Search

This paper introduces a novel supervised classification method based on dynamic clustering (DC) and K-nearest neighbor (KNN) learning algorithms, denoted DC-KNN. The aim is to improve the accuracy of a classifier by using a DC method to discover the hidden patterns of the apriori groups of the training set. It provides a partitioning of each group into a predetermined number of subgroups. A new objective function is designed for the DC variant, based on a trade-off between the compactness and separation of all subgroups in the original groups. Moreover, the proposed DC method uses adaptive distances which assign a set of weights to the variables of each cluster, which depend on both their intra-cluster and inter-cluster structure. DC-KNN performs the minimization of a suitable objective function. Next, the KNN algorithm takes into account objects by assigning them to the label of subgroups. Furthermore, the classification step is performed according to two KNN competing algorithms. The proposed strategies have been evaluated using both synthetic data and widely used real datasets from public repositories. The achieved results have confirmed the effectiveness and robustness of the strategy in improving classification accuracy in comparison to alternative approaches. [ABSTRACT FROM AUTHOR]

Accurate segmentation of breast ultrasound (BUS) images is crucial for early diagnosis and treatment of breast cancer. Further, the task of segmenting lesions in BUS images continues to pose significant challenges due to the limitations of convolutional neural networks (CNNs) in capturing long-range dependencies and obtaining global context information. Existing methods relying solely on CNNs have struggled to address these issues. Recently, ConvNeXts have emerged as a promising architecture for CNNs, while transformers have demonstrated outstanding performance in diverse computer vision tasks, including the analysis of medical images. In this paper, we propose a novel breast lesion segmentation network CS-Net that combines the strengths of ConvNeXt and Swin Transformer models to enhance the performance of the U-Net architecture. Our network operates on BUS images and adopts an end-to-end approach to perform segmentation. To address the limitations of CNNs, we design a hybrid encoder that incorporates modified ConvNeXt convolutions and Swin Transformer. Furthermore, to enhance capturing the spatial and channel attention in feature maps we incorporate the Coordinate Attention Module. Second, we design an Encoder-Decoder Features Fusion Module that facilitates the fusion of low-level features from the encoder with high-level semantic features from the decoder during the image reconstruction. Experimental results demonstrate the superiority of our network over state-of-the-art image segmentation methods for BUS lesions segmentation. [ABSTRACT FROM AUTHOR]

The price prediction is valuable in energy management system (EMS) because it allows making informed decisions and solving the problem of the uncertainty related to the future ignorance based only on the past knowledge. To this goal, we present in this paper a two-steps EMS in order to control the different operations of a micro-grid (MG). In the first step, we exploit the advantages of the Bidirectional Long-Short Term Memory (BiLSTM) deep learning model to predict the next daily electricity price based on time series. In the second step, we use a type-2 fuzzy logic controller to decide which energy source will exploit the excess energy produced or meet the MG need. Real data is used in this paper to test the effectiveness of the proposed EMS whose superiority is proved through the test period. The BiLSTM forecasting model better performs compared to other related algorithms designed to the electricity price prediction. In addition, the proposed decision-making process can reduce the total MG cost and protect the batteries against the deep discharge and maximum charge in order to prolong their lifespan. We expect that this work can contribute to meet the real-world needs in the management of the electrical system. [ABSTRACT FROM AUTHOR]

