Salim Chikhi

Color quantization and its impact on color histogram based image retrieval accuracy

The comparison of color histograms is one of the most widely used techniques for Content-Based Image Retrieval. Before establishing a color histogram in a defined model (RGB, HSV or others), a process of quantization is often used to reduce the number of used colors. In this paper, we present the results of an experimental investigation studying the impact of this process on the accuracy of research results and thus will determine the number of intensities most appropriate for a color quantization for the best accuracy of research through tests applied on an image database of 500 color images.

Dynamic fuzzy logic and reinforcement learning for adaptive energy efficient routing in mobile ad-hoc networks

To ensure good network performance, a routing protocol for MANETs must change its routing policy online to account for changes in network conditions and to deal with routing information imprecision.The main focus of this paper is on the use of fuzzy logic and reinforcement learning.A dynamic membership function is defined to enhance the adaptivity of legacy fuzzy logic systems.A fuzzy extension of a reinforcement learning based routing protocol for MANETs is presented.Dynamic fuzzy logic is more appropriate than reinforcement learning for adaptive energy aware routing in MANETs. In this paper, a dynamic fuzzy energy state based AODV (DFES-AODV) routing protocol for Mobile Ad-hoc NETworks (MANETs) is presented. In DFES-AODV route discovery phase, each node uses a Mamdani fuzzy logic system (FLS) to decide its Route REQuests (RREQs) forwarding probability. The FLS inputs are residual battery level and energy drain rate of mobile node. Unlike previous related-works, membership function of residual energy input is made dynamic. Also, a zero-order Takagi Sugeno FLS with the same inputs is used as a means of generalization for state-space in SARSA-AODV a reinforcement learning based energy-aware routing protocol. The simulation study confirms that using a dynamic fuzzy system ensures more energy efficiency in comparison to its static counterpart. Moreover, DFES-AODV exhibits similar performance to SARSA-AODV and its fuzzy extension FSARSA-AODV. Therefore, the use of dynamic fuzzy logic for adaptive routing in MANETs is recommended.

An adaptive energy-aware routing protocol for MANETs using the SARSA reinforcement learning algorithm

In MANETs (Mobile Ad-hoc NETworks), communicating nodes are powered by batteries which could not be re-charged in many practical usage scenarios. Hence, maximizing network lifetime is a critical optimization objective in routing protocols design for MANETs. To meet this objective, energy-consumption should be balanced among all mobile nodes. In this paper, we formulate the energy-aware route discovery problem in a reactive routing protocol as a Reinforcement Learning (RL) problem that we solve using the SARSA RL algorithm. We have implemented our proposed RL-model on the top of AODV a well-known reactive routing protocol for MANETs. Furthermore, we show through simulations the efficiency of our proposal, against an implementation of the Energy-Aware Probability routing protocol.

Adaptive maximum-lifetime routing in mobile ad-hoc networks using temporal difference reinforcement learning

Mobile ad-hoc NETworks (MANETs) are very dynamic environments. A routing protocol for MANETs should be adaptive in order to operate correctly in presence of variable network conditions. Reinforcement learning (RL) is a recently used technique to achieve adaptive routing in MANETs. In comparison to other machine learning and computational intelligence techniques, RL achieves optimal results at low processing and medium memory costs. To deal with adaptive energy-aware routing issue in MANETs, a RL-based maximum-lifetime routing strategy is proposed. Each mobile node learns how to adjust its route-request packets forwarding-rate according to its energy profile. In terms of RL-resolution methods, Q-Learning, SARSA, Q(λ) and SARSA(λ) which are Temporal difference RL-algorithms are used. The proposed RL model is implemented on the top of AODV routing protocol. Simulation results show that the RL-based AODV achieved good performances in comparison to Time-Delay and Probability based AODV. Particularly, the Q-Learning based AODV has marked the best global performances in terms of energy efficiency and end to end delay.

Comparative study of the use of geometrical moments for Arabic handwriting recognition

Moments and functions of moments have been employed as pattern features in numerous applications to recognize two-dimensional image patterns. These pattern features extract global properties of the image such as the shape area, the center of the mass, the moment of inertia, and so on. This paper shows the use of different moments to extract features from offline Arabic words. Invariants moment of Hu, Zernike moments, Pseudo Zernike moments, Tchebichef moments, and Legendre moments have been applied to the IFN/ENIT database with a neural network classifier and the results have been compared. Our results show that pseudo Zernike moments yields the best recognition accuracy of 89%.

A Survey of Reinforcement Learning Based Routing Protocols for Mobile Ad-Hoc Networks

Designing mobility and power aware routing protocols have made the main focus of the early contributions to the field of Mobile Ad-hoc NETworks (MANETs). However, almost all conventional routing protocols for MANETs suffer from their lack of adaptivity leading to their performance degradation under varying network conditions. In fact, this is due to both simplistic conception hypotheses they made about the network and to the use of some prefixed parameters in protocols implementations. Currently, artificial intelligence methods like Reinforcement Learning (RL) are widely used to design adaptive routing strategies for MANETs. In this paper, we present a comprehensive survey of RL-based routing protocols for MANETs. Besides, we propose some future research directions in this area.

A novel fuzzy approach for handwritten Arabic character recognition

AbstractThe aim of our work is to present a new method based on structural characteristics and a fuzzy classifier for off-line recognition of handwritten Arabic characters in all their forms (beginning, end, middle and isolated). The proposed method can be integrated in any handwritten Arabic words recognition system based on an explicit segmentation process. First, three preprocessing operations are applied on character images: thinning, contour tracing and connected components detection. These operations extract structural characteristics used to divide the set of characters into five subsets. Next, features are extracted using invariant pseudo-Zernike moments. Classification was done using the Fuzzy ARTMAP neural network, which is very fast in training and supports incremental learning. Five Fuzzy ARTMAP neural networks were employed; each one is designed to recognize one subset of characters. The recognition process is achieved in two steps: in the first one, a clustering method affects characters to one of the five character subsets. In the second one, the pseudo-Zernike features are used by the appropriate Fuzzy ARTMAP classifier to identify the character. Training process and tests were performed on a set of character images manually extracted from the IFN/ENIT database. A height recognition rate was reportedn.

Evolving cellular automata by parallel quantum genetic algorithm

Evolving solutions rather than computing them certainly represents a promising programming approach. Evolutionary computation has already been known in computer science since more than 4 decades. More recently, another alternative of evolutionary algorithms was invented: quantum genetic algorithms. In this paper, we outline the approach of quantum genetic algorithm (QGA) by giving an example where it serves to automatically program cellular automata (CA) rules. Our results have shown that QGA can be a very promising tool for exploring CA search spaces.

A new large Arabic database for offline handwriting recognition

To evaluate the performances of handwriting recognition systems, it is necessary to compare them objectively on the same database. A few freely databases are available for Arabic handwriting recognition, for this reason we have developed a new database of Algerian village names to be available freely for research and academic use. Up to now the database contains 1,209 forms including 26,580 binary and greyscale images representing 886 Algerian village names, collected from 1,209 writers. We also describe a new character segmentation algorithm for offline handwritten Arabic words. This algorithm uses a set of fuzzy ART neural networks as classifiers and Zernike invariant moments as features. The proposed system was trained and tested for the first time using the new database. A height segmentation and recognition accuracy were reported.

Routing in Mobile Ad-hoc Networks as a Reinforcement Learning Task

Communicating nodes in Mobile Ad-hoc NETworks (MANETs) must deal with routing in an efficient and adaptive way. Efficiency feature is strongly recommended since both bandwidth and energy are scarce resources in MANETs. Besides, adaptivity is crucial to accomplish the routing task correctly in presence of varying network conditions in terms of mobility, links quality and traffic load. Our focus, in this paper, is on the application of Reinforcement Learning (RL) technique to achieve adaptive routing in MANETs. Particularly, we try to underline the main design-issues that arise when dealing with adaptive-routing as a Reinforcement Learning task.