Search

This study focuses on the improved initialization of initial centroids instead of random selection for the K-means algorithm. The random selection of initial seeds is a major drawback of the original Kmeans algorithm because it leads to less reliable result of clustering the data. The modified approach of the k-means algorithm integrates the computation of the weighted mean to improve the seeds initialization. This paper shows the comparison of K-Means and Modified K-Means algorithm, using the first simple dataset of four objects and the dataset for service vehicles. The two simple applications proved that the Modified K- Means of selecting initial centroids is more reliable than K-Means Algorithm. Clustering is better achieved in the modified k-means algorithm. Article DOI: http://dx.doi.org/10.20319/mijst.2016.22.4864 This work is licensed under the Creative Commons Attribution-Non-commercial 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.

This paper considers the identification problem of multi-input-output-error autoregressive systems. A hierarchical gradient based iterative (H-GI) algorithm and a hierarchical least squares based iterative (H-LSI) algorithm are presented by using the hierarchical identification principle. A gradient based iterative (GI) algorithm and a least squares based iterative (LSI) algorithm are presented for comparison. The simulation results indicate that the H-LSI algorithm can obtain more accurate parameter estimates than the LSI algorithm, and the H-GI algorithm converges faster than the GI algorithm.

The junction tree algorithm is currently the most popular algorithm for exact inference on Bayesian networks. To improve the time complexity of the junction tree algorithm, we need to find a triangulation with the optimal total table size. For this purpose, Ottosen and Vomlel have proposed a depth-first search (DFS) algorithm. They also introduced several techniques to improve the DFS algorithm, including dynamic clique maintenance and coalescing map pruning. Nevertheless, the efficiency and scalability of that algorithm leave much room for improvement. First, the dynamic clique maintenance allows to recompute some cliques. Second, in the worst case, the DFS algorithm explores the search space of all elimination orders, which has size n!, where n is the number of variables in the Bayesian network. To mitigate these problems, we propose an extended depth-first search (EDFS) algorithm. The new EDFS algorithm introduces the following two techniques as improvements to the DFS algorithm: (1) a new dynamic clique maintenance algorithm that computes only those cliques that contain a new edge, and (2) a new pruning rule, called pivot clique pruning. The new dynamic clique maintenance algorithm explores a smaller search space and runs faster than the Ottosen and Vomlel approach. This improvement can decrease the overhead cost of the DFS algorithm, and the pivot clique pruning reduces the size of the search space by a factor of O ( n 2 ) . Our empirical results show that our proposed algorithm finds an optimal triangulation markedly faster than the state-of-the-art algorithm does. A state-of-the-art algorithm for optimal triangulation is proposed.Evaluated the degree of impact for employing the total table size as optimality criterion.A novel pruning rule for triangulation algorithm is proposed.A fast dynamic clique maintenance algorithm is proposed.

In this paper the hybrid and modified versions of the PSO algorithm applied to improvement of the search characteristics of the classical PSO algorithm in the development problem of the SVM classifier have been offered and investigated. A herewith two hybrid versions of the PSO algorithm assume the use of the classical Grid Search (GS) algorithm and the Design of Experiment (DOE) algorithm accordingly, and the modified version of the PSO algorithm realizes the simultaneous search of the kernel function type, the parameters values of the kernel function, and also the regularization parameter value. Besides, the questions of applicability of the k nearest neighbors (kNN) algorithm in the development problem of the SVM classifier have been considered.

A semi-real-time pattern-matching algorithm consisting of an offline and online stage is proposed. The approach of the proposed algorithm is to perform a significant amount of the calculation required by pattern matching in the offline stage. This necessitates only a small amount of calculation in the online process to reject a great number of mismatched positions. The proposed algorithm first uses triangle inequality and orthogonal decomposition to derive the lower bounds of the distances between the pattern and the candidate windows of the base image. Then, mismatched candidate windows are rejected if their lower bounds exceed an adaptive threshold. The proposed method accelerates the online processing effectively while yielding the identical result as a full search algorithm. The proposed algorithm was compared with other state-of-the-art algorithms and the result confirms that the proposed algorithm has a distinct speed advantage over the other algorithms for online processing.

