Advances in Swarm Intelligence

Inverse problems are ill-posed: the error function has its minimum in a flat elongated valley or surrounded by many local minima. Local optimization methods give unpredictable results if no prior information is available. Traditionally this has generated mistrust for the use of inverse methods. Stochastic approaches to inverse problems consists in shift attention to the probability of existence of certain kinds of models (called equivalent) instead of “looking for the true model”. Also, inverse problems are ill-conditioned and often the observed data are noisy. Global optimization methods have become a good alternative to sample the model space efficiently. These methods are very robust since they solve the inverse problem as a sampling problem, but they are hampered by dimensionality issues and high computational costs needed to solve the forward problem (predictions). In this paper we show how our research over the last three years on particle swarm optimizers can be used to solve and evaluate inverse problems efficiently. Although PSO is a stochastic algorithm, it can be physically interpreted as a stochastic damped mass-spring system. This analogy allowed us to introduce the PSO continuous model, to deduce a whole family of PSO algorithms, and to provide some results of its convergence based on the stochastic stability of the particle trajectories. This makes PSO a particularly interesting algorithm, different from other global algorithms which are purely heuristic.

We include the results of an application of our PSO algorithm to the prediction of phosphorylation sites in proteins, an important mechanism for regulation of biological function. Our PSO optimization methods have enabled us to predict phosphorylation sites with higher accuracy and with better generalization, than other reports on similar studies in literature. Our preliminary studies on 984 protein sequences show that our algorithm can predict phosphorylation sites with a training accuracy of 92.5% and a testing accuracy 91.4%, when combined with a neural network based algorithm called Extreme Learning Machine.

A computational method that automatically builds dynamical models of swarming systems from positional data is introduced. As an initial test for the approach, the classical Vicsek model is used to make samples for the computer algorithm and retrieve a model. Time dependent separation measures are introduced in order to characterize the dynamics of a system and then compare the behaviors of the source and retrieved model. Cases of low and high density interactions are considered to verify the generality of the models. The results show the retrieved models are capable of emulating the collective behavior well, especially when the interaction structure resembles the one of the source model.

Swarm intelligent, cooperative object recognition forms part of cooperative construction research. A simulation model was designed and utilised to assess the suitability of a swarm of agents to identify and collect different objects, termed the Simplified Hexagonal Model. An agent in this system cannot assess different object types alone. Key to the efficiency of the system is avoiding stagnation whilst maintaining robustness. This paper examines the energy efficiency of the system when the probability of an agent moving away from an object it is trying to identify is varied. The probability of an agent moving away from an unidentified object per time-step was varied from 1:12 to 1:400. Both low and high probabilities increased the energy required to complete the task. This was more pronounced when using fewer agents. The reduced chance that the required number of agents were surrounding the same objects at the same time caused the increase.

Particle Swarm Optimization (PSO), proposed by Professor Kennedy and Eberhart in 1995, attracts many attentions to solve for a lot of optimization problems nowadays. Due to its simplicity of setting-parameters and computational efficiency, it becomes one of the most popular algorithms in optimizations. However, the discrepancy of PSO is the low dimensionality of the problem can be solved. Once the optimized function becomes complicated, the efficiency gained in PSO degradates rapidly. More complex algorithms on PSO required. Therefore, different algorithms will be applied to different problems with difficulties. Three different algorithms are suggested to solve different problems accordinately. In summary, proposed PSO algorithms apply well to problems with different difficulties in the final simulations.

Particle swarm optimization (PSO) algorithm can be viewed as a series of iterative matrix computation and its population diversity can be considered as an observation of the distribution of matrix elements. In this paper, PSO algorithm is first represented in the matrix format, then the PSO normalized population diversities are defined and discussed based on matrix analysis. Based on the analysis of the relationship between pairs of vectors in PSO solution matrix, different population diversities are defined for separable and non-separable problems, respectively. Experiments on benchmark functions are conducted and simulation results illustrate the effectiveness and usefulness of the proposed normalized population diversities.

This paper introduces an enhancement to the particle swarm optimization algorithms that models a characteristic of social groups: the disagreements between individuals. After a short introduction, we describe the new concept theoretically and define a special type of particle swarm optimization with disagreements: the 6

σ

-PSOD. Based on it, we conduct some tests proving that it can perform better, having strengthened neighborhood focus using partial disagreements and enhanced exploration capabilities through extreme disagreements.

Landslide is an increasing problem and a cause for concern in densely populated areas. In addition to field studies and laboratory experiments, engineers also embrace computer technology to find better solutions and to achieve better landslide analysis. Traditionally, the main analysis framework is the limit equilibrium analysis, in which the state of limit equilibrium in the target slope is assumed to be reached along the entire sliding surface at the same time of failure. However, the assumption of circular sliding surfaces would prevent the search of non-circular sliding surfaces with lower factors of safety. To address this problem, particle swarm optimization (PSO) is implemented in this study using a computer program written in C# to automatically discover the optimal results in the target function. The results show that the PSO-based approach offers many interesting outcomes.

Cutting and packing (C & P) problem is to allocate a set of items to larger rectangular standardized units by minimizing the waste. Bin packing, strip packing and cutting stock problem is well-known classical C & P problem. An overview is provided of several meta-heuristics algorithms of swarm intelligence from the literature for the 2D C & P problem. The objective of this paper is to present and categorize the solution approaches in the literature for 2D regular and irregular C & P problem. The focus is hereby on the analysis of the methods and application of swarm intelligence algorithms.

