Artificial Intelligence and Deep Learning applications are well-developed as a part of human life... more Artificial Intelligence and Deep Learning applications are well-developed as a part of human life. In the field of object recognition, Convolutional Neural Network (CNN) based methods are getting more and more important and challenging. However, existing CNN methods do not perform well on datasets that exhibit high similarities, resulting in confusion between different classes. In this study, we propose a new Deep Learning approach for recognizing date fruit categories based on the Deep Convolutional Neural Network (DCNN). The modified fine-tuning (MFTs-Net) approach can recognize with high accuracy the different date fruit categories. In order to train and to test the robustness of our proposed method, we have collected a dataset that takes into account different date fruit categories. The presented dataset is challenging as it contains classes of a unique object and presents high similarities concerning the shape, color and texture of date fruit. We show that the MFTs-Net CNN we implemented, trained and tested using the collected dataset can recognize with high accuracy the different date categories compared with state-of-the-arts works. The presented methodology works perfectly with very small datasets, which is one of the main strengths of the proposed method. Our MFTs-Net architecture performs perfectly on test data with an accuracy of 98%. 1
In many applications of Wireless Sensor Networks (WSNs), event detection is the main purpose of u... more In many applications of Wireless Sensor Networks (WSNs), event detection is the main purpose of users. Moreover, determining where and when that event occurs is crucial; thus, the positions of nodes must be identified. Subsequently, in a range-free case, the Distance Vector-Hop (DV-Hop) heuristic is the commonly used localization algorithm because of its simplicity and low cost. The DV-Hop algorithm consists of a set of reference nodes, namely, anchors, to periodically broadcast their current positions and assist nearby unknown nodes during localization. Another potential solution includes the use of only one mobile anchor instead of these sets of anchors. This solution presents a new challenge in the localization of rang-free WSNs because of its favorable results and reduced cost. In this paper, we propose an analytical probabilistic model for multi-hop distance estimation between mobile anchor nodes and unknown nodes. We derive a non-linear analytic function that provides the relation between the hop counts and distance estimation. Moreover, based on the recursive least square algorithm, we present a new formulation of the original DV-Hop localization algorithm, namely, online DV-Hop localization, in WSNs. Finally, different scenarios of path planning and simulation results are conducted.
Abstract: Let G and H be two simple undirected graphs. An embedding of the graph G into the graph... more Abstract: Let G and H be two simple undirected graphs. An embedding of the graph G into the graph H is an injective mapping f from the vertices of G to the vertices of H. The study of graph embedding arises naturally in a number of computational problems: portability of algorithms across various parallel architectures and layout of circuits in VLSI. Akers and Krishnamurthy proposed the pancake as an alternative to the hypercube and its variations for interconnecting processors in parallel computers.
How to cite this article: Karima Femmam and Smain Femmam, Improving the dimensionality reduction ... more How to cite this article: Karima Femmam and Smain Femmam, Improving the dimensionality reduction of PCA using bivariate copulas, Advances and Applications in Statistics 86(1) (2023), 47-64.
Proceedings of the 3rd International Conference on Intelligent Information Processing
In this paper, we develop a new methodology and a new conceptual implementation based on the ZigB... more In this paper, we develop a new methodology and a new conceptual implementation based on the ZigBee sensor communication platform, used for perceiving materials of unknown environments. In this implementation, we distinguish different unknown environments by actively contacting and testing them, and by analysing the resulting signals using a new experimental ZigBee sensor. For this implementation, we identify sensor-derived measures that are diagnostic of material properties, and use these measures to classify the unknown environments in their different perception class. The experiment is based on a wireless communication between the unknown environments and the analysis landed station for characterizations. The parameter of classification is the internal angle of friction of the unknown environment based on its own material.