We introduced a new harmony memory initialization method.We introduced a new pitch adjustment method based on DE/best/1 mutation strategy.We comprehensively studied the parameter setting of our algorithm.We compared our algorithm with the state of the art variants of HS algorithm.We compared our algorithm with the state of the art variants of DE algorithm. The performance of the Harmony Search (HS) algorithm is highly dependent on the parameter settings and the initialization of the Harmony Memory (HM). To address these issues, this paper presents a new variant of the HS algorithm, which is called the DH/best algorithm, for the optimization of globally continuous problems. The proposed DH/best algorithm introduces a new improvisation method that differs from the conventional HS in two respects. First, the random initialization of the HM is replaced with a new method that effectively initializes the harmonies and reduces randomness. Second, the conventional pitch adjustment method is replaced by a new pitch adjustment method that is inspired by a Differential Evolution (DE) mutation strategy known as DE/best/1. Two sets of experiments are performed to evaluate the proposed algorithm. In the first experiment, the DH/best algorithm is compared with other variants of HS based on 12 optimization functions. In the second experiment, the complete CEC2014 problem set is used to compare the performance of the DH/best algorithm with six well-known optimization algorithms from different families. The experimental results demonstrate the superiority of the proposed algorithm in convergence, precision, and robustness.

This paper proposes a computationally efficient optimization algorithm for segmenting colour satellite images.CS algorithm incorporating Mantegna's and McCulloch's method for modeling levy flight is presented.PSO, DPSO, ABC and CS algorithms are compared with the proposed algorithm.All these optimization algorithms are exploited using three different objective functions.Performance assessment metrics demonstrated the improvement in the efficiency of the proposed algorithm. Satellite image segmentation is challenging due to the presence of weakly correlated and ambiguous multiple regions of interest. Several bio-inspired algorithms were developed to generate optimum threshold values for segmenting such images efficiently. Their exhaustive search nature makes them computationally expensive when extended to multilevel thresholding. In this paper, we propose a computationally efficient image segmentation algorithm, called CSMcCulloch, incorporating McCulloch's method for l e ? v y flight generation in Cuckoo Search (CS) algorithm. We have also investigated the impact of Mantegna's method for l e ? v y flight generation in CS algorithm (CSMantegna) by comparing it with the conventional CS algorithm which uses the simplified version of the same. CSMantegna algorithm resulted in improved segmentation quality with an expense of computational time. The performance of the proposed CSMcCulloch algorithm is compared with other bio-inspired algorithms such as Particle Swarm Optimization (PSO) algorithm, Darwinian Particle Swarm Optimization (DPSO) algorithm, Artificial Bee Colony (ABC) algorithm, Cuckoo Search (CS) algorithm and CSMantegna algorithm using Otsu's method, Kapur entropy and Tsallis entropy as objective functions. Experimental results were validated by measuring PSNR, MSE, FSIM and CPU running time for all the cases investigated. The proposed CSMcCulloch algorithm evolved to be most promising, and computationally efficient for segmenting satellite images. Convergence rate analysis also reveals that the proposed algorithm outperforms others in attaining stable global optimum thresholds. The experiments results encourages related researches in computer vision, remote sensing and image processing applications.

In this paper, we propose a distributed adaptive learning algorithm to train the coefficients of a widely linear autoregressive moving average model by measurements collected by the nodes of a network. We assume that each node uses the augmented complex adaptive infinite impulse response (ACA-IIR) filter as the learning rule, and nodes interact with each other under an incremental mode of cooperation. To derive the proposed algorithm, called the incremental ACAIIR (IACA-IIR), we firstly formulate the distributed adaptive learning problem as an unconstrained minimization problem. Then, we apply stochastic gradient optimization argument to solve it and derive the proposed algorithm. We further find the step size range where the stability of the proposed algorithm is guaranteed. We also introduce a reduced-complexity version of the IACA-IIR algorithm. Since the proposed algorithm relies on the augmented complex statistics, it can be used to model both types of complex-valued signals (proper and improper signals). To evaluate the performance of the proposed algorithm, we use both synthetic and real-world complex signals in our simulations. The results exhibit superior performance of the proposed algorithm over the non-cooperative ACA-IIR algorithm.