In Particle Swarm Optimization (PSO), setting the inertia weight w is one of the most important topics. The inertia weight was introduced into PSO to balance between itsglobal and local search abilities. In this paper, first, wepropose a method to adaptively adjust the inertia weight based on particle’s velocity information. Second, we utilize both position and velocity information to adaptively adjust the inertia weight. The proposed methodsare then tested on benchmark functions. The simulation results illustrate the effectiveness and efficiency of the proposed algorithm by comparing it with other existingPSOs.

A nonlinear inertia weight variation for dynamic adaptation in particle swarm optimization (NDWPSO) was presented to solve the problem that it easily stuck at a local minimum point and its convergence speed is slow, when the linear decreasing inertia weight PSO (LDWPSO) adapt to the complex nonlinear optimization process. The rate of particle evolution changing was introduced in this new algorithm and the inertia weight was formulated as a function of this factor according to its impact on the search performance of the swarm. In each iteration process, the weight was changed dynamically based on the current rate of evolutionary changing value, which provides the algorithm with effective dynamic adaptability. The algorithm of LDWPSO and NDWPSO were tested with three benchmark functions. The experiments show that the convergence speed of NDWPSO is significantly superior to LDWPSO, and the convergence accuracy is improved.

This paper talks about the problems in particle swarm optimization (PSO), including local optimum and difficulty in improving solution accuracy by fine tuning. We presents a new variation of Adaptive Tribe-PSO model where nonlinear updating of inertia weight and a particle’s fitness with Tribe-PSO model are combined to improve the speed of convergence as well as fine tune the search in the multidimensional space. The method proved to be a powerful global optimization algorithm.

The continuous PSO algorithm has been widely researched and also applied as an intelligent computational technique to solve problems requiring iterative solutions based on some predefined objective function. However, the research on binary version of PSO (DBPSO) is still underway. The major research concerns are to accelerate the convergence speed retaining the search ability and reliability of the algorithm. To achieve this, a novel hybrid binary particle swarm optimization (HBPSO) algorithm is proposed in this paper. It combines the PSO’s concept and GA. In the existing standard binary PSO (DBPSO) two new operators such as crossover and mutation are incorporated to accelerate the convergence speed and to avoid possible stuck in local optimum thereby maintaining population diversity. The proposed HBPSO algorithm has been studied on 6 bench mark optimization problems. The experimental results such as minimum fitness, mean fitness, and variance of fitness over 50 consecutive trials on each objective function indicate that the HBPSO algorithm consistently outperforms the DBPSO and its variants in terms of convergence speed and search accuracy on a bulk of bench mark problems.

By the penalty function method we transform zero-one nonlinear programming problems into unconstrained zero-one integer optimization problems. A particle swarm optimization algorithm with chaos and gene density mutation is given to solve unconstrained the zero-one nonlinear program problems. We use chaos to initialize populations and use the 0-1 integer operation in updating positions to produce 0-1 integer points. We use the fitness variance and gene density strategy to determine whether the population premature phenomenon or not. If it appears that we use the gene density mutation to increase the population diversity or restart and reset the population by chaos technique. Numerical simulations show that the proposed algorithm for most test functions is feasible, effective and has high precision.

Particle Swarm Optimization (PSO) is a metaheuristic that is highly used to solve mono- and multi-objective optimization problems. Two well-differentiated PSO versions have been defined - one that operates in a continuous solution space and one for binary spaces. In this paper, a new version of the Binary PSO algorithm is presented. This version improves its operation by a suitable positioning of the velocity vector. To achieve this, a new modified version of the continuous gBest PSO algorithm is used. The method proposed has been compared with two alternative methods to solve four known test functions. The results obtained have been satisfactory.

Many real world optimization problems are dynamic in which global optimum and local optimum change over time. Particle swarm optimization has performed well to find and track optimum in dynamic environments. In this paper, we propose a new particle swarm optimization algorithm for dynamic environments. The proposed algorithm utilizes FCM to adapt exclusion radios and utilize a local search on best swarm to accelerate progress of algorithm and adjust inertia weight adaptively. To improve the search performance, when the search areas of two swarms are overlapped, the worse swarms will be removed. Moreover, in order to track quickly the changes in the environment, all particles in the swarm convert to quantum particles when a change in the environment is detected. Experimental results on different dynamic environments modeled by moving peaks benchmark show that the proposed algorithm outperforms other PSO algorithms, for all evaluated environments.

Premature convergence when solving multimodal problems is still the main limitation which affects the performance of the PSO. To avoid of premature, an improved PSO algorithm with an adaptive updating mechanism (IPSO) is proposed in this paper. When the algorithm converges to a local optimum, the updating mechanism begins to work so that the stagnated algorithm obtains energy for optimization. That is, the updating mechanism refreshes the swarm and expands the range for exploration. In this way, the algorithm can achieve a good balance between global exploration and local exploitation by the combination of the basic PSO evolution and updating mechanism. The proposed method is tested with a set of 10 standard optimization benchmark problems and the results are compared with those obtained through other 4 existing PSO algorithms. The simulation results elucidate that the proposed method produces the near global optimal solution, especially for those complex multimodal functions whose solution is difficult to be found by the other 4 algorithms. It is also observed from the comparison the IPSO is capable of producing a quality of optimal solution with faster rate.

Born and death is the nature of lives, but most swarm intelligence algorithm did not reflect this important property. Based on Particle Swarm Optimization, the concept of life span is introduced to control the activity generation of particles. Furthermore, the differential operator is applied to enhance the convergence and precision. The performance of propose algorithm, along with PSO and DE, is tested on benchmark functions. Results show that life span and differential operator greatly improved PSO and with well-balanced exploration and exploitation characteristic.