Handling datasets nowadays has become a crucial task, since today's world is heavily dependent on... more Handling datasets nowadays has become a crucial task, since today's world is heavily dependent on data information. However, many data tend to be big and contain redundancy which makes them difficult to deal with. Due to that, data pre-processing became almost necessary before using any data, and one of the main tasks in data preprocessing is dimensionality reduction. In this paper we propose a new approach for dimensionality reduction using feature selection method based on bivariate copulas. This approach is a direct application of copulas to describe and model the inter-correlation between any two dimensionsbivariate analysis. The study will first show how we use the bivariate method to detect redundant dimensions and eliminate them, and then compare the quality of the results against most-known selection methods in term of accuracy, using statistical precision and classification models.
Chaos represents interesting features that are suitable for the cryptography domain. One-dimensio... more Chaos represents interesting features that are suitable for the cryptography domain. One-dimensional chaotic maps are widely used for security issues due to their simplicity and chaotic behavior versus other multidimensional chaotic maps that can be complex for hardware implementation and hard to analyze. However, classical one-dimensional chaotic maps present a reduced range of chaotic behavior. In this paper, we propose two new piecewise compound one-dimensional chaotic maps; an Altered Sine-Logistic map based on Tent map (ASLT) and a combined Cubic-Tent map (CT). The proposed compound maps combine classical and simple one-dimensional chaotic maps to produce an extensive range of chaotic behavior. The ASLT system comprises a combined Sine-Tent map in the first piece of the function and a combined Logistic-Tent map in the second piece of the function. Then, the CT map is based on the nonlinear fusion operation between the Cubic map and the piecewise Tent map. Simulation results and...
Abstract: In this chapter, we present a new type of model particularly suited to solving physical... more Abstract: In this chapter, we present a new type of model particularly suited to solving physical problems in non-Euclidean spaces. These spaces may mix the concepts of proximity (such as geographic information systems), topological notions (such as social networks), information concepts and algebraic concepts (such as the invariance of certain local symmetries).
Many recent papers deal with the improvement of tools to characterize ultrasonic signal in NonDes... more Many recent papers deal with the improvement of tools to characterize ultrasonic signal in NonDestructive Control of materials. The spectral signature of an ultrasonic signal can be characterized by a set of ultrasonic resonances related to the material shape, size and physical properties. Many methods are used for characterizing this signature either in frequency domain or in time domain. In both cases, one cannot take into account all the physical phenomena. This paper is related to the problem of characterization of materials by use of time-frequency methods. It is concerned with an application of Wigner-Ville transforms in non-destructive evaluation of materials by ultrasound, for a better segmentation of the signal (energies). This segmentation will allow to the spectral ratio method to provide better results.
Wireless Sensor Network (WSN) architectures are widely used in a variety of practical application... more Wireless Sensor Network (WSN) architectures are widely used in a variety of practical applications. In most cases of application, the event information transmitted by a sensor node via the network has no significance without the knowledge of its accurate geographical localization. In this paper, a method based on Machine Learning Technique (MLT) is proposed to improve node accuracy localization in WSN. We propose a Single Hidden Layer Extreme Learning Machine (SHL-ELM) and a Two Hidden Layer Extreme Learning Machine (THL-ELM) based methods for nodes localization in WSN. The suggested methods are applied in different Multi-hop WSN deployment cases. We focused on range-free localization algorithm in isotropic case and irregular environments. Simulation results demonstrate that the proposed THL-ELM algorithm greatly minimizes the average localization errors when compared to the Single Hidden Layer Extreme Learning Machine and the Distance Vector Hop (DV- Hop) algorithm.
Variations in scaling behavior in the flux and emissions of distant astronomical sources with res... more Variations in scaling behavior in the flux and emissions of distant astronomical sources with respect to their cosmic time are important phenomena that can provide valuable information about the dynamics within the sources and their cosmological evolution with time. Different studies have been applying linear analysis to understand and model quasars' light curves. Here, we study the multifractal behavior of selected quasars' radio emissions in their observed frame (at 22 and 37 GHz bands) and their rest frame. To this end, we apply the wavelet transform-based multifractal analysis formalism called wavelet transform modulus maxima. In addition, we verify whether the autoregressive integrated moving average (ARIMA) models fit our data. In our work, we observe strong multifractal behavior for all the sources. Additionally, we find that the degree of multifractality is strongly similar for each source and significantly different between sources at 22 and 37 GHz. This similarity implies that the two frequencies have the same radiation region and mechanism, whereas the difference indicates that the sources have intrinsically different dynamics. Furthermore, we show that the degree of multifractality is the same in the observed and rest frames of the quasars, i.e., multifractality is an intrinsic property of radio quasars. Finally, we show that the ARIMA models fit the 3C 345 quasar at 22 GHz and partially fit most of the time series, with the exception of the 3C 273 and 3C 279 quasars at 37 GHz, for which the models are found to be inadequate.
In robotics, three large families of actuators are mainly used: electric, hydraulic and pneumatic... more In robotics, three large families of actuators are mainly used: electric, hydraulic and pneumatic actuators. The conversion of energy into motion is a key point of actuators. It is possible that smart actuators can be achieved with self‐calibration, information processing, communications and control systems or servos. Actuators are chosen according to the tasks achieved and required performance. When great precision, important torque performance, significant velocities and acceleration are desirable, people resort to electric actuators or direct current motors. This chapter presents the modeling of the dynamic behavior of rotating DC machines. The dynamic model of the pneumatic system is composed of two stages: a dynamic stage that takes the mechanical aspects into consideration and a second stage corresponding to the pneumatic actuators. The hydraulic actuator modeling is based on the study of the flux stages provided by a fixed pressure.
Thirty-Sixth Southeastern Symposium on System Theory, 2004. Proceedings of the
A delay-dependent approach analysis to robust filtering is proposed for linear discrete-time syst... more A delay-dependent approach analysis to robust filtering is proposed for linear discrete-time systems with multiple delays in the state. The uncertain parameters are supposed to reside in a polytope and the attention is focused on the analysis and design of robust filters guaranteeing a prescribed noise attenuation level. The proposed filter analysis methodology includes some recently appeared results.
Artificial Intelligence and Deep Learning applications are well-developed as a part of human life... more Artificial Intelligence and Deep Learning applications are well-developed as a part of human life. In the field of object recognition, Convolutional Neural Network (CNN) based methods are getting more and more important and challenging. However, existing CNN methods do not perform well on datasets that exhibit high similarities, resulting in confusion between different classes. In this study, we propose a new Deep Learning approach for recognizing date fruit categories based on the Deep Convolutional Neural Network (DCNN). The modified fine-tuning (MFTs-Net) approach can recognize with high accuracy the different date fruit categories. In order to train and to test the robustness of our proposed method, we have collected a dataset that takes into account different date fruit categories. The presented dataset is challenging as it contains classes of a unique object and presents high similarities concerning the shape, color and texture of date fruit. We show that the MFTs-Net CNN we implemented, trained and tested using the collected dataset can recognize with high accuracy the different date categories compared with state-of-the-arts works. The presented methodology works perfectly with very small datasets, which is one of the main strengths of the proposed method. Our MFTs-Net architecture performs perfectly on test data with an accuracy of 98%. 1
In many applications of Wireless Sensor Networks (WSNs), event detection is the main purpose of u... more In many applications of Wireless Sensor Networks (WSNs), event detection is the main purpose of users. Moreover, determining where and when that event occurs is crucial; thus, the positions of nodes must be identified. Subsequently, in a range-free case, the Distance Vector-Hop (DV-Hop) heuristic is the commonly used localization algorithm because of its simplicity and low cost. The DV-Hop algorithm consists of a set of reference nodes, namely, anchors, to periodically broadcast their current positions and assist nearby unknown nodes during localization. Another potential solution includes the use of only one mobile anchor instead of these sets of anchors. This solution presents a new challenge in the localization of rang-free WSNs because of its favorable results and reduced cost. In this paper, we propose an analytical probabilistic model for multi-hop distance estimation between mobile anchor nodes and unknown nodes. We derive a non-linear analytic function that provides the relation between the hop counts and distance estimation. Moreover, based on the recursive least square algorithm, we present a new formulation of the original DV-Hop localization algorithm, namely, online DV-Hop localization, in WSNs. Finally, different scenarios of path planning and simulation results are conducted.
Abstract: Let G and H be two simple undirected graphs. An embedding of the graph G into the graph... more Abstract: Let G and H be two simple undirected graphs. An embedding of the graph G into the graph H is an injective mapping f from the vertices of G to the vertices of H. The study of graph embedding arises naturally in a number of computational problems: portability of algorithms across various parallel architectures and layout of circuits in VLSI. Akers and Krishnamurthy proposed the pancake as an alternative to the hypercube and its variations for interconnecting processors in parallel computers.
How to cite this article: Karima Femmam and Smain Femmam, Improving the dimensionality reduction ... more How to cite this article: Karima Femmam and Smain Femmam, Improving the dimensionality reduction of PCA using bivariate copulas, Advances and Applications in Statistics 86(1) (2023), 47-64.
Proceedings of the 3rd International Conference on Intelligent Information Processing
In this paper, we develop a new methodology and a new conceptual implementation based on the ZigB... more In this paper, we develop a new methodology and a new conceptual implementation based on the ZigBee sensor communication platform, used for perceiving materials of unknown environments. In this implementation, we distinguish different unknown environments by actively contacting and testing them, and by analysing the resulting signals using a new experimental ZigBee sensor. For this implementation, we identify sensor-derived measures that are diagnostic of material properties, and use these measures to classify the unknown environments in their different perception class. The experiment is based on a wireless communication between the unknown environments and the analysis landed station for characterizations. The parameter of classification is the internal angle of friction of the unknown environment based on its own material.
Handling datasets nowadays has become a crucial task, since today's world is heavily dependent on... more Handling datasets nowadays has become a crucial task, since today's world is heavily dependent on data information. However, many data tend to be big and contain redundancy which makes them difficult to deal with. Due to that, data pre-processing became almost necessary before using any data, and one of the main tasks in data preprocessing is dimensionality reduction. In this paper we propose a new approach for dimensionality reduction using feature selection method based on bivariate copulas. This approach is a direct application of copulas to describe and model the inter-correlation between any two dimensionsbivariate analysis. The study will first show how we use the bivariate method to detect redundant dimensions and eliminate them, and then compare the quality of the results against most-known selection methods in term of accuracy, using statistical precision and classification models.
Chaos represents interesting features that are suitable for the cryptography domain. One-dimensio... more Chaos represents interesting features that are suitable for the cryptography domain. One-dimensional chaotic maps are widely used for security issues due to their simplicity and chaotic behavior versus other multidimensional chaotic maps that can be complex for hardware implementation and hard to analyze. However, classical one-dimensional chaotic maps present a reduced range of chaotic behavior. In this paper, we propose two new piecewise compound one-dimensional chaotic maps; an Altered Sine-Logistic map based on Tent map (ASLT) and a combined Cubic-Tent map (CT). The proposed compound maps combine classical and simple one-dimensional chaotic maps to produce an extensive range of chaotic behavior. The ASLT system comprises a combined Sine-Tent map in the first piece of the function and a combined Logistic-Tent map in the second piece of the function. Then, the CT map is based on the nonlinear fusion operation between the Cubic map and the piecewise Tent map. Simulation results and...
Abstract: In this chapter, we present a new type of model particularly suited to solving physical... more Abstract: In this chapter, we present a new type of model particularly suited to solving physical problems in non-Euclidean spaces. These spaces may mix the concepts of proximity (such as geographic information systems), topological notions (such as social networks), information concepts and algebraic concepts (such as the invariance of certain local symmetries).
Many recent papers deal with the improvement of tools to characterize ultrasonic signal in NonDes... more Many recent papers deal with the improvement of tools to characterize ultrasonic signal in NonDestructive Control of materials. The spectral signature of an ultrasonic signal can be characterized by a set of ultrasonic resonances related to the material shape, size and physical properties. Many methods are used for characterizing this signature either in frequency domain or in time domain. In both cases, one cannot take into account all the physical phenomena. This paper is related to the problem of characterization of materials by use of time-frequency methods. It is concerned with an application of Wigner-Ville transforms in non-destructive evaluation of materials by ultrasound, for a better segmentation of the signal (energies). This segmentation will allow to the spectral ratio method to provide better results.
Wireless Sensor Network (WSN) architectures are widely used in a variety of practical application... more Wireless Sensor Network (WSN) architectures are widely used in a variety of practical applications. In most cases of application, the event information transmitted by a sensor node via the network has no significance without the knowledge of its accurate geographical localization. In this paper, a method based on Machine Learning Technique (MLT) is proposed to improve node accuracy localization in WSN. We propose a Single Hidden Layer Extreme Learning Machine (SHL-ELM) and a Two Hidden Layer Extreme Learning Machine (THL-ELM) based methods for nodes localization in WSN. The suggested methods are applied in different Multi-hop WSN deployment cases. We focused on range-free localization algorithm in isotropic case and irregular environments. Simulation results demonstrate that the proposed THL-ELM algorithm greatly minimizes the average localization errors when compared to the Single Hidden Layer Extreme Learning Machine and the Distance Vector Hop (DV- Hop) algorithm.
Variations in scaling behavior in the flux and emissions of distant astronomical sources with res... more Variations in scaling behavior in the flux and emissions of distant astronomical sources with respect to their cosmic time are important phenomena that can provide valuable information about the dynamics within the sources and their cosmological evolution with time. Different studies have been applying linear analysis to understand and model quasars' light curves. Here, we study the multifractal behavior of selected quasars' radio emissions in their observed frame (at 22 and 37 GHz bands) and their rest frame. To this end, we apply the wavelet transform-based multifractal analysis formalism called wavelet transform modulus maxima. In addition, we verify whether the autoregressive integrated moving average (ARIMA) models fit our data. In our work, we observe strong multifractal behavior for all the sources. Additionally, we find that the degree of multifractality is strongly similar for each source and significantly different between sources at 22 and 37 GHz. This similarity implies that the two frequencies have the same radiation region and mechanism, whereas the difference indicates that the sources have intrinsically different dynamics. Furthermore, we show that the degree of multifractality is the same in the observed and rest frames of the quasars, i.e., multifractality is an intrinsic property of radio quasars. Finally, we show that the ARIMA models fit the 3C 345 quasar at 22 GHz and partially fit most of the time series, with the exception of the 3C 273 and 3C 279 quasars at 37 GHz, for which the models are found to be inadequate.
In robotics, three large families of actuators are mainly used: electric, hydraulic and pneumatic... more In robotics, three large families of actuators are mainly used: electric, hydraulic and pneumatic actuators. The conversion of energy into motion is a key point of actuators. It is possible that smart actuators can be achieved with self‐calibration, information processing, communications and control systems or servos. Actuators are chosen according to the tasks achieved and required performance. When great precision, important torque performance, significant velocities and acceleration are desirable, people resort to electric actuators or direct current motors. This chapter presents the modeling of the dynamic behavior of rotating DC machines. The dynamic model of the pneumatic system is composed of two stages: a dynamic stage that takes the mechanical aspects into consideration and a second stage corresponding to the pneumatic actuators. The hydraulic actuator modeling is based on the study of the flux stages provided by a fixed pressure.
Thirty-Sixth Southeastern Symposium on System Theory, 2004. Proceedings of the
A delay-dependent approach analysis to robust filtering is proposed for linear discrete-time syst... more A delay-dependent approach analysis to robust filtering is proposed for linear discrete-time systems with multiple delays in the state. The uncertain parameters are supposed to reside in a polytope and the attention is focused on the analysis and design of robust filters guaranteeing a prescribed noise attenuation level. The proposed filter analysis methodology includes some recently appeared results.
Uploads
Papers by Smain Femmam