Diabetic retinopathy is a severe retinal disease that can blur or distort the vision of the patient. It is one of the leading causes of blindness. Early detection of diabetic retinopathy can significantly help in the treatment. The recent development in the field of AI and especially Deep learning provides ambitious solutions that can be exploited to predict, forecast and diagnose several diseases in their early phases. This work aims towards finding an automatic way to classify a given set of retina images in order to detect the diabetic retinopathy. Deep learning concepts have been used with a convolutional neural network (CNN) algorithm to build a multi-classification model that can detect and classify disease levels automatically. In this study, a CNN architecture has been applied with several parameters on a dataset of diabetic retinopathy with different structures. At the current stage of this work, obtained results are highly encouraging.

Reconfigurable manufacturing system (RMS) is a recent manufacturing paradigm, which can easily adjust its capacity and functionality for rapid responsiveness to sudden changes in the market. The core component of RMS is called reconfigurable machine tool (RMT), which has a modular structure. The RMTs can be reconfigured into many configurations. This ability allows RMS to manufacture many types of products with high quantities. In this paper, the scheduling of products in a multi-product line is fulfilled based on three criteria: profit over cost, due date, and reconfiguration responsiveness index. The latter is the combination of reconfiguration time and reconfiguration reliability of machines. An integrated approach of maximum deviation method (MDM) and multi-criteria decision-making (MCDM) approach called technique for order preference by similarity to ideal solution (TOPSIS) is proposed as a solution approach for getting the optimal scheduling of the products to be manufactured in RMS. Weights of criteria have been calculated using MDM and ranking of products is obtained using TOPSIS. A numerical example is presented to illustrate the scheduling of products in RMS.

This paper aims to examine the effects of aging on the efficiency of facial recognition algorithms in terms of age groups and age difference intervals. A comparative analysis of the recognition performance of two approaches is conducted for different age groups and different length time intervals between images. The first approach uses a two-dimensional discrete cosine transform (2D-DCT) and a Kernel Fisher Analysis (KFA) as description tools; classification is made using a k-NN classifier based on Euclidean distance. However, the second one is performed in two ways: first, we considered face as a single entity, then we viewed face as an independent component set. This approach makes use of Convolutional Neural Networks (CNN) for description and Support vector machines (SVM) for classification. Achieved results using the publicly accessible FG-NET face database prove that age groups influence the performance of face recognition algorithms. Also, time length lapses between images can significantly reduce the performance of face recognition.

Multi-objective optimization evolutionary algorithms (MOEAs) are considered as the most suitable heuristic methods for solving multi-objective optimization problems (MOPs). These MOEAs aim to search for a uniformly distributed, near-optimal and near-complete Pareto front for a given MOP. However, MOEAs fail to achieve their aim completely because of their fixed population size. To overcome this limit, an evolutionary approach of multi-objective optimization was proposed; the dynamic multi-objective evolutionary algorithms (DMOEAs). This paper deals with improving the user requirements (i.e., getting a set of optimal solutions in minimum computational time). Although, DMOEA has the distinction of dynamic population size, being an evolutionary algorithm means that it will certainly be characterized by long execution time. One of the main reasons for adapting parallel evolutionary algorithms (PEAs) is to obtain efficient results with an execution time much lower than the one of their sequential counterparts in order to tackle more complex problems. Thus, we propose a parallel version of DMOEA (i.e., PDMOEA). As experimental results, the proposed PDMOEA enhances DMOEA in terms of three criteria: improving the objective space, minimization of computational time and converging to the desired population size.

The fifth-generation (5G) standard is the last telecommunication technology, widely considered to have the most important characteristics in the future network industry. The 5G system infrastructure contains three principle interfaces, each one follows a set of protocols defined by the 3rd Generation Partnership Project group (3GPP). For the next generation network, 3GPP specified two authentication methods systematized in two protocols namely 5G Authentication and Key Agreement (5G-AKA) and Extensible Authentication Protocol (EAP). Such protocols are provided to ensure the authentication between system entities. These two protocols are critical systems, thus their reliability and correctness must be guaranteed. In this paper, we aim to formally re-examine 5G-AKA protocol using micro Common Representation Language 2 (mCRL2) language to verify such a security protocol. The mCRL2 language and its associated toolset are formal tools used for modeling, validation, and verification of concurrent systems and protocols. In this context, the authentication protocol 5G-AKA model is built using Algebra of Communication Processes (ACP), its properties are specified using Modal mu-Calculus and the properties analysis exploits Model-Checker provided with mCRL2. Indeed, we propose a new mCRL2 model of 3GPP specification considering 5G-AKA protocol and we specify some properties that describe necessary requirements to evaluate the correctness of the protocol where the parsed properties of Deadlock Freedom, Reachability, Liveness and Safety are positively assessed.

Formal methods are widely exploited in the performance evaluation of Wireless Sensor Networks (WSNs) protocols and algorithms. These methods help researchers to model and to analyse mathematically such protocols. Numerical results obtained by analysis and performance evaluation can be employed to prove the correctness and consistency of the designed models. However, these methods face a scalability problem when the number of components becomes very high, which is often the case in WSNs. To overcome this challenge, this paper proposes to use a Machine Learning (ML) solution to provide predictions when the number of nodes increases and the formal model becomes enable to make the analysis. Indeed, this work deals with the application of effective Artificial Neural Networks (ANNs) for the prediction of a set of crucial performance metrics of CSMA/CA-MAC protocol in WSNs when the number of nodes increases significantly in the network. This prediction process is based on prior results obtained by the formal model when the number of nodes was manageable by that formal model.

Indexing stage in information retrieval process has a great importance as an essential tool for the performance of recall and precision. Despite the many studies that have been done on the indexing conducted in the last few decades, to our knowledge, no study has investigated whether indexing real-time based on keywords extraction is efficient to perform of recall and precision. Moreover, relatively fewer Arabic text indexing studies are currently available despite the enormous efforts put together to satisfy the needs of the growing number of Arabic internet users. This paper suggests a method for Arabic text indexing based on keywords extraction. The proposed method consists of two stages. The first stage conducts a real-time indexing. The second stage is a keywords extraction and updating of initial index taking into account the output of keywords extraction process. We illustrate application and the performance of this method of indexing using an Arabic text editor (SIRAT) developed and designed for this aim. We also illustrate the process of building a new form of Arabic corpus appropriate to conduct the necessary experiments. Our findings show that SIRAT successfully identifies the keywords most relevant to the document. Finally, the main contribution of this experiment is to demonstrate the effectiveness of this method compared to other methods. In addition, the paper proposes a solution to issues and deficiencies Arabic language processing suffers from, especially regarding corpora building and keywords extraction evaluation systems.

This paper presents a teaching-learning-based optimization algorithm for discrete large-scale multi-objective problems (DLM-TLBO). Unlike the previous variants, the learning strategy used by each individual and the acquired knowledge are defined based on its level. The proposed approach is used to solve a bi-objective object clustering task (B-OCT) in a swarm robotic system, as a case study. The simple robots have as mission the gathering of a number of objects distributed randomly, while respecting two objectives: maximizing the clustering quality, and minimizing the energy consumed by these robots. The simulation results of the proposed algorithm are compared to those obtained by the well-known algorithm NSGA-II. The results show the superiority of the proposed DLM-TLBO in terms of the quality of the obtained Pareto front approximation and convergence speed.

This article describes how the classical algorithm of shape context (SC) is still unable to capture the part structure of some complex shapes. To overcome this insufficiency, the authors propose a novel shape-based retrieval approach that is called HybMAS-GA using a multi-agent system (MAS) and a genetic algorithm (GA). They define a new distance called approximate distance (AD) to define a SC method by AD, which called approximate distance shape context (ADSC) descriptor. Furthermore, the authors' proposed HybMAS-GA is a star architecture where all shape context agents, N, are directly linked to a coordinator agent. Each retrieval agent must perform either a SC or an ADSC method to obtain a similar shape, started from its own initial configuration of sample points. This combination increases the efficiency of the proposed HybMAS-GA algorithm and ensures its convergence to an optimal images retrieval as it is shown through experimental results.