Datacenters are expanding in size and complexity to the point where anomaly detection and infrastructure monitoring become critical challenges. One potential strategy for dealing with the reliability of computational nodes in a datacenter is to identify cluster nodes or virtual machines exhibiting anomalous behavior. Throughout this paper, we introduce a novel clustering approach for analyzing cluster node behavior while running various workloads in a system based on resource usage details (CPU utilization, network events, etc.). The new clustering technique aims at boosting the efficiency of fuzzy clustering algorithms based on the maximum likelihood estimation (MLE) scheme. We propose the use of a recently developed object-to-group distance since it does not involve the computation of distances among all pairs of objects to assign the objects to the most appropriate group. The experimental findings under realistic settings demonstrate that the newly implemented algorithm outperforms many similar algorithms that have been used frequently in such tasks. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) is one of the most common consequences of diabetes. It affects the retina, causing blood vessel damage which can lead to loss of vision. Saving patients from losing their sight or at least slowing the progress of this disease depends mainly on the early detection of this pathology, on top of the detection of its specific stage. Furthermore, the early detection of diabetic retinopathy and the follow-up of the patient's condition remains an arduous task, whether for an experienced expert ophthalmologist or a computer-aided diagnosis technician. In this paper, we aim to propose a new automatic diabetic retinopathy severity level detection method. The proposed approach merges the pyramid hierarchy of the discrete wavelet transform of the retina fundus image with the modified capsule network and the modified inception block proposed, in addition to a new deep hybrid model that concatenates the inception block with capsule networks. The performance of our proposed approach has been validated on the APTOS dataset, as it achieved a high training accuracy of 97.71% and a high testing accuracy score of 86.54%, which is considered one of the best scores achieved in this field using the same dataset. [ABSTRACT FROM AUTHOR]

Self-driving systems require the creation of perception, which means that they must learn to interact with their environment, gather information and perform tasks while driving. Localization and mapping are essential concepts that any autonomous vehicle must perceive, as they provide information about the location and distance between objects. The absence of GPS, for example, in tunnels and urban canyons, makes it necessary to develop robust methods based on other vehicle equipment, such as Lidar, cameras, IMU, or to combine them all. Based on Lidar measurements, our work presents an architecture consisting of two main phases of mapping and localization; The mapping phase creates a global map of the environment based on non-semantic features using a fuzzy c-means algorithm instead of a semantic algorithm as it was done in some state-of-the-art work. In addition, the remaining clusters were filtered using the DBSCAN algorithm. The localization phase adopted here followed the particle filter architecture; motion update, measurement update, and resampling to estimate the positions. The main contribution of this work is the novel extension of selecting particles to reduce computational time and maintain long-term localization reliability. We have exhibited a method to select relevant particles after motion updates, based on two approaches: clustering with the k-means algorithm and the sigma points algorithm, which were thoroughly examined on short sequences of the Kitti dataset to discover the best one. The selected approach thoroughly tested on the Pandaset dataset. In addition, we tested our method on a long sequence dataset and compared it with the most recent methods. The analysis performed demonstrated the speed of our method and its ability to capture the features needed for real-time localization. Furthermore, it outperformed the well-known localization methods. [ABSTRACT FROM AUTHOR]

Soft clustering is able to handle overlapping data thanks to membership functions that allow examples to belong to several clusters. However, the formula linking membership functions to centers and the adoption of fixed time steps, when updating the parameters, prevents research from exploring areas that may be promising. To alleviate this complexity and enlarge the search space, we propose a constrained optimization model that de-assigns memberships from centers. To take advantage of the ability of neural networks to understand the characteristics of the data and of dynamic systems to remember previous groupings, we solve the proposed model by the third-order gradient recurrent network whose stability point is formed by the memberships and the centers. In this respect, the fixed time step Euler-Cauchy algorithm prevents the search from exploring promising regions. To remedy this problem, we adopt a variable time step that allows a maximum decay of the Lyaponuv function at each iteration. We demonstrate that dissociating the membership coefficients from the group centers results in a less complex optimization model with a larger set of feasible solutions offering better optimal fuzzy clustering. We compared our method to different clustering methods basing on silhouette, separability, compactness criterion, and on Dunn's index; our method has shown its superiority on academic data sets. We used our method to group plants based on 24 nutrients which results in a better classification in comparison with the unidimensional methods. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy is one of the most dangerous complications of diabetes. It affects the eyes causing damage to the blood vessels of the retina. Eventually, as the disease develops, it is possible to lose sight. The main cure for this pathology is based on the early detection which plays a crucial role in slowing the progress of the underlying disease and protecting many patients from losing their sight. However, the detection of diabetic retinopathy at its early stages remains an arduous task that requires human expert interpretation of fundus images in order to vigilantly follow-up the patient. In this paper, we shall propose a new automatic diabetic retinopathy detection method that based on deep-learning. The aforementioned approach is composed of two main steps: an initial pre-processing step where the deformable registration is applied on the retina to occupy the entire image and eliminate the effect of the background on the classification process. The second step is the classification phase in which we train four convolutional neural networks (CNN) models (Densenet-121, Xception, Inception-v3, Resnet-50) to detect the stage of diabetic retinopathy. The performance of our proposed architecture has been tested on the APTOS 2019 dataset. As the latter is relatively small, a transfer learning is adopted by pre-training the mentioned CNNs on the ImageNet dataset and fine-tuning them on the APTOS dataset. In the testing phase, the final prediction is obtained by a system of voting based on the output of the four convolutional neural networks. Our model has performed an accuracy of 85.28% in the testing phase. [ABSTRACT FROM AUTHOR]