In this paper, we propose a particle swarm optimization (PSO) based hybrid testing technique named as “pseudo dynamic testing” to generate test data for C programs by fulfilling one of the most demanding test adequacy criteria: the all-path testing criterion using an interpreter. The proposed methodology attempts to solve many of the structural testing problems such as dynamic variables, input dependent array index, abstract function calls, infeasible paths and loop handling. The key algorithms and heuristics are given which are easy enough to implement, scalable and effective. The technique is employed on real world programs to show the robustness of this technique. The set of test inputs generated are not redundant as each leads to a different program path.

The optimization of reactive power compensation plays an important role in power system planning and designing. A mathematical model in the 10kV distribution network is established in this paper. Its objective function is the cost of investment in equipment of reactive power compensation and active power loss of the system should be the least. The node voltages beyond limited and the generator reactive power output beyond limited will be expressed in the way of penalty function. In this paper, particle swarm optimization will be used. Using PSO’s characteristic of high convergence efficiency, the speed of reactive power optimization will be improved. Using the binary PSO, the algorithm can better adapt to solve the problem.

An attempt has been made in this paper to solve the non-linear and ill-posed Electrical Impedance Tomography (EIT) inverse problem using clustering-based particle swam optimization (PSO). To enhance optimal search capability in such an ultra high dimension space and improve the quality of the reconstructed image, an adaptive PSO algorithm combined with a modified Newton–Raphson algorithm and a conductivity-based clustering algorithm was proposed. The modifications are performed on the reduction of dimension by dividing all mesh into clusters and initializing particles using the result of the modified Newton–Raphson type algorithm. Numerical simulation results indicated that the proposed method has a faster convergence to optimal solution and higher spatial resolution on a reconstructed image.

A robust optimization approach is proposed to solve the problem of supply chain collaboration under a demand uncertain environment. The proposed approach is universal and able to adapt to various demand models. First, the uncertain demand is described by a set of sampling sequences, and the total cost of supply chain is calculated based on these sequences to evaluate a collaboration scheme. Then a particle swarm optimization (PSO) is employed to find the optimal collaboration scheme which leads to a minimum total cost of supply chain. Numerical experiments show that the proposed approach can produce a robust solution that is insensible to the effect of demand uncertainty.

An optimal evacuation route plan has to be established to overcome the problem of poor coordination and uneven distribution of vehicles before or during disaster. This article introduces the evacuation vehicle routing problem (EVRP) as a new variant to the vehicle routing problem (VRP). EVRP is a process of moving vehicles from a vehicle location to the potentially flooded area (PFA), and from PFA to relief center using a number of capacitated vehicles. This paper examines the application of a multi-valued discrete particle swarm optimization (DPSO) for routing of vehicles from vehicle location to PFA. A solution representation is adopted and modified from the solution of the shortest path problem (SPP) to accommodate this problem. Experimental results were tested based on the objective function of finding a minimum total travelling time using datasets from a flash flood evacuation operation. DPSO was found to yield better results than a genetic algorithm (GA).

Process parameter window selection in semiconductor manufacturing field is usually the problem to find out the ranges of input parameters that meet production requirements, which requires allocating optima of a multimodal function efficiently. To achieve good results under the conditions of multimodal model and process control requirement, a NichePSO algorithm based method for parameter window selection is presented in this paper. Both simulation results and production validation data indicate it is an effective method for process parameter window selection.

Location based services are rapidly gaining popularity in various mobile applications. Such services rely particularly on the capability to accurately determine the location of the user. Several techniques are already available to provide localization for static or mobile applications, GPS being the most popular. However, due to some limitations of GPS such as low accuracy, unavailability in indoor environments and lower signal quality in urban areas with high rise buildings, complementary solutions are essential to offer satisfactory service at all places all the time. This paper demonstrates the use of a widely available WiFi networking infrastructure for accurate and low-cost indoor localization. Most existing WiFi-based localization approaches employ radio signal strength indicator (RSSI) fingerprinting technique, which requires a great deal of pre-deployment effort. Our swarm-inspired optimization algorithm applies a simpler and efficient technique based on the radio propagation model of the wireless signal. The proposed technique is evaluated in simulation and is demonstrated to achieve excellent average localization error of about 4 meters in an area of 50 x 50 square meters, under noisy RSSI measurements.

An improved method for embedding a secret message into a cover image with least significant bit (LSB) substitution in discrete cosine transformation (DCT) domain was proposed. The secret message was first split into partitions, while the cover image was divided into blocks of size 2x2, and DCT was used to convert the blocks from spatial domain to frequency domain. Then, Particle Swarm Optimization (PSO) algorithm was applied to search for an optimal substitution matrix T to transform the split partitions for an optimal embedding. Next, the transformed part of secret message was embedded into the AC coefficients of the transformed image blocks by LSB substitution. Experimental results show the proposed method can keep the quality of the stego-image better, while the security of the hidden secret message is increased by use of the substitution matrix T.

In this paper, a novel numerical integration method based on Particle Swarm Optimization (PSO) was presented. PSO is a technique based on the cooperation between particles. The exchange of information between these particles allows to resolve difficult problems. This approach is carefully handled and tested with some numerical examples.

