Régis Audran Mogo Wafo, Thomas Messi Nguélé, Armel Jacques Nzekon Nzeko'o, Djam Youh Xaviera
<div><p>Graph analysis applications are increasingly used on large-scale datasets, where memory access patterns have a significant impact on performance. In such applications, poor data locality leads to a high number of cache misses, which in turn degrades execution time. To address this issue, several graph numbering techniques have been proposed to reorganize data in memory, mainly based on structural properties of the graph. More recently, execution trace-based approaches such as NumBaClus have been introduced, leveraging clustering techniques to group nodes with similar access patterns. However, these methods do not explicitly capture the co-occurrence of nodes within the same execution contexts. In this paper, we propose NumBaFrIt, a new graph numbering approach based on frequent itemset mining of execution traces. By modeling execution traces as transactional data, our method identifies groups of nodes that are frequently accessed together and reorganizes them in memory to improve data locality. Experimental results show that NumBaFrIt improves base numbering and significantly enhances existing ordering strategies when used as a preprocessing step, leading to better cache efficiency and reduced execution time. This is for example the case on user machine with 3072 KB of cache memory, our proposal got the best cache misses reduction through the combination NumBaFrit-sup70_cn-order with 48.96% compared to 17.15% of cache misses reduction gotten with Cn-order_cl-h (an existing numbering of NumBaClus).</p></div>
The commonly employed test for detecting change-point are quite generally based on asymptotic approximations which may not provide good size control in finite samples (Diebold et Chen (1996)). There has been a number of recent studies (BAI and PERRON, 1998, SHAO and ZHANG, 2010) that seek to improve the reliability of common tests for change-point detection using EDGEWORTH and bootstrap methods. Yet the latter remain approximate.In this companion paper, we propose two tests for change-point detection in linear regression contexts. The first is a Monte Carlo (MC) test for one change-point. The second is a new sequential test for detecting one or more change-point in time series. The finite-sample behaviour of the proposed tests is investigated through Monte Carlo experiments. The results show that tests which are widely used in the studies dealing with climatic changes in West and Central Africa exhibit important size distortions, while MC test achieve complete size control and display good power.
Program Increment (PI) Planning is a critical synchronization event in scaled Agile methodologies, yet remains largely informal, with limited analytical rigor or strategic alignment. This paper introduces a formal metamodel and a model-driven framework to support structured PI Planning in large-scale software development. Our metamodel captures essential planning artifacts, including features, teams, iterations, risks, and dependencies, and is enriched with OCL constraints for validation and traceability. We extend the model to align planning with business value streams and introduce stakeholder-specific viewpoints for tailored analysis. Compared to tools like Jira or Mural, our approach provides formal semantics, automated reasoning, and extensibility. This work lays the foundation for a Domain-Specific Language (DSL) for PI Planning, bridging the gap between Agile practice and formal modeling.
This study evaluates and compares the performance of two prominent fingerprint recognition methods: the Generalized Hough Transform (GHT) and Siamese neural networks. Using the diverse and well suited CASIA database, we conducted an exhaustive comparison on a representative sample of 628 fingerprints per model.Siamese neural networks demonstrated exceptional accuracy, surpassing 99.2\%. Their efficiency in processing large datasets highlights their suitability for biometric applications requiring detailed analysis. In contrast, the Generalized Hough Transform achieved a strong accuracy of 97.7\%, proving its reliability even with more constrained data.Our analysis underscores the distinct strengths of each method, particularly the adaptability of Siamese networks to complex fingerprint variations. Their robust generalization capabilities make them ideal for large-scale biometric systems.Performance was assessed using key metrics, including the confusion matrix (visualizing predicted vs. actual matches) and the F1-score (balancing precision and recall).These findings advance our understanding of biometric security challenges and suggest promising directions for future research in this rapidly evolving field.
The objective of this study is to propose solutions for predicting drinking water consumption over aperiod of 5 days. To achieve this, machine learning methods such as neural network algorithms areused to confront different results in order to obtain a model that optimizes daily water consumptionthe most. The algorithms used include Multi-Layer Perceptron (MLP), Simple Recurrent NeuralNetwork, Long Short-Term Memory (LSTM) recurrent neural network, and Gated Recurrent Unit(GRU) recurrent neural network. Since the data comes from multiple sources, the algorithms areused to train models on each dataset. At the end of this study, the model that optimizes the best isselected for each dataset after comparing the results of different algorithms. Based on the variousresults obtained, the performance of the algorithms varies depending on the periods and the data.For example, the best performance in predicting consumption in the KUA area on the J+5 dayperiod is attributed to the MLP algorithm based on mean absolute errors (MAE).
NoSQL data stores have introduced a new way of designing database systems to meet the recent needs of applications and services operating in areas such as the World Wide Web, Big Data, and Data Analytics. They offer a means to store and access high volumes of partially structured data by enhancing the flexibility of the data model and integrating distributed architecture at their core, thus providing better properties of high data availability and low data latency.This paper reviews the various design approaches of NoSQL data stores, providing up-to-date information on their data models, request processing, scalability, storage management, data dis- tribution modes, and use cases. It also addresses multi-model and cloud-oriented NoSQL stores and offer a comprehensive description of a wide range of NoSQL stores with the use of a rich taxonomy.
Gérard Nzebop Ndenoka, Maurice Tchuente, Emmanuel Simeu, Valery Monthe
GRAFCET is a powerful graphical modeling language for the specification of controllers in discrete event systems. It considers hierarchical structures as well as structural and semantic constraints. In this paper, we propose to use a GRAFCET specification model in a Model Driven Engineering (MDE) approach for multi-target synthesis of embedded logic control systems based on microcontrollers. In this approach, a GRAFCET metamodel is associated with a microcontroller metamodel which characterizes the microcontroller platform features to be considered when generating code. The GRAFCET metamodel includes the modeling of expressions to facilitate model verification and an easy interpretation of Grafcet events and time constraints. Transformation rules for generation of C-programmable microcontroller code are then presented. As application, we present a platform based on Eclipse EMF, Object Constraint Language (OCL) and Acceleo code generation engine.
Armel Jacques Nzekon Nzeko'o, Hamza Adamou, Thomas Messi Nguele, Bleriot Pagnaul Betndam Tchamba
Nowadays, e-commerce, streaming and social networks platforms play an important role in our daily lives. However, the ever-increasing addition of items on these platforms (items on Amazon, videos on Netflix and YouTube, posts on Facebook and Instagram) makes it difficult for users to select items that interest them. The integration of recommender systems into these platforms aims to offer each user a small list of items that match their preferences. To improve the performance of these recommender systems, some work in the literature incorporate explicit or implicit trust between platform users through trust-based recommender systems. Indeed, many of these works are based on explicit trust, when each user designates those whom they trust in the platform. But this information is rare in most real-world platforms. Thus, other work propose to estimate the implicit trust that each user can grant to another. However, work that estimates implicit trust does not take into account the temporal dynamics of users' past following actions and even less the fact that a user can influence another on one category of item and not on another. In this paper, we propose time and content aware strategies to estimate social influence of one user on another. The resulting time and content aware implicit trust are integrated to trust-based recommender systems build on K-Nearest Neighbors (KNN) and Graph-based techniques. Experiments done for rating predictions with KNN and Top-N recommendations with Graph model show that time and content aware implicit trust make it possible to improve the performance of the KNN according to the RMSE metric by 7% and 10%, and the performance of the graph model according to the NDCG@10 metric by 59% and 08% respectively on the Ciao and Epinions datasets.
Régis Audran Mogo Wafo, Thomas Messi Nguelé, Armel Jacques Nzekon Nzeko'o, Djam Youh Xaviera
Social graph analysis is generally based on a local exploration of the underlying graph. That is, the analysis of a node of the graph is often done after having analyzed nodes located in its vicinity. However, over the time, networks are bound to grow with the addition of new members, which inevitably leads to the enlargement of the corresponding graphs. At this level we therefore have a problem because more the size of the graph increases, more the execution time of graph analysis applications too. This is due to the very large number of nodes that will need to be treated. Some recent work in-faces this problem by exploiting the properties of social networks such as the community structure to renumber the nodes of the graph in order to reduce cache misses. Reducing cache misses in an application allows to reduce the execution time of this application. In this paper, we argue that combining existing graph ordering with a new numbering that exploit execution traces analysis can allow to improve cache misses reduction and hence execution time reduction. The idea is to build graph numbering using execution traces of graph analysis applications and then combine it with an existing graph numbering (such as cn-order). To build this new ordering, we define a new distance and then used it to analyse execution traces with well known clustering algorithms K-means (for Kmeans-order) and hierarchical clustering (for cl-hier-order). Experiments on a user machine (dual-core) and four cores of Grid'5000 node (Neowise) show that this combination improves slightly existing graph ordering (cn-order, numbaco, rabbit and gorder) in almost all the cases (the two cores of dual-core, all the four cores of neowise), with PageRank graph application and astro-ph dataset. For example, on neowise with one thread and Astro-ph dataset, the best performance is given with the combination kmeans-order_cn-order which allows to reduce by 42.59% the cache misses (compared to the second numbaco with 40.79%) and therefore by 7.27 % the time of execution (compared to 6.89% for the second numbaco).
