common phenomenon characterized by uncontrolled urban growth, has far-reaching socio-economic and environmental implications. It’s a complex phenomenon, and finding a better way to tackle it is essential. Accurate simulation and prediction of urban sprawl features would facilitate decision-making in urban planning and the formulation of city growth policies. This article provides an overview of the techniques used to this end. Initially, it highlights the use of a certain category of so-called traditional methods, such as statistical models or classical machine learning methods. It then focuses particularly on the intersection of deep learning and urban sprawl modelling, examining how deep learning methods are being exploited to simulate and predict urban sprawl. I finally studies hybrid approaches that combine deep learning with agent-based models, cellular automata, or other techniques offer a synergistic way to leverage the strengths of different methodologies for urban sprawl modelling.

Identifying influential nodes in a network is a major issue due to the great deal of applications concerned, such as disease spreading and rumor dynamics. That is why, a plethora of centrality measures has emerged over the years in order to rank nodes according to their topological importance in the network. Local metrics such as degree centrality make use of a very limited information and are easy to compute. Global metrics such as betweenness centrality exploit the information of the whole network structure at the cost of a very high computational complexity. Recent works have shown that combining multiple metrics is a promising strategy to quantify the node's influential ability. Our work is in this line. In this paper, we introduce a multi-attributes centrality measure called M-Centrality that combines the information on the position of the node in the network with the local information on its nearest neighborhood. The position is measured by the K-shell decomposition, and the degree variation in the neighborhood of the node quantifies the influence of the local context. In order to examine the performances of the proposed measure, we conduct experiments on small and large scale real-world networks from the perspectives of transmission dynamics and network connectivity. According to the empirical results, the M-Centrality outperforms its alternatives in identifying both influential spreaders and nodes essential to maintain the network connectivity. In addition, its low computational complexity makes it easily applied to large scale networks.

In this paper, a robust hybrid watermarking method based on discrete wavelet transform (DWT), discrete cosine transform (DCT), and scale-invariant feature transformation (SIFT) is proposed. Indeed, it is of prime interest to develop robust feature-based image watermarking schemes to withstand both image processing attacks and geometric distortions while preserving good imperceptibility. To this end, a robust watermark is embedded in the DWT-DCT domain to withstand image processing manipulations, while SIFT is used to protect the watermark from geometric attacks. First, the watermark is embedded in the middle band of the discrete cosine transform (DCT) coefficients of the HL1 band of the discrete wavelet transform (DWT). Then, the SIFT feature points are registered to be used in the extraction process to correct the geometric transformations. Extensive experiments have been conducted to assess the effectiveness of the proposed scheme. The results demonstrate its high robustness against standard image processing attacks and geometric manipulations while preserving a high imperceptibility. Furthermore, it compares favorably with alternative methods.

Minimizing the financial cost of sensors network deployment while ensuring a good coverage of the measured variable have become the main priority in network topology optimization. In this work, the problem is considered of optimization of air quality network for pollution monitoring in Dunkirk in France. Modelized data of PM10 by Atmospheric Dispersion Modelling System (ADMS) is taken as a ground truth. In this study, we used Inverse Distance Weighting (IDW) interpolation to compute the cost function of Root Mean Square Error (RMSE). Pollution rates in the region is estimated by interpolation from the positions of measurement stations. The Root Mean Square Error between interpolated and the ground truth pollution values reflect the quality of the network. The position of pollution sensors is optimized by means of the Genetic Algorithm to minimize the RMSE. A random population is chosen as initial input to this algorithm, afterwards by means of crossover and mutation techniques it tries to generate populations that would produce individuals achieving a smaller (RMSE). In addition to optimal station positioning, this approach also allows estimating how RMSE of pollution estimation depends on the number of measuring stations. We observe the optimization precision dependence on the sensors number. The obtained configuration was analyzed and compared with the actual national measurement network of ATMO Hauts-de-France. As for this deployed national network, that have most of its sensors poisoned next to the emission sources (most industrial), the optimal topology resulted by genetic algorithm has only one coastal sensor (next to the industrial zone), and the rest are distributed all over the studied zone. Besides to sensors found to be in the sea part of the city. This obtained topology could be explained by the background pollution contribution in the atmospheric pollution of the city.

Nowadays information and communication technologies are playing a decisive role in helping the financial institutions to deal with the management of credit risk. There have been significant advances in scorecard model for credit risk management. Practitioners and policy makers have invested in implementing and exploring a variety of new models individually. Coordinating and sharing information groups, however, achieved less progress. One of several causes of the 2008 financial crisis was in data architecture and information technology infrastructure. To remedy this problem the Basel Committee on Banking Supervision (BCBS) outlined a set of principles called BCBS 239. Using Ontology Design Patterns (ODPs) and BCBS 239, credit risk scorecard and applicant ontologies are proposed to improve the decision making process in credit loan. Both ontologies were validated, distributed in Ontology Web Language (OWL) files and checked in the test cases using SPARQL. Thus, making their (re)usability and expandability easier in financial institutions. These ontologies will also make sharing data more effective and less costly.

Air quality models (AQMs) are useful for studying various types of air pollutions and provide the possibility to reveal the contributors of air pollutants. Existing AQMs have been used in many scenarios having a variety of goals, e.g., focusing on some study areas and specific spatial units. Previous AQM reviews typically cover one of the forming elements of AQMs. In this review, we identify the role and relevance of every component for building AQMs, including (1) the existing techniques for building AQMs, (2) how the availability of the various types of datasets affects the performance, and (3) common validation methods. We present recommendations for building an AQM depending on the goal and the available datasets, pointing out their limitations and potentials. Based on more than 40 works on air quality, we concluded that the main utilized methods in air pollution estimation are land-use regression (LUR), machine learning, and hybrid methods. In addition, when incorporating LUR methods with traffic variables, it gives promising results; however, when using kriging or inverse distance weighting techniques, the monitoring stations measurements of air pollution data are enough to have good results. We aim to provide a short manual for people who want to build an AQM given the constraints at hands such as the availability of datasets and technical/computing resources.

span lang="EN-US">This paper proposes an ontological scorecard model for credit risk management. The purpose of credit scoring model is to reduce the possibility of potential losses with regard to issued loans. Loans are provided according to strict criteria which contain information about the client, loan structure, the purpose, repayment source and collateral. Several techniques have been used for credit risk assessment before granting a loan. Ontology design patterns is used here to enable the implementation of domain knowledge using the OWL rules and to improve the decision making process in credit monitoring. The modeling of our ontology will make the data publication simpler and graph structures intuitive, thus making its reusability and expandability easier.</span