A* algorithm is widely used in the game for diameter, is currently one of the more popular heuristic search algorithm, but the algorithm has the problem of searching time and path. In this paper, a bidirectional search A* algorithm is proposed to improve the search efficiency and ensure the accuracy of the search, and effectively solve the problem of path twists and turns. Then, the algorithm is implemented by experiment simulation, which improves the effectiveness and feasibility of the algorithm in the large map.

This paper proposes a discrete-time learning algorithm for fast image restoration using a novel L2-norm noise constrained estimation. The noise constrained estimation approach can relax the need of the optimal regularization parameter to be estimated. Performance analysis shows that the proposed algorithm can converge globally to a robust optimal weight vector. Compared with the cooperative neural fusion (CNF) algorithm minimizing L1-norm estimation, the proposed algorithm only needs O(N) multiplication operationper iteration, instead of O ( N 2 ) multiplication operation required by the CNF algorithm. Moreover, the proposed fusion approach overcomes the difficulty of estimating the noise error set in the CNF approach. Simulation results show by comparison that under the non-optimal regularization parameter, the proposed algorithm can obtain a better restored estimate in Gaussian mixture noise and can run much faster compared to the CNF algorithm.

This paper proposes a hybrid bat algorithm with natural-inspired algorithms for continuous optimization problem. In this study, the proposed algorithm combines the reproduction step from weed algorithm and genetic algorithm. The reproduction step is applied to clone each bat population by fitness values and the genetic algorithm is applied in order to expand the population. The algorithm is evaluated on eighteen benchmark problems. The computational results of the proposed algorithm are compared with the methods in the literature which are self-adaptive differential evolution (DE), traditional DE algorithm, intersection mutation differential evolution (IMDE) algorithm, and the JDE self-adaptive algorithm. Findings show that the algorithm produces several solutions obtained by the previously published methods especially for the continuous unimodal function, the quartic function, the multimodal function and the discontinuous step function. In addition, the finding shows that the proposed algorithm can produce optimal solutions efficiently on benchmark instances within short computational time.

Fuzzy rule-based classification systems have been used extensively in data mining. This paper proposes a fuzzy rule- based classification algorithm based on a quantum ant optimization algorithm. A method of generating the hierarchical rules with different granularity hybridization is used to generate the initial rule set. This method can obtain an original rule set with a smaller number of rules. The modified quantum ant optimization algorithm is used to generate the optimal individual. Compared to other similar algorithms, the algorithm proposed in this paper demonstrates higher classification accuracy and a higher convergence rate. The algorithm is proved to be convergent on theory. Some experiments have been conducted on the algorithm, and the results proved that the algorithm is feasible.

This paper illustrates an enhanced algorithm based on one of the swarm intelligence techniques for constructing the Phylogenetic tree (PT), which is used to study the relationship between species. The main scheme is to formulate a PT, an NP- complete problem through an evolutionary algorithm called Artificial Bee Colony (ABC). The tradeoff between the accuracy and the computational time taken for constructing the tree makes way for new variants of algorithms. A new variant of ABC algorithm is proposed to promote the convergence rate of general ABC algorithm through recommending a new formula for searching solution. In addition, a searching step has been included so that it constructs the tree faster with a nearly optimal solution. Experimental results are compared with the ABC algorithm, Genetic Algorithm and the state-of-the-art techniques like unweighted pair group method using arithmetic mean, Neighbour-joining and Relaxed Neighbor Joining. For results discussion, we used one of the standard dataset Treesilla. The results show that the Enhanced ABC (EABC) algorithm converges faster than others. The claim is supported by a distance metric called the Robinson-Foulds distance that finds the dissimilarity of the PT, constructed by different algorithms.

In risk assessment and predictive policing, biased data can yield biased results.