A VSD system, which consists of an inverter & an induction motor, is now widely used in all kinds of application. But from the view point of an end user, neither the motor parameters in the mathematics model nor the vector controller structure are known. In this paper a PSO algorithm is programmed with IEC61131-3 language to estimate the parameters for the VSD system, based on the hardware of a vector controlled inverter, in order to reach the similar dynamic performance as a DC motor system. The PSO algorithm could be a kind of alternative approach of present parameter identification functions, for its requirements on the speed of CPU and volume of memory are low, while it converges quickly. It’s especially helpful for the adjustment of complicated control system, when the technical requirements are clear & measurable.

The Emergency Evacuation Simulation (EES) has been increasingly becoming a hotspot in the field of transportation. PSO-based EES is a good choice as its low computation complexity compared with some other algorithms, especially in an emergency. The selection of fitness function of each particle in PSO is a key problem for EES. This paper will introduce some fitness functions for EES and present a new fitness function called Triple-Distance Safe Degree (TDSD). Through theoretical analysis and experimental validation, the TDSD is proved to be much better than other fitness functions introduced in this paper.

Meta-heuristic algorithms inspired by nature have been used in a wide range of optimization problems. These types of algorithms have gained popularity in the field of artificial neural networks (ANN). On the other hand, spiking neural networks are a new type of ANN that simulates the behaviour of a biological neural network in a more realistic manner. Furthermore, these neural models have been applied to solve some pattern recognition problems. In this paper, it is proposed the use of the particle swarm optimization (PSO) algorithm to adjust the synaptic weights of a spiking neuron when it is applied to solve a pattern classification task. Given a set of input patterns belonging to

K

classes, each input pattern is transformed into an input signal. Then, the spiking neuron is stimulated during

T

ms and the firing rate is computed. After adjusting the synaptic weights of the neural model using the PSO algorithm, input patterns belonging to the same class will generate similar firing rates. On the contrary, input patterns belonging to other classes will generate firing rates different enough to discriminate among the classes. At last, a comparison between the PSO algorithm and a differential evolution algorithm is presented when the spiking neural model is applied to solve non-linear and real object recognition problems.

In this paper, we propose a hybridization between an ant-based clustering algorithm: CAC (Communicating Ants for Clustering) algorithm [5] and a clustering aggregation algorithm: the Furthest algorithm [6]. The CAC algorithm takes inspiration from the sound communication properties of real ants. In this algorithm, artificial ants communicate directly with each other in order to achieve the clustering task. The Furthest algorithm takes as inputs

m

clusterings given by

m

different runs of the CAC algorithm, and tries to find a clustering that matches, as possible, all the clusterings given as inputs. This hybridization shows an improvement of the obtained results.

This paper presents a multi-cellular-ant algorithm for large scale capacitated vehicle routing problem with restrictive vehicle capability. The problem is divided into corresponding smaller ones by a decomposition methodology. Relative relationship between subsystems will be solved through cooperative performance among cellular ants to avoid trivial solutions. The empirical results composed with adaptive ant colony algorithm and traditional collaboration show more efficiency and availability.

In this paper, we propose a fast and novel probabilistic fiber tracking method for Diffusion tensor imaging (DTI) data using the ant colony tracking technique, which considers both the local fiber orientation distribution and the global fiber path in collaborative manner. We first construct a global optimization model that captures both global fiber path and the uncertainties in local fiber orientation. Then, a global fiber tracking algorithm is presented using a novel learning strategy where the probability associated with a fiber is iteratively maximized. Finally, the proposed algorithm is validated and compared to alternative methods using both synthetic and real data.

To obtain the best food source, bees communicate their forage information by waggle dance, which indicates direction, distance, and quality of the food source they found. In this paper we propose a multi-function routing algorithm (BMFR) inspired by bees’ foraging behaviors for network-on-chip (NoC). We utilize a bee agent model to exchange the states among nodes. According to these states we establish a probability model to choose the output direction. We analyze the performance of BMFR on uniform traffic pattern. Finally, we compare BMFR with XY routing algorithm on uniform and tornado traffic patterns.

A new mapping algorithm is proposed based on Artificial Bee Colony (ABC) model to solve the problem of energy aware mapping optimization in Network-on-Chip (NoC) design. The optimal mapping result can be achieved by transmission of the information among various individuals. The comparison of the proposed algorithm with Genetic Algorithm (GA) and Max-Min Ant System (MMAS) based mapping algorithm shows that the new algorithm has lower energy consumption and faster convergence rate. Simulations are carried out and the results show the ABC based method could save energy by 15.5% in MMS, 5.1% in MPEG-4 decoder and 12.9% in VOPD compared to MMAS, respectively.

Resource constrained project scheduling (RCPSP) is one of the most crucial problems in project problem. The aim of RCPSP, which is NP-hard, is to minimize the project duration. Sometimes the activity durations are not known in advance and are random variables. These problems are called stochastic resource constrained project scheduling problems or stochastic RCPSP. Various algorithms such as genetic algorithm and GRASP have been applied on stochastic RCPSP. Bee algorithm is a metaheuristic based on the intelligent behavior of honey bee swarms. The goal of this study is adopting the artificial bee colony (ABC) algorithm to solve stochastic RCPSP and investigating its performance on the stochastic RCPSP. Simulation results show that proposed algorithm is an effective method for solving the stochastic resource constrained project scheduling problem. With regard to the problems with high distribution variability, the ABC algorithm is more effective than the other algorithms in the literature.

Human being is the most intelligent animal in this world. Intuitively, optimization algorithm inspired by human being creative problem solving process should be superior to the optimization algorithms inspired by collective behavior of insects like ants, bee,

etc

. In this paper, we introduce a novel brain storm optimization algorithm, which was inspired by the human brainstorming process. Two benchmark functions were tested to validate the effectiveness and usefulness of the proposed algorithm.