The article titled "Correction: Multi-task approach based on combined CNN-transformer for efficient segmentation and classification of breast tumors in ultrasound images" discusses a correction made to the original article. The authors reported that the wrong version of the abstract and keywords were mistakenly inserted into the article. The corrected abstract and keywords describe a proposed multi-task framework for segmenting and classifying tumors in breast ultrasound images using a hybrid architecture of convolutional neural networks (CNNs) and Vision Transformers (ViTs). The framework incorporates various modules to enhance feature extraction and fusion, resulting in improved segmentation and classification performance. The corrected article provides experimental results that demonstrate the superiority of the proposed framework over recent works in terms of segmentation accuracy and classification performance. [Extracted from the article]

Wind energy is a green source of electricity that is growing faster than other renewable energies. However, dependent mainly on wind speed, this source is characterized by the randomness and fluctuation that makes challenging optimal management. In order to remedy this inconvenience, it is essential to predict meteorological data or power produced by generators. In this paper, we present a wind power forecasting approach based on regularized extreme learning machine algorithm (R-ELM), particle swarm optimization method (PSO), and AutoEncoder network (AE) so-called AutoEncoder-optimal regularized extreme learning machine (AE-ORELM). Firstly, we train the AE model by the ELM algorithm. Then, the output weights resulting are used as the input weights of the R-ELM model. Furthermore, the PSO method is used to optimally select hyperparameters of the whole model, namely the regularization parameter and the number of hidden nodes in the hidden layer. The simulation results show that the proposed AE-ORELM can achieve better testing accuracy with a faster training time compared to related models. [ABSTRACT FROM AUTHOR]

The rapid growth in virtualization solutions has driven the widespread adoption of cloud computing paradigms among various industries and applications. This has led to a growing need for XaaS solutions and equipment to enable teleworking. To meet this need, cloud operators and datacenters have to overtake several challenges related to continuity, the quality of services provided, data security, and anomaly detection issues. Mainly, anomaly detection methods play a critical role in detecting virtual machines' abnormal behaviours that can potentially violate service level agreements established with users. Unsupervised machine learning techniques are among the most commonly used technologies for implementing anomaly detection systems. This paper introduces a novel clustering approach for analyzing virtual machine behaviour while running workloads in a system based on resource usage details (such as CPU utilization and downtime events). The proposed algorithm is inspired by the intuitive mechanism of flocking birds in nature to form reasonable clusters. Each starling movement's direction depends on self-information and information provided by other close starlings during the flight. Analogically, after associating a weight with each data sample to guide the formation of meaningful groups, each data element determines its next position in the feature space based on its current position and surroundings. Based on a realistic dataset and clustering validity indices, the experimental evaluation shows that the new weighted fuzzy c-means algorithm provides interesting results and outperforms the corresponding standard algorithm (weighted fuzzy c-means). [ABSTRACT FROM AUTHOR]