We introduce a fast Branch-and-Bound algorithm for optimal feature selection based on a U-curve assumption for the cost function. The U-curve assumption, which is based on the peaking phenomenon of the classification error, postulates that the cost over the chains of the Boolean lattice that represents the search space describes a U-shaped curve. The proposed algorithm is an improvement over the original algorithm for U-curve feature selection introduced recently. Extensive simulation experiments are carried out to assess the performance of the proposed algorithm (IUBB), comparing it to the original algorithm (UBB), as well as exhaustive search and Generalized Sequential Forward Search. The results show that the IUBB algorithm makes fewer evaluations and achieves better solutions under a fixed computational budget. We also show that the IUBB algorithm is robust with respect to violations of the U-curve assumption. We investigate the application of the IUBB algorithm in the design of imaging W -operators and in classification feature selection, using the average mean conditional entropy (MCE) as the cost function for the search.

In the optimal experimental design literature, the G-optimality is defined as minimizing the maximum prediction variance over the entire experimental design space. Although the G-optimality is a highly desirable property in many applications, there are few computer algorithms developed for constructing G-optimal designs. Some existing methods employ an exhaustive search over all candidate designs, which is time-consuming and inefficient. In this paper, a new algorithm for constructing G-optimal experimental designs is developed for both linear and generalized linear models. The new algorithm is made based on the clustering of candidate or evaluation points over the design space and it is a combination of point exchange algorithm and coordinate exchange algorithm. In addition, a robust design algorithm is proposed for generalized linear models with modification of an existing method. The proposed algorithm are compared with the methods proposed by Rodriguez et al. [Generating and assessing exact G-optimal ...

In this paper, a self-learning genetic algorithm for solving combinatorial optimization problems is considered. The self-learning consists in changing the values of the probabilities of crossing and mutation in accordance with changing the value of the fitness function after operations in the next iteration of the algorithm. The results of comparing the proposed algorithm with the Holland algorithm by the problems of multiprocessor job scheduling and subset sum problem are presented.

In this paper the authors propose an adaptive estimation algorithm for in-network processing of complex signals over distributed networks. In the proposed algorithm, as the incremental augmented complex least mean square (IAC-LMS) algorithm, nodes of the network are allowed to collaborate via incremental cooperation mode to exploit the spatial dimension; while at the same time are equipped with LMS learning rules to endow the network with adaptation. The authors have extracted closed-form expressions that show how IAC-LMS algorithm performs in the steady-state. The authors further have derived the required conditions for mean and mean-square stability of the proposed algorithm. The authors use both synthetic benchmarks and real world non-circular data to evaluate the performance of the proposed algorithm. Simulation results also reveal that the IAC-LMS algorithm is able to estimate both second order circular (proper) and non-circular (improper) signals. Moreover, IAC-LMS algorithm outperforms the non-cooperative solution.

The problem of local minimum cannot be avoided when it comes to nonlinear optimization in the learning algorithm of neural network parameters, and the larger the optimization space is, the more obvious the problem becomes. This paper proposes a new type of hybrid learning algorithm for three-layered feed-forward neural networks. This algorithm is based on three-layered feed-forward neural networks with output layer function, namely linear function, combining a quasi Newton algorithm with adaptive decoupled step and momentum (QNADSM) and iterative least square method to export. Simulation proves that this hybrid algorithm has strong self-adaptive capability, small calculation amount and fast convergence speed. It is an effective engineer practical algorithm.

Ranking data points with respect to a given preference criterion is an example of a preference learning task. In this paper, we investigate the generalization performance of the regularized ranking algorithm associated with least square ranking loss in a reproducing kernel Hilbert space, and use the method of computing hold-out estimates for the proposed algorithm. Based on using the hold-out method, we obtain fast learning rate for this algorithm.

Bat Algorithm is a recently-developed method in the field of computational intelligence. In this paper is presented an improved version of a Bat Meta-heuristic Algorithm, (IBACH), for solving integer programming problems. The proposed algorithm uses chaotic behaviour to generate a candidate solution in behaviors similar to acoustic monophony. Numerical results show that the IBACH is able to obtain the optimal results in comparison to traditional methods (branch and bound), particle swarm optimization algorithm (PSO), standard Bat algorithm and other harmony search algorithms. However, the benefits of this proposed algorithm is in its ability to obtain the optimal solution within less computation, which save time in comparison with the branch and bound algorithm (exact solution method). Index Terms—Bat algorithm; meta-heuristics; optimization; chaos; integer programming.