With the huge increase of large-scale multimedia over Internet, especially images, building Content-Based Image Retrieval (CBIR) systems for large-scale images has become a big challenge. One of the drawbacks associated with CBIR is the very long execution time. In this article, we propose a fast Content-Based Image Retrieval system using Spark (CBIR-S) targeting large-scale images. Our system is composed of two steps. (i) image indexation step, in which we use MapReduce distributed model on Spark in order to speed up the indexation process. We also use a memory-centric distributed storage system, called Tachyon, to enhance the write operation (ii) image retrieving step which we speed up by using a parallel k-Nearest Neighbors (k-NN) search method based on MapReduce model implemented under Apache Spark, in addition to exploiting the cache method of spark framework. We have showed, through a wide set of experiments, the effectiveness of our approach in terms of processing time.

This study concerns a swarm of autonomous reactive mobile robots, qualified of naïve because of their simple constitution, having the mission of gathering objects randomly distributed while respecting two contradictory objectives: maximizing quality of the emergent heap-formation and minimizing energy consumed by aforesaid robots. This problem poses two challenges: it is a multi-objective optimization problem and it is a hard problem. To solve it, one of renowned multi-objective evolutionary algorithms is used. Obtained solution, via a simulation process, unveils a close relationship between behavioral-rules and consumed energy; it represents the sought behavioral model, optimizing the grouping quality and energy consumption. Its reliability is shown by evaluating its robustness, scalability, and flexibility. Also, it is compared with a single-objective behavioral model. Results' analysis proves its high robustness, its superiority in terms of scalability and flexibility, and its longevity measured based on the activity time of the robotic system that it integrates.

The growth of Internet of Thing (IoT) implies the availability of a very large number of services which may be similar or the same, managing the Quality of Service (QoS) helps to differentiate one service from another. The service composition provides the ability to perform complex activities by combining the functionality of several services within a single process. Very few works have presented an adaptive service composition solution managing QoS attributes, moreover in the field of healthcare, which is one of the most difficult and delicate as it concerns the precious human life.In this paper, we will present an adaptive QoS-Aware Service Composition Approach (P-MPGA) based on multi-population genetic algorithm in Fog-IoT healthcare environment. To enhance Cloud-IoT architecture, we introduce a Fog-IoT 5-layared architecture. Secondly, we implement a QoS-Aware Multi-Population Genetic Algorithm (P-MPGA), we considered 12 QoS dimensions, i.e., Availability (A), Cost (C), Documentation (D), Location (L), Memory Resources (M), Precision (P), Reliability (R), Response time (Rt), Reputation (Rp), Security (S), Service Classification (Sc), Success rate (Sr), Throughput (T). Our P-MPGA algorithm implements a smart selection method which allows us to select the right service. Also, P-MPGA implements a monitoring system that monitors services to manage dynamic change of IoT environments. Experimental results show the excellent results of P-MPGA in terms of execution time, average fitness values and execution time / best fitness value ratio despite the increase in population. P-MPGA can quickly achieve a composite service satisfying user’s QoS needs, which makes it suitable for a large scale IoT environment.

In this digital age embracing robotics across various areas of life, especially intellectual ones, have reaped great benefits owing to this modern technology. Therefore, the learning field has not remained unchanged given current evolutions as the schooling conditions have been improved through these smart devices. However, teachers still face some difficulties when choosing pedagogical methods and means for effective language learning for children. Thus, this paper aims to measure the effectiveness of social robots in facilitating children’s learning of a second language (L2). For this purpose, the term L2 learning and its subordinate concepts have been distinguished, and then the different methods of language learning were discussed. The latest research regarding social robots in the educational context was also discussed when reviewing the literature. An experimental study conducted on a sample of children illustrated that the use of the social robot significantly helped them in the L2 learning regarding the assimilation fast, retention, and correct pronunciation of its vocabulary. Finally, this study concludes that the social robot would be a good solution and recommends their widespread use in education given its role in improving the schooling conditions of children.

Association rules are the specific data mining methods aiming to discover explicit relations between the different attributes in a large dataset. However, in reality, several datasets may contain both numeric and categorical attributes. Recently, many meta-heuristic algorithms that mimic the nature are developed for solving continuous problems. This article proposes a new algorithm, DCSA-QAR, for mining quantitative association rules based on crow search algorithm (CSA). To accomplish this, new operators are defined to increase the ability to explore the searching space and ensure the transition from the continuous to the discrete version of CSA. Moreover, a new discretization algorithm is adopted for numerical attributes taking into account dependencies probably that exist between attributes. Finally, to evaluate the performance, DCSA-QAR is compared with particle swarm optimization and mono and multi-objective evolutionary approaches for mining association rules. The results obtained over real-world datasets show the outstanding performance of DCSA-QAR in terms of quality measures.

On one hand, there are many proposed intrusion detection systems (IDSs) in the literature. On the other hand, many studies try to deduce the important features that can best detect attacks. This paper presents a new and an easy-to-implement approach to intrusion detection, named distance sum-based k-nearest neighbors (DS-kNN), which is an improved version of k-NN classifier. Given a data sample to classify, DS-kNN computes the distance sum of the k-nearest neighbors of the data sample in each of the possible classes of the dataset. Then, the data sample is assigned to the class having the smallest sum. The experimental results show that the DS-kNN classifier performs better than the original k-NN algorithm in terms of accuracy, detection rate, false positive, and attacks classification. The authors mainly compare DS-kNN to CANN, but also to SVM, S-NDAE, and DBN. The obtained results also show that the approach is very competitive.

The design and the evaluation of communication protocols in WSNs is a crucial issue. Generally, researchers use simulation methods to evaluate them. However, formal modelling and analysis techniques are an efficient alternative to simulation methods. Indeed, these techniques allow performance evaluation and model verification. In this paper, a formal approach is proposed to modelling and to evaluating the Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) MAC protocol with a star topology. Moreover, the proposed approach deals with some properties that are not stated in most existing works. The approach uses Hierarchical Timed Coloured Petri Nets (HTCPNs) formalism to model the protocol and exploits the CPN-Tools to analyse the generated models. HTCPNs provide timed aspect which facilitates the consideration of time constraints inherent to the CSMA/CA protocol.

The growth of Internet of Thing (IoT) implies the availability of a very large number of services which may be similar or the same, managing the Quality of Service (QoS) helps to differentiate one service from another. The service composition provides the ability to perform complex activities by combining the functionality of several services within a single process. Very few works have presented an adaptive service composition solution managing QoS attributes, moreover in the field of healthcare, which is one of the most difficult and delicate as it concerns the precious human life.In this paper, we will present an adaptive QoS-Aware Service Composition Approach (P-MPGA) based on multi-population genetic algorithm in Fog-IoT healthcare environment. To enhance Cloud-IoT architecture, we introduce a Fog-IoT 5-layared architecture. Secondly, we implement a QoS-Aware Multi-Population Genetic Algorithm (P-MPGA), we considered 12 QoS dimensions, i.e., Availability (A), Cost (C), Documentation (D), Location (L), Memory Resources (M), Precision (P), Reliability (R), Response time (Rt), Reputation (Rp), Security (S), Service Classification (Sc), Success rate (Sr), Throughput (T). Our P-MPGA algorithm implements a smart selection method which allows us to select the right service. Also, P-MPGA implements a monitoring system that monitors services to manage dynamic change of IoT environments. Experimental results show the excellent results of P-MPGA in terms of execution time, average fitness values and execution time / best fitness value ratio despite the increase in population. P-MPGA can quickly achieve a composite service satisfying user’s QoS needs, which makes it suitable for a large scale IoT environment

Designing a good MAC protocol remains a challenge. Such a protocol has to guarantee access to the medium while reducing energy consumption. With the appearance of energy harvesting-wireless sensor networks (EH-WSNs), energy is no longer a problem but the challenge now is that each sensor remains in its energetically sustainable state as much as possible. This paper proposes a formal study of on demand MAC (ODMAC) one of the well-known protocols proposed for EH-WSNs. An analysis through statistical model checking is made where properties that guarantee the protocol's correctness are verified and a performance evaluation of important aspects is achieved.