Souhir Bouaziz

A hybrid learning algorithm for evolving Flexible Beta Basis Function Neural Tree Model

Abstract In this paper, a tree-based encoding method is introduced to represent the Beta basis function neural network. The proposed model called Flexible Beta Basis Function Neural Tree (FBBFNT) can be created and optimized based on the predefined Beta operator sets. A hybrid learning algorithm is used to evolving FBBFNT Model: the structure is developed using the Extended Genetic Programming (EGP) and the Beta parameters and connected weights are optimized by the Opposite-based Particle Swarm Optimization algorithm (OPSO). The performance of the proposed method is evaluated for benchmark problems drawn from control system and time series prediction area and is compared with those of related methods.

Evolving flexible beta operator neural trees (FBONT) for time series forecasting

In this paper, a new time-series forecasting model based on the Flexible Beta Operator Neural Tree (FBONT) is introduced. The FBONT model which has a tree-structural representation is considered as a special Beta basis function multi-layer neural network. Based on the pre-defined Beta operator sets, the FBONT can be formed and optimized. The FBONT structure is developed using the Extended Genetic Programming (EGP) and the Beta parameters and connected weights are optimized by the Particle Swarm Optimization algorithm (PSO). The performance of the proposed method is evaluated using time series forecasting problems and compared with those of related methods.

Hybrid harmony search algorithm for global optimization

This paper proposes two hybrid optimization methods based on Harmony Search algorithm (HS) and two different nature-inspired metaheuristic algorithms. In the first contribution, the combination was between the Improved Harmony Search (IHS) and the Particle Swarm Optimization (PSO). The second contribution merged the IHS with the Differential Evolution (DE) operators. The basic idea of hybridization was to ameliorate all the harmony memory vectors by adapting the PSO velocity or the DE operators in order to increase the convergence speed. The new algorithms (IHSPSO and IHSDE) have been compared to the IHS, DE, PSO and some other algorithms like DHS and HSDM. The DHS and HSDM are two existing algorithms, which use different hybridization concepts between HS and DE. All of these algorithms have been evaluated by different test Benchmark functions. The results demonstrated that the hybrid algorithm IHSDE have the better convergence speed into the global optimum than the IHSPSO and the standard IHS, DE and PSO.

A hybrid computational chemotaxis in bacterial foraging optimization algorithm for global numerical optimization

This paper first proposes a simple scheme for adapting the chemotactic step size of the Bacterial Foraging Optimization Algorithm (BFOA), and then this new adaptation and two very popular optimization techniques called Particle Swarm Optimization (PSO) and Differential Evolution (DE) are coupled in a new hybrid approach named Adaptive Chemotactic Bacterial Swarm Foraging Optimization with Differential Evolution Strategy (ACBSFO _DES). This novel technique has been shown to overcome the problems of premature convergence and slow of both the classical BFOA and the other BFOA hybrid variants over several benchmark problems.

Extended immune programming and opposite-based PSO for evolving flexible beta basis function neural tree

In this paper, a new hybrid learning algorithm based on the global optimization techniques, is introduced to evolve the Flexible Beta Basis Function Neural Tree (FBBFNT). The structure is developed using the Extended Immune Programming (EIP) and the Beta parameters and connected weights are optimized using the Opposite-based Particle Swarm Optimization (OPSO) algorithm. The performance of the proposed method is evaluated for time series prediction area and is compared with those of associated methods.

Evolving flexible beta basis function neural tree for nonlinear systems

In this paper, a new evolving artificial neural network using evolutionary computation is introduced. Based on the pre-defined Beta operator sets, this model called Flexible Beta Basis Function Neural Tree (FBBFNT), can be created and learned. The structure is developed using the Extended Immune Programming (EIP). The Beta parameters and connected weights are optimized using the Hybrid Bacterial Foraging Optimization algorithm. The performance of the proposed method is evaluated for nonlinear systems and compared with those of related methods.

Evolving flexible beta basis function neural tree using extended genetic programming & Hybrid Artificial Bee Colony

Display OmittedThe hybrid learning algorithm for FBBFNT model. A new hybrid learning algorithm is introduced to evolve the flexible beta basis function neural tree (FBBFNT) model.The Extended Genetic Programming (EGP) is used to optimize the structure of the FBBFNT.A new hybridization between Artificial Bee Colony (ABC) and Opposite-based Particle Swarm Optimization (OPSO) is proposed to optimize the parameters of FBBFNT.The proposed model is evaluated for benchmark problems drawn from time series prediction area. In this paper, a new hybrid learning algorithm is introduced to evolve the flexible beta basis function neural tree (FBBFNT). The structure is developed using the Extended Genetic Programming (EGP) and the Beta parameters and connected weights are optimized by the Hybrid Artificial Bee Colony algorithm. This hybridization is essentially based on replacing the random Artificial Bee Colony (ABC) position with the guided Opposite-based Particle Swarm Optimization (OPSO) position. Such modification can minimize the delay which might be lead by the random position, in reaching the global solution. The performance of the proposed model is evaluated for benchmark problems drawn from time series prediction area and is compared with those of related methods.

Universal approximation propriety of Flexible Beta Basis Function Neural Tree

In this paper, the universal approximation propriety is proved for the Flexible Beta Basis Function Neural Tree (FBBFNT) model. This model is a tree-encoding method for designing Beta basis function neural network. The performance of FBBFNT is evaluated for benchmark problems drawn from time series approximation area and is compared with other methods in the literature.

The Adaptive Chemotactic Foraging with Differential Evolution algorithm

This work proposes the application of a novel evolutionary approach called the Adaptive Chemotactic Foraging with Differential Evolution algorithm (ACF_DE) on benchmark problems. This method is based on the well-known Bacterial Foraging Optimization Algorithm (BFOA), applying appropriate Differential Evolution operators and including an adaptation scheme of the chemotaxis step size to concentrate the search in the desired optimal zone. The hybrid system is compared with those of related methods on benchmark problems showing its high performance in overcoming slow and premature convergence.

Fuzzy modeling system based on hybrid evolutionary approach

In this paper, we introduce a new evolutionary methodology to design fuzzy inference systems. An innovative hybrid stages of learning method and tuning method, contains Subtractive clustering, Adaptive Neuro-Fuzzy Inference System (ANFIS) and particle swarm optimization (PSO), is developed to generate evolutional fuzzy modeling systems with high accuracy. For the purpose of illustration and validation of the approach, some data sets have been exploited. Empirical results illustrate that the proposed method is efficient.