Human Group Optimization (HGO) algorithm, derived from the previously proposed seeker optimization algorithm (SOA), is a novel swarm intelligence algorithm by simulating human behaviors, especially human searching/foraging behaviors. In this paper, a canonical HGO with local search (L-HGO) is proposed. Based on the benchmark functions provided by CEC2005, the proposed algorithm is compared with several versions of differential evolution (DE) algorithms, particle swarm optimization (PSO) algorithms and covariance matrix adaptation evolution strategy (CMA-ES). The simulation results show that the proposed HGO is competitive or, even, superior to the considered other algorithms for some employed functions.

For improving the convergence of Cat Swarm Optimization (CSO), we propose a new algorithm of CSO namely, Average-Inertia Weighted CSO (AICSO). For achieving this, we added a new parameter to the position update equation as an inertia weight and used a new form of the velocity update equation in the tracing mode ofalgorithm. Experimental results using Griewank, Rastrigin and Ackley functions demonstrate that the proposed algorithm has much better convergence than pure CSO.

Under the given system weight constraint, we consider the problem of maximizing the system lifetime and minimizing the system cost. The lifetimes of components in the system are characterized by type-2 fuzzy variables. The numbers of redundant elements of each components are the decision variables. We use the reduction methods to reduce the type-2 fuzzy lifetimes. Then, we propose a goal programming model for this system. We suggest an approximation approach (AA) to the reliability and design an AA-based particle swarm optimization (PSO) algorithm to solve the fuzzy model.

A population-based algorithm, oriented search algorithm (OSA), is proposed to optimize functions in this paper. In OSA, the search-individual imitates human random search behavior, and the search-object simulates an intelligent agent that can transmit oriented information to search-individuals. OSA is tested on thirteen complex benchmark functions. The results are compared with those of particle swarm optimization with inertia weight (PSO-w), particle swarm optimization with constriction factor (PSO-cf) and comprehensive learning particle swarm optimizer (CLPSO). The results show that OSA is superior in convergence efficiency, search precision, convergence property and has the strong ability to escape from the local sub-optima.

Indirect reciprocity is a key mechanism for the evolution of human cooperation. There are normally two choices in the standard model of indirect reciprocity which works through reputation. Here we introduced the role of costly punishment into the model. The players could have the third choice besides cooperation and defection. The dynamics of cooperation in indirect reciprocity is analyzed under the social norms which depend on the action of the donor and the reputation of the recipient. It is found that those strategies using costly punishment which allow the evolutionary stability of cooperation typically reduce the average payoff of the population and there is only a small parameter region where costly punishment is evolutionary stable and more efficient. The computer simulations based on agent in finite populations are performed and the result is agreement with our theoretical predictions.

A collaborative optimization algorithm under a control framework is developed for the asymmetric traveling salesman problem (ATSP). The collaborative approach is not just a simple combination of two methods, but a deep collaboration in a manner like the feedback control. A notable feature of the approach is to make use of the collaboration to reduce the search space while maintaining the optimality. Compared with the previous work of the reduction procedure by Carpaneto, Dell’Amico et al. (1995) we designed a tighter and more generalized reduction procedure to make the collaborative method more powerful. Computational experiments on benchmark problems are given to exemplify the approach.

Dynamic composition and reconfiguration of service-oriented Internetware systems are of paramount importance as we can not pre-define everything during the design time of a software system. Recent biology studies show that the slime mold

Physarum polycephalum

– a single-cell organism – can form a veined network that explores the available space and connects food sources in the absence of central control mechanisms. Inspired by the formation and behavior of such biological adaptive networks, a new bionic approach is proposed for dynamic service composition and reconfiguration of Internetware systems. Simulation experiments were conducted. The experimental results show that the proposed approach is effective and efficient. It is hoped that this paper will shed new light in Internetware system design and construction.

In this paper, we propose a novel search interval forecasting (SIF) optimization algorithm for global numerical optimization. In the SIF algorithm, the information accumulated in the previous iteration of the evolution is utilized to forecast area where better optimization value can be located with the highest probability for the next searching operation. Five types of searching strategies are designed to accommodate different situations, which are determined by the history information. A suit of benchmark functions are used to test the SIF algorithm. The simulation results illustrate the good performance of SIF, especially for solving large scale optimization problems.

This paper proposes a Danger Theory (DT) based learning (DTL) model for combining classifiers. Mimicking the mechanism of DT, three main components of the DTL model, namely signal I, danger signal and danger zone, are well designed and implemented so as to define an immune based interaction between two grounding classifiers of the model. In addition, a self-trigger process is added to solve conflictions between the two grounding classifiers. The proposed DTL model is expected to present a more accuracy learning method by combining classifiers in a way inspired from DT. To illustrate the application prospects of the DTL model, we apply it to a typical learning problem — e-mail classification, and investigate its performance on four benchmark corpora using 10-fold cross validation. It is shown that the proposed DTL model can effectively promote the performance of the grounding classifiers.

The interest of hybridizing different nature inspired algorithms has been growing in recent years. As a relatively new algorithm in this field, Biogeography Based Optimization(BBO) shows great potential in solving numerical optimization problems and some practical problems like TSP. In this paper, we proposed an algorithm which combines Biogeography Based Optimization (BBO) and Clonal Selection Algorithm (BBOCSA). Several benchmark functions are used for comparison among the hybrid and other nature inspired algorithms (BBO, CSA, PSO and GA). Simulation results show that clone selection can enhance the ability of exploration of BBO and the proposed hybrid algorithm has better performance than the other algorithms on some benchmarks.

Biogeography-based optimization algorithm(BBO) is a new kind of optimization algorithm based on Biogeography. It mimics the migration strategy of animals to solve the problem of optimization. In this paper, the clone selection strategy is combined with biogeography for solving the problem of sensors selection of aircraft. It is compared with other classical nature inspired algorithms. The comparison results show that BBOCSA is an effective algorithm for optimization problem in practice. It provides a new method for this kind of problem.

focusing on “distinctiveness” and “effectiveness” of algorithms inspired by natural immune system, a novel artificial immune network algorithm named immune feature extracting network (IFEN) is proposed to realize the function of feature data extracting in this paper. Based on comprehensive analysis of mechanism of natural immune system and current researching works of artificial immune system (AIS), a modified paradigm of artificial immune network (AIN) and a new mutation operation are designed to adapt to decrease the size of sample data set and extract feature data from the data set with noise. The proposed algorithm is supposed to be used as a data preprocessing method with functions of data compression and data cleansing. Preliminary experiments show that the quality of the data set processed by IFEN is apparently improved and the size is compressed.

Differential Evolution (DE) algorithm is a successful optimization method in continuous space and has been successfully applied in many different areas. The operators used in DE are simple, however, the mechanism in which the operators are defined, makes it impossible to apply the standard DE directly to the problems in binary space. A novel binary encoding DE (BDE) was proposed to extend DE for solving the optimization problems in binary space. A mixed expression, which constitutes of an arithmetical expression and a logical expression, was used to construct a new mutation operator. And then with a predefined probability, the result of the mutation operator was flipped. Initial experiment results indicate the novel BDE is useful and effective.

Differential Evolution algorithm is a simple yet reliable and robust evolutionary algorithm for numeric optimization. However, fine-tuning control parameters of DE algorithm is a tedious and time-consuming task thus became a major challenge for its application. This paper introduces a novel self-adaptive method for tuning the amplification parameters

F

of DE dynamically. This method sampled appropriate

F

value from a probabilistic model build on periodic learning experience. The performance of proposed MSDE is investigated and compared with other state-of-art self-adaptive approaches. Moreover, the influence of learning frequency of MSDE is investigated.

Differential evolution is a powerful evolution algorithm for optimization of real valued and multimodal functions. To accelerate its convergence rate and enhance its performance, this paper introduces a top-p-best trigonometric mutation strategy and a self-adaptation method for controlling the crossover rate (

CR

). The performance of the proposed algorithm is investigated on a comprehensive set of 13 benchmark functions. Numerical results and statistical analysis show that the proposed algorithm boosts the convergence rate yet maintaining the robustness of the DE algorithm.

During the past decade, the particle swarm optimization (PSO) with various versions showed competitiveness on the constrained optimization problems. In this paper, an improved Gaussian particle swarm optimization algorithm (GPSO) is proposed to improve the diversity and local search ability of the population. A mutation operator based on differential evolution (DE) is designed and employed to update the personal best position of the particle and the global best position of the population. The purpose is to improve the local search ability of GPSO and the probability to find the global optima. The regeneration strategy is employed to update the stagnated particle so as further to improve the diversity of GPSO. A simple feasibility-based method is employed to compare the performances of different particles. Simulation results of three constrained engineering optimization problems demonstrate the effectiveness of the proposed algorithm.

Due to their efficiency and adaptability, bio-inspired algorithms have shown their usefulness in a wide range of different non-linear optimization problems. In this paper, we compare two ways of training an artificial neural network (ANN): Particle Swarm Optimization (PSO) and Differential Evolution (DE) algorithms. The main contribution of this paper is to show which of these two algorithms provides the best accuracy during the learning phase of an ANN. First of all, we explain how the ANN training phase could be seen as an optimization problem. Then, we explain how PSO and DE could be applied to find the best synaptic weights of the ANN. Finally, we perform a comparison between PSO and DE approaches when used to train an ANN applied to different non-linear problems.

In the last few years bio-inspired neural networks have interested an increasing number of researchers. In this paper, a novel approach is proposed to solve the problem of identifying the topology and parameters in Hindmarsh-Rose-neuron networks. The approach introduces generalized extremal optimization (GEO), a relatively new heuristic algorithm derived from co-evolution to solve the identification problem. Simulation results show that the proposed approach compares favorably with other heuristic algorithms based methods in existing literatures with smaller estimation errors. And it presents satisfying results even with noisy data.

In this paper, the delay-dependent stability problem for a class of neural networks of neutral-type with interval time-varying delays and nonlinear perturbations is investigated. A novel stability criterion is obtained in terms of linear matrix inequality (LMI) by employing a Lyapunov-Krasovskii functional. The proposed criteria can be checked easily by the LMI Control Toolbox in Matlab. In addition, two examples are given to show the effectiveness of the obtained result.

Air pollution time series is often characterized as chaotic in nature. The prediction using traditional statistical techniques and artificial neural network with back-propagation (BP) algorithm, which is most widely applied, do not give reliable prediction results. The new algorithm is therefore proposed to predict the chaotic time series based on the generalized Chebyshev neural network technique. In addition, the new algorithm has no problems such as local minima, slow convergence arising from the steepest descent-like algorithm. Finally, to illustrate the power of the Chebyshev Neural Network (CNN), a simulation example is presented to show good performance that extracts useful information from the weight functions for understanding relations inherent in the given patterns, and the trained CNN has good performance both on generalization and calculating precision.

Financial time series has been standard complex problem in the field of forecasting due to its non-linearity and high volatility. Though various neural networks such as back propagation, radial basis, recurrent and evolutionary etc. can be used for time series forecasting, each of them suffer from some flaws. Performances are more varied for different time series with loss of generalization. Each of the method poses some pros and cons for it. In this paper, we use ensembles of neural networks to get better performance for the financial time series forecasting. For neural network ensemble four different modules has been used and results of them are finally integrated using integrator to get the final output. Gating has been used as integration techniques for the ensembles modules. Empirical results obtained from ensemble approach confirm the out performance of forecast results than single module results.

Software reliability models are used for the estimation and prediction of software reliability. In a situation where reliability data is lacking and numerous models are available, the key to quantitative analysis of software reliability lies in the selection of an optimal model. This paper describes a model selection method which involves an encoding scheme with multiple evaluation metrics and uses back-propagation (BP) neural network to perform clustering algorithm. Finally, by utilizing 20 sets of failure data that are collected in actual software development projects, a simulation experiment is made. The result shows the method is both correct and feasible.

Atavistic evolutionary strategy for genetic algorithm is put forward according to the atavistic phenomena existing in the process of biological evolution, and the framework of the new strategy is given also. The effectiveness analysis of the new strategy is discussed by three characteristics of the reproduction operators. The introduction of atavistic evolutionary strategy is highly compatible with the minimum induction pattern, and increases the population diversity to a certain extent. The experimental results show that the new strategy improves the performance of genetic algorithm on convergence time and solution quality.

This paper presents an improved co-evolution genetic algorithm (ICGA), which uses the methodology of game theory to solve the mode deception and premature convergence problem. In ICGA, groups become different players in the game. Mutation operator is designed to simulate the situation in the evolutionary stable strategy. Information transfer mode is added to ICGA to provide greater decision-making space. ICGA is used to solve large-scale deceptive problems and an optimal control problem. Results of numerical tests validate the algorithm’s excellent performance.

This paper proposed an improved decomposition approach for structuring elements of arbitrary shape. For the model of this method, we use the recursive model which decomposes a given structuring element into a variable-size matrix dilated by a fixed-size matrix and with union of a residue component, such procedures repeated until the variable-size matrix is smaller than a predefined threshold. For the algorithm of our method, we proposed an improved GA based on the ring topology of migration model and the power-rank fitness scaling strategy. The experiments demonstrate that our method is more robust than Park’s method, Anelli’s method, and Shih’s method, and gave the final decomposition tree of different SE shapes such as the letter “V”, heart, and umbrella.

Minimizing computing energy consumption has many benefits, such as environment protection, cost savings, etc. An important research problem is the energy aware task scheduling for cloud computing. For many diverse computers in a typical cloud computing system, great energy reduction can be achieved by smart optimization methods. The objective of energy aware task scheduling is to efficiently complete all assigned tasks to minimize energy consumption with various constraints. Genetic Algorithm (GA) is a popular and effective optimization algorithm. However, it is much slower than other traditional search algorithms such as heuristic algorithm. In this paper, we propose a shadow price guided algorithm (SGA) to improve the performance of energy aware task scheduling. Experiment results have shown that our energy aware task scheduling algorithm using the new SGA is more effective and faster than the standard GA.

Crossing minimization problem in a bipartite graph is a well-known NP-Complete problem. Drawing the directed/undirected graphs such that they are easy to understand and remember requires some drawing aesthetics and crossing minimization is one of them. In this paper, we investigate an intelligent evolutionary technique i.e. Genetic Algorithm (GA) for bipartite drawing problem (BDP). Two techniques GA1 and GA2 are proposed based on Genetic Algorithm. It is shown that these techniques outperform previously known heuristics e.g., MinSort (M-Sort) and BaryCenter (BC) as well as a genetic algorithm based level permutation problem (LPP), especially when applied to low density graphs. The solution is tested over various parameter values of genetic bipartite drawing problem. Experimental results show the promising capability of the proposed solution over previously known heuristics.

In many challenging numerical optimization problems, the conflict between exploitation and exploration abilities of EAs must be balanced in an effective and efficient way. In the previous research, in order to address this issue, the Two-Stage ensemble Evolutionary Algorithm (TSEA) was originally proposed for engineering application. In TSEA, the optimization is divided into two relatively separate stages, which aims at handling the exploitation and exploration in a more reasonable way. In this paper, we try to extend the application area of TSEA from specific engineering problems to general numerical optimization problems by altering its sub-optimizers. The experimental studies presented in this paper contain three aspects: (1) The benefits of the TSEA framework are experimentally investigated by comparing TSEA with its sub-optimizers on 26 test functions; then (2) TSEA is compared with diverse state-of-the-art evolutionary algorithms (EAs) to comprehensively show its advantages; (3) To benchmark the performance of TSEA further, we compare it with 4 classical memetic algorithms (MAs) on CEC05 test functions. The experimental results definitely demonstrate the excellent effectiveness, efficiency and reliability of TSEA.