Thomas Messi Nguelé, Armel Jacques Nzekon Nzeko'o, Damase Donald Onana
Recent work has shown that deep learning algorithms are efficient for various tasks, whether in Natural Language Processing (NLP) or in Computer Vision (CV). One of the particularities of these algorithms is that they are so efficient as the amount of data used is large. However, sequential execution of these algorithms on large amounts of data can take a very long time. In this paper, we consider the problem of training Recurrent Neural Network (RNN) for hate (aggressive) messages detection task. We first compared the sequential execution of three variants of RNN, we have shown that Long Short Time Memory (LSTM) provides better metric performance, but implies more important execution time in comparison with Gated Recurrent Unit (GRU) and standard RNN. To have both good metric performance and reduced execution time, we proceeded to a parallel implementation of the training algorithms. We proposed a parallel algorithm based on an implicit aggregation strategy in comparison to the existing approach which is based on a strategy with an aggregation function. We have shown that the convergence of this proposed parallel algorithm is close to that of the sequential algorithm. The experimental results on an 32-core machine at 1.5 GHz and 62 Go of RAM show that better results are obtained with the parallelization strategy that we proposed. For example, with an LSTM on a dataset having more than 100k comments, we obtained an f-measure of 0.922 and a speedup of 7 with our approach, compared to a f-measure of 0.874 and a speedup of 5 with an explicit aggregation between workers.
Under-resourced languages encounter substantial obstacles in speech recognition owing to the scarcity of resources and limited data availability, which impedes their development and widespread adoption. This paper presents a representation learning model that leverages existing frameworks based on self-supervised learning techniques—specifically, Contrastive Predictive Coding (CPC), wav2vec, and a bidirectional variant of CPC—by integrating them with multilingual learning approaches. We apply this model to three African languages: Wolof, Swahili, and Fongbe. Our evaluation of the resulting representations in a downstream task, automatic speech recognition, utilizing an architecture analogous to DeepSpeech, reveals the model’s capacity to discern language specific linguistic features. The results demonstrate promising performance, achieving Word Error Rates (WER) of 61% for Fongbe, 72% for Wolof, and 88% for Swahili. These findings underscore the potential of our approach in advancing speech recognition capabilities for under-resourced languages, particularly within the African linguistic landscape.
In the Middle Ages, some illuminations were intended for the elites of society and served as a means of communication for them to extend their social influences and to represent their social environments. They constitute an information system based on symbolic components linked together by semantic and influential relationships whose structure is close to models representing social relationships and networks. Today, understanding these illuminations and extracting their implicit messages, expressed through the combination of metaphorical graphic elements, are a difficult task reserved for experts. To help these latter and address the semantic heterogeneity of illuminations, this article explores the synergy between knowledge representation techniques and the analysis of medieval documents to build a knowledge model describing these medieval paintings. It proposes a formal ontology composed of items describing the explicit and visible knowledge of medieval illuminations and others expressing their implicit messages. The considered illuminations are part of those ordered or linked to the Burgundy duke, Philiphe le Bon
In this paper, we propose a hybrid approach for solving the symmetric traveling salesman problem. The proposed approach combines the ant colony algorithm (ACO) with neural networks based on the attention mechanism. The idea is to use the predictive capacity of neural networks to guide the behaviour of ants in choosing the next cities to visit and to use the prediction results of the latter to update the pheromone matrix, thereby improving the quality of the solutions obtained. In concrete terms, attention is focused on the most promising cities by taking into account both distance and pheromone information thanks to the attention mechanism, which makes it possible to assign weights to each city according to its degree of relevance. These weights are then used to predict the next towns to visit for each city. Experimental results on instancesTSP from the TSPLIB library demonstrate that this hybrid approach is better compared to the classic ACO.
Fabien Moffo, Auguste Noumsi Woguia, Joseph Mvogo Ngono, Samuel Bowong
Cough is the most common symptom of lung disease. COVID-19, a respiratory illness, has caused over 700 million positive cases and 7 million deaths worldwide. An effective, affordable, and widely available diagnostic tool is crucial in combating lung disease and the COVID-19 pandemic. Deep learning and machine learning algorithms could be used to analyze the cough sounds of infected patients and make predictions. Our research lab and the COUGHVID research lab provide the cough data. This diagnostic approach can distinguish cough sounds from COVID-19 patients and people suffering from other ailments as well as healthy people using deep learning and feature extraction from Mel spectrograms. The model used is a variant of ConvNet. This ConvNet model can easily capture features in MFCC Vectors and enable convolution parallelism, which increases processing speed. ConvNet attains translational invariance in features through the sharing of weights between layers. During data acquisition for model training, it is important to consider quiet environments to reduce errors in audio quality. The architecture of the convolutional neural networks gives an F1-score of 89%, an accuracy of 90.33% and sensitivity of 87.3%. This system has the potential to significantly impact society by reducing virus transmission, expediting patient treatment, and freeing up hospital resources. Early detection of COVID-19 can prevent disease progression and enhance screening effectiveness.
Juvet Karnel Sadie, Leonel Moyou Metcheka, René Ndoundam
Text steganography is a mechanism of hiding secret text message inside another text as a covering message. In this paper, we propose a text steganographic scheme based on color coding. This includes two different methods: the first based on permutation, and the second based on numeration systems. Given a secret message and a cover text, the proposed schemes embed the secret message in the cover text by making it colored. The stego-text is then send to the receiver by mail. After experiments, the results obtained show that our models perform a better hiding process in terms of hiding capacity as compared to the scheme of Aruna Malik et al. on which our idea is based.
Nzeuhang Yannick Yomie, Yonta Paulin Melatagia, Lecouteux Benjamin
Recently popularized self-supervised models appear as a solution to the problem of low data availability via parsimonious learning transfer. We investigate the effectiveness of these multilingual acoustic models, in this case wav2vec 2.0 XLSR-53 and wav2vec 2.0 XLSR-128, for the transcription task of the Ewondo language (spoken in Cameroon). The experiments were conducted on 11 minutes of speech constructed from 103 read sentences. Despite a strong generalization capacity of multilingual acoustic model, preliminary results show that the distance between XLSR embedded languages (English, French, Spanish, German, Mandarin, . . . ) and Ewondo strongly impacts the performance of the transcription model. The highest performances obtained are around 69% on the WER and 28.1% on the CER. An analysis of these preliminary results is carried out andthen interpreted; in order to ultimately propose effective ways of improvement.
Six time series related to atmospheric phenomena are used as inputs for experiments offorecasting with singular spectrum analysis (SSA). Existing methods for SSA parametersselection are compared throughout their forecasting accuracy relatively to an optimal aposteriori selection and to a naive forecasting methods. The comparison shows that awidespread practice of selecting longer windows leads often to poorer predictions. It alsoconfirms that the choices of the window length and of the grouping are essential. Withthe mean error of rainfall forecasting below 1.5%, SSA appears as a viable alternative forhorizons beyond two weeks.
Lot sizing is important in production planning. It consists of determining a production plan that meets the orders and other constraints while minimizing the production cost. Here, we consider a Discrete Lot Sizing and Scheduling Problem (DLSP), specifically the Pigment Sequencing Problem (PSP). We have developed a solution that uses Genetic Algorithms to tackle the PSP. Our approach introduces adaptive techniques for each step of the genetic algorithm, including initialization, selection, crossover, and mutation. We conducted a series of experiments to assess the performance of our approach across some multiple trials using publicly available instances of the PSP. Our experimental results demonstrate that Genetic Algorithms are practical and effective approaches for solving DLSP.
This article proposes a method for extracting knowledge in association rules using the classical measure of implication intensity. We then applied our method to data from mathematics didactics studies. The aim of the didactic study was to identify the relationships between students' difficulties and skills when demonstrating a mathematical proposition formulated in French. The results of our study show that our methodology is effective in extracting interesting rules. In addition, the results of our didactic analysis showed the dependency between understanding a mathematical statement in French, competence in translating it formally and proving it.
The coffee berry borer (CBB) Hypothenemus hampei (Coleoptera: Scolytidae) is the most important insect pest affecting coffee production worldwide and generating huge economic losses. As most of its life cycle occurs inside the coffee berry, its control is extremely difficult. To tackle this issue, we solve an optimal control problem based on a berry age-structured dynamical model that describes the infestation dynamics of coffee berries by CBB during a cropping season. This problem consists in applying a bio-insecticide at discrete times in order to maximise the economic profit of healthy coffee berries, while minimising the CBB population for the next cropping season. We derive analytically the first-order necessary optimality conditions of the control problem. Numerical simulations are provided to illustrate the effectiveness of the optimal control strategy.