Live migration of virtual machines proves to be inexorable in providing load balancing among physical devices and allowing scalability and flexibility in resource allocation. The existing approaches exhibit different policies, distinct performance characteristics, and side effects such as power consumption and performance degradation. Therefore, determining the most optimal live migration algorithm in certain situations remains an open challenge. In this work, a new prediction-based model to manage the live migration process of VMs is introduced. Our adaptive model dynamically identifies the optimal live migration algorithm for a given performance metric based on a prior diagnosis of the system. The model is developed by considering the assumption of different workloads alongside certain resource constraints for any of the currently available migration algorithms. The proposed model consists of an ensemble-learning strategy that involves linear and non-parametric regression methods to predict six live migration key metrics, provided by the operator and/or the user, for each live migration algorithm. Our model allows considering the best combination which is constituted of the algorithm-metric pair to migrate a VM. The experimental results show that the proposed model allows to significantly alleviate the service level agreement violation rate by between 31 % and 60 % , along with decreasing the total CPU time required for the prediction process. [ABSTRACT FROM AUTHOR]

In E-Learning Systems, the automatic detection of the learners' learning styles provides a concrete way for instructors to personalize the learning to be made available to learners. The classification techniques are the most used techniques to automatically detect the learning styles by processing data coming from learner interactions with the system. By using these classification techniques, considerable results are obtained by several approaches with various learning style models. The performance of these approaches varies from one approach to another depending on the data used. However, these approaches have some limitations related to robustness. Indeed, a common feature of these approaches is that they consider only a single course in their trials. Whereas to construct a robust classifier, a representative set of data is crucial. Subsequently, a robust approach for automatically detecting learning styles must take into account the wealth of information to be processed. Therefore, it must consider that the data have to be gathered from learners' learning behaviors that correspond to several courses. In this paper, we propose a robust classifier which can be able to identify the learning style of the learner in E Learning System. The learning behavior of the learner is captured on varied contexts typically on varied courses appertaining to a specific subject matter. The web usage mining is used for capturing the learners' behaviors and then, the learning styles are mapped to Felder-Silverman Learning Style Model (FSLSM) categories. Fuzzy C Means (FCM) algorithm is used to cluster the captured learning behavioral data into FSLSM categories. The experiment results show the performance of our approach although the captured data are gathered from the learners' learning behaviors corresponding to several courses. [ABSTRACT FROM AUTHOR]

In cloud computing, the virtual machine placement is a critical process which aims to identify the most appropriate physical machine to host the virtual machine. It has a significant impact on the performance, resource usage and energy consumption of the datacenters. In order to reduce the number of active physical machines in a datacenter, several virtual machine placement schemes have already been designed and proposed. This study investigates how do four different methods compare to each other in terms of accuracy and efficiency for solving the virtual machine placement as a knapsack problem. A new approach has been adopted which focuses on maximizing the use of a server's central processing unit resource considering a certain capacity threshold. The compared methods are classified; two belong to the category of the exact methods, i.e., branch and bound and dynamic programming, while the other two represent the approximate approach, i.e., genetic algorithm and ant colony optimization algorithm. Experimental results show that the metaheuristic ant colony optimization algorithm outperforms the other three algorithms in terms of efficiency. [ABSTRACT FROM AUTHOR]

Recommendation systems have become based on graph neural networks (GNN) as many fields, and this is due to the advantages that represent this kind of neural networks compared to the classical ones; notably, the representation of concrete realities by taking the relationships between data into consideration and understanding them in a better way. In this paper, we have proposed an item-based recommender system using a deep GraphSAGE model, which learns item embeddings from the user–item matrix and uses them for recommending items that are similar to the ones that users have interacted with before. Furthermore, we have discussed the common problems that usually arise when using deep GNN-based architectures, and which can negatively affect the performance of our recommender system, in particular, the over-smoothing problem. To this end, we have integrated the Jumping Knowledge connections (JK) strategy in our system, using a new method called Ordinal Aggregation Network (OAN) as a layer aggregator to tackle this kind of problem. To evaluate the recommendations, we have used the required metrics that are designated for this purpose: Hits@n and NDCG@n, and we have also measured the duration of training of every model. The experimental results that we have made show that our method has improved the performance of a recommender system concretely and efficiently compared to other aggregation methods. In addition, they have suggested that deep GraphSAGE with Jumping Knowledge connections (JK) would be empirically promising. [ABSTRACT FROM AUTHOR]