In this paper, we propose an applicable genetic programming approach to solve the problems of binary image analysis and gray scale image enhancement. Given a section of original image and the corresponding goal image, the proposed algorithm evolves for generations and produces a mathematic morphological operation sequence, and the result performed by which is close to the goal. When the operation sequence is applied to the whole image, the objective of image analysis is achieved. In this sequence, only basic morphological operations— erosion and dilation, and logical operations are used. The well-defined chromosome structure leads brings about more complex morphological operations can be composed in a short sequence. Because of a reasonable evolution strategy, the evolution effectiveness of this algorithm is guaranteed. Tested by the binary image features analysis, this algorithm runs faster and is more accurate and intelligible than previous works. In addition, when this algorithm is applied to infrared finger vein gray scale images to enhance the region of interest, more accurate features are extracted and the accuracy of discrimination is promoted.

This work develops a new fuzzy version of single-source capacitated facility location problem (FLP), in which a set of capacitated facilities is selected to provide service to demand points with possibility distributions at the minimal total cost. Since the proposed FLP includes credibility service level constraints and 0–1 decision variables, its solution method is a challenge issue for research, and usually relies on metaheuristics and approximation approach. However, for frequently used trapezoidal, Gamma and Normal fuzzy demands, the FLPs are equivalent to deterministic 0-1 programming problems. As a consequence, the equivalent 0-1 programming problems can be solved by general purpose software or conventional optimization algorithms. At the end of this paper, we demonstrate the developed modeling idea via numerical experiments.

Based on granular spaces, some relational problems with fuzzy equivalence relations is studied, and three results are obtained as follows. Firstly, the dynamic property of a fuzzy equivalence relation on its granular space is discussed. Secondly, the ordering relationship between fuzzy equivalence relations and their granular spaces is researched, and they are order-preserving. Furthermore, the collaborative clustering of fuzzy equivalence relations on granular spaces by their intersection operation is given, which the collaborative clustering derived from the fuzzy equivalence relations obtained by the intersection operation is a thinner or more precise consistent cluster. These conclusions will help us pursue an even deeper understanding of the essence of granular computing.

Data fusion using diverse biological data has been applied to predict the protein function in recent years. In this paper, fuzzy integral fusion based on fuzzy measure is used to integrate the probabilistic outputs of different classifiers. Support vector machines as base learners are applied to predict the functions of examples from each data source. Fuzzy density values are determined by Particle Swarm Algorithm and an improved

λ

-measure is used. We compare our improved fuzzy measure to typical one. The experimental results show that our method has the better results.

K-means is one of the most commonly used clustering methods for analyzing gene expression data, where it is sensitive to the choice of initial clustering centroids and tends to be trapped in local optima. To overcome these problems, a memetic K-means (MKMA) algorithm, which is a hybridization of particle swarm optimizer (PSO) based memetic algorithm (MA) and K-means, is proposed in this paper. In particular, the PSO based MA is used to minimize the within-cluster sum of squares and the K-means is used to iteratively fine-tune the locations of the centers. The experimental results on two gene expression datasets indicate that MKMA is capable of obtaining more compact clusters than K-means, Fuzzy K-means, and the other PSO based K-means namely PK-means. MKMA is also demonstrated to attain faster convergence rate and more robustness against the random choice of initial centroids.

This paper presents the implementation of a Hybrid Particle Swarm Optimization with Biased Mutation (HPSOBM) algorithm to solve the load flow computation in electrical power systems. The load flow study obtains the system status in the steady-state and it is widely used in the power system operation, planning and control. The proposed methodology is applied in a different computational model, which is based on the minimization of the power mismatches in the system buses. This new model searches for a greater convergence, and also a larger application in comparison with traditional numerical methods. In order to illustrate the proposed algorithm some simulations were conducted using the IEEE 14 bus system.

Avoiding collision of nano-particles during manipulation operations and selecting the best route and lowest Atomic Force Microscopy (AFM) movement are major concerns in the area of nano-space. To apply the lowest force on the cantilever from fluid environment forces, we try to minimize AFM movements. Our proposed method calculates the optimum routing for AFM probe movement for nano-particles transmission using hybrid GA (Genetic algorithm) and PSO-AS (Particle Swarm Optimization- Ant System) simulates it in various type of medium. We consider the collision of the probe with minor barriers. An optimized AFM path minimizes the time and energy required for nano-particle manipulation. For movement of the nano-particles, we seek an efficient probe pattern. A second goal is to transfer the nano-particles without undesired collision. The optimum routing method will increase the speed of the process. Our proposed model, utilizes both Mathematical and Matlab software to simulate the process.

Exchange rate fluctuation has a significant effect on the risk of marketing business, economic development and financial stability. Accurate prediction for exchange rate may reduce commercial and economic risk arisen by exchange rate fluctuation. In this study, we propose an intelligent approach to the forecasting problem of the CNY-USD exchange rate, where a neuro-fuzzy self-organizing system is used as the intelligent predictor. For learning purpose, a novel hybrid learning method is devised for the intelligent predictor, where the well-known particle swarm optimization (PSO) algorithm and the recursive least squares estimator (RLSE) algorithm are involved. The proposed learning method is called the PSO-RLSE-PSO method. Experiments for time series forecasting of the CNY-USD exchange rate are conducted. For performance, the intelligent predictor is trained by several different methods. The experimental results show that the proposed approach has excellent forecasting performance.

This paper provides a novel hybrid model to solve credit scoring problems. This model is based on RBF neural network with genetic algorithm and its principal character is that Central position, center spread and weights of RBF neural network are encode as genes of Chromosome in genetic algorithm. And then using genetic algorithm trains RBF neural network circularly. A real world credit dataset in the University of California Irvine Machine Learning Repository are selected for the experiment. Numerical experiment shows that the model possesses fast learning ability and excellent generalization ability, and verifies that the novel model has better preference.