K-Means algorithm is a one of the most famous unsupervised clustering algorithm. It has many disadvantages, such as sensitivity to the initial clustering centers and computes all the data points multiple times when facing the increasing data volume. In order to overcome the above limitations, this paper proposes to make use of density idea to find k cluster centers by adjusting the threshold. Finally, we design and implementation of the K-Means algorithm on the modern Graphic Processing Unit (GPU). The ratio of distance between classes to distance within classes and speedup are used as evaluation criteria. The experiments indicate that the proposed algorithm significantly improves the stability and efficiency of K-Means algorithm.

A new Evolutionary algorithm was presented based on intermediate value theorem of continuous function. The global search ability and the local search ability of this algorithm are well balanced, and operators used in the algorithm are simple, Further, small size of population scale is used. Initial numerical experiments show that the mean evolutionary algorithm is better than differential evolution algorithm in solving high dimension function optimization.

In this work, we have developed a new bio-inspired neural network algorithms for blockbased motion estimation. The main goal is to bridge the gap between algorithmic and biological vision by suggesting a bio-inspired motion estimation model based on neural network. We simplify the matching criterion for the block matching algorithm to reduce the hardware complexity and a number of input ports which maintaining the good quality. This paper implements the optimized algorithm in the reference model of H.264 compiled by VC6.0, and chooses six typical video sequences for simulation. The results show that our algorithm can reduce the average search points up to 82% to the full search black-matching algorithm. The optimized algorithm has reduced the motion estimation by 13.792% compared with UMHexagonS, and it gets better optimization to video testing sequences with low complexity.

The optimization of the Electromagnetism-like Mechanism (EM) algorithm is analyzed, and a Hybrid Chaos Search (CS) Electromagnetism-like Mechanism Algorithm (CEM) is proposed in this paper. The new algorithm combines the advantages of both the CS method and the EM algorithm. To make the initial population more uniform, the CEM algorithm constructs it with the theory of good point set in number theory. The CS method, rather than the random linear search algorithm, is used in the new algorithm, which can effectively prevent the search process from terminating prematurely. Besides, an adaptive movement is used in the improved algorithm to speed up convergence, and a genetic coefficient is added to the update formula to update the locations of particles so that the particles are more likely to move into the other feasible regions. Experimental results show that the improved algorithm can converge to global optimums more effectively and accurately.

For improving the effect of multi-target tracking in dense target and clutter scenario, a centralized fuzzy optimal assignment algorithm (CMS-FOA) of three-sensor multi-target system is proposed. And on the base of this, a generalized probabilistic data association algorithm (CMS-FOAGPDA) based on CMS-FOA algorithm is presented. The fusion algorithm gets effective 3-tuple of measurement set by using components of several satisfactory solutions of the fuzzy optimal assignment problem and then uses generalized probabilistic data association algorithm to calculate the update states of targets. Simulation results show that, in the aspect of multi- target tracking accuracy, CMS-FOA algorithm is superior to the optimal assignment (CMS-OA) algorithm based on state estimate and CMS-FOAGPDA algorithm is better than CMS-FOA algorithm. But considering the time spent, CMS-FOA algorithm spends a minimum of time and CMS-FOAGPDA algorithm is exactly on the contrary. Therefore, compared with CMS-OA algorithm, the two algorithms presented in the study each has its advantages and should be chosen according to the needs of the actual application when in use.

Based on the problems that traditional image de-noising algorithms easily to lose details features and always has low signal-to-noise ratio, inspired by the cross breeding method and the genetic algorithm, this paper proposes an image hybrid wavelet transform image de-noising algorithm based on Bayesian estimation. The proposed algorithm uses the Bayesian wavelet de-noised image as the male parent, and the Wiener filtering image as the female parent, the picked individual was prepared for the crossover and mutation operations. The optimal offspring will be choosing as the final solution algorithm and the decoding reduction as the de-noised image. The peak signal to noise ratio of the algorithm is much higher than that of the traditional algorithm, and the visual effect is better. The experimental results show that this method can not only eliminate the image noise effectively, but also can preserve the image edges and other details.

In this work, we present a new distributed algorithm for a non-convex and nonsmooth dictionary learning problem. The proposed algorithm, named proximal primal-dual algorithm with increasing penalty (Prox-PDA-IP), is a primal-dual scheme, where the primal step minimizes certain approximation of the augmented Lagrangian of the problem, and the dual step performs an approximate dual ascent. We provide a proof outline for convergence to stationary points, which is mainly based on constructing a new potential function that is guaranteed to decrease after some finite number of iterations. Numerical results are presented to validate the effectiveness of the proposed algorithm.

The efficiency of the Wang-Mendel (WM) algorithm is severely affected by the number of fuzzy rules and data scale. Thus, this paper proposes a reduced weighted WM algorithm to solve the problem by balancing the completeness and the computation time. The clustering algorithm is first introduced to obtain the cluster centers. Then, only the cluster centers are used to generate fuzzy rules, namely, the most important fuzzy rules are obtained. Finally, the weighted average is used to improve the accuracy of the WM algorithm. The proposed algorithm can save much computation time and storage space. The results of the experiments demonstrate that the proposed algorithm has high efficiency with high precision.

Nowadays, nesting problem has been encountered in many manufacturing industry. Nesting problem is given lots of layout elements and using algorithm to looking for the most suitable positions of every layout element in template to save the resource. In this paper, the two-dimensional problem is considered. The width of template is assumed to be fixed, and the heuristic and genetic algorithm is used to lessen the height. The no-fit polygon (NFP) is used to be the decode function of genetic algorithm. In the genetic algorithm, we use a combination of the height and overlap parameter considering that some combination's height is lower because of the overlap. The experimental results show the effectiveness of the proposed method.

The improvement of pattern matching algorithm is a key to improve the efficiency of network intrusion detection. Firstly, this thesis analyzes the current mainstream of single mode and multi-mode matching algorithm; then an improved multi-mode matching KR-BM2 algorithm based on KR algorithm and BM2 algorithm is proposed. BM2 algorithm using dual characters determines to move distance of the pattern string to the right, and uses the hash algorithm to reduce the number of comparisons in each match. KR-BM2 algorithm realizes the multi-pattern matching algorithm and the matching performance is improved. Finally the improvement of the proposed algorithm in the time performance is verified by experimental results.

Traditional K-means clustering methods have great attachment to the selection of the initial value and easily get into the local extreme value. This paper proposes a synthetic clustering algorithm of rough set and K-means based on Ant colony algorithm. While the rough set theory presents processing method of uncertain boundary objects, Ant colony algorithm is a bionic optimization algorithm, which has strong robustness, easily with other method unifies, solving efficiency higher characteristic.. Therefore, the K-means algorithm based on Ant colony algorithm in this paper combines rough set theory with simulated annealing algorithm and K-means, in which K means cluster number and initial cluster centers can be obtained dynamically with the principle of maximum minimum, and processing boundary objects with upper and lower approximation of rough set theory. Finally, the UCIs Iris set is used to test the algorithm. The experimental results show that the algorithm has higher accuracy rate, faster execution time and more stable performance.

In this paper, based on the combination of Genetic algorithm and BP algorithm, a new algorithm is proposed in this paper. The BP operator is embedded in the genetic operation in the algorithm, the algorithm effectively assimilates the global optimization of genetic algorithm and fast convergence of BP algorithm, and it encodes the construction and the weights hybrid with real code and binary code, achieving the same step optimization of structure and weights. The simulation results show that, the new algorithm can quickly converge to the global optimal solution, but also can obtain the best approximation of weights in the network structure.

In this paper we combine the largest minimum distance algorithm and the traditional K-Means algorithm to propose an improved K-Means clustering algorithm. This improved algorithm can make up the shortcomings for the traditional K-Means algorithm to determine the initial focal point. The improved K-Means algorithm effectively solved two disadvantages of the traditional algorithm, the first one is greater dependence to choice the initial focal point, and another one is easy to be trapped in local minimum [1][2].

This paper presents a fault diagnosis method of BP neural network based on Levenberg-Marquardt learning algorithm. First, the use of principal component analysis to reduce the dimension of the fault sample reduced BP neural network input variables. Then use the Levenberg-Marquardt learning algorithm to adjust the network weights. Levenberg-Marquardt learning algorithm is combination of the Gauss - Newton algorithm and steepest descent algorithm. It has Gauss - Newton algorithm of local convergence and gradient descent algorithm of the global characteristic. So it has higher convergence speed, reduces the training time, to a certain extent, overcomes the problem of traditional BP network convergence speed slow and easy to fall into local minimum point. Simulation results demonstrate the correctness and accuracy of this fault diagnosis method.

Using quantization techniques, Laloe (2009) defined a new algorithm called Alter. This $L^1$-based algorithm is proved to be convergent, but suffers two shortcomings. First, the number of clusters $K$ has to be supplied by the user. Second, it has high complexity. In this article, we adapt the idea of $X$-means algorithm (Pelleg and Moore; 2000) to offer solutions for these problems. This fast algorithm is used as a building-block which quickly estimates $K$ by optimizing locally the Bayesian Information Criterion (BIC). Our algorithm combines advantages of $X$-means (calculation of $K$ and speed) and Alter (convergence and parameter-free). Finally, an aggregative step is performed to adjust the relevance of the final clustering according to BIC criterion. We confront here our algorithm to different real and simulated data sets, which shows its relevance.

A coordinate direction search algorithm is designed to train artificial neural network error function. The algorithm searches all possible directions in the error space. An acceleration step is introduced for quick convergence. The step is taken when successive search by the algorithm reduces the function value. The repeated successful search directions provide information for orthogonal move. This direction of search is defined as leap-frog step. The algorithm is suitable when complex geometry of the error surface is present in the form of stiff ridges, valleys, contours, or flat surfaces. Quite often derivative-based training algorithm terminates in local minimum. The leaf-frog step allows the algorithm to escape local minimum. The algorithm is derivative free and is convenient when the derivative information of an error function is not available. The algorithm converges to minimum value and is robust. This algorithm is a different class and is not a random search or a heuristic optimization method. It ...

Based on the assumption about the sparse representation of natural images and the theory of compressed sensing, very promising results about single-image super-resolution were obtained by an excellent algorithm introduced by Yang et al. [45]. However, their success could not be well explained theoretically. The lack of theoretical insight has hindered the further improvement of the algorithm. In this paper, Yang's algorithm is revisited in the view of learning theory. According to this point, Yang's algorithm can be considered as a linear regression method in a special feature space which is named as sparse coding space by us. In fact, it has been shown that Yang's algorithm is a result of optimal linear estimation in sparse coding space. More importantly, our theoretical analysis suggests that Yang's algorithm can be improved by using more flexible regression methods than the linear regression method. Following the idea, a novel single-image super-resolution algorithm which is designed based on the framework of L"2-Boosting is proposed in the paper. The experimental results show the effectiveness of the proposed algorithm by comparing with other methods, which verify our theoretical analysis about Yang's algorithm.

An adaptive genetic algorithm was proposed to optimization bound in order to speed up the convergence of Gaussian mean shift. In practical question, as it's difficult to give a critical value definitely. According to the engine nonlinear of the corresponding oil and performance parameters, in gradient genetic algorithm, BP algorithm of local search is introduced. The adaptive value of chromosomes group gets quickly improved with the search in one coding field getting avoided due to the utilization of knowledge of chromosomes in problem-domain. The crossover and mutation operations are added so that chromosomes will not fall into the local minimum point in neighborhood. The experimental results prove that the convergence speed of the proposed method is non-linear and the use of gradient genetic algorithm is a fast algorithm that can support the global optimization of chromosomes in a group of process of iteration.

In image matching research, how to ensure that best match’s accuracy of the premise and a significant reduction in the amount of computing is the focus of concern by researchers. Search strategy to find the best match location of the image matching process to determine the amount of computing of image matching, in the existing image matching method are used to traverse search strategy, it is difficult to reduce the amount of computing. This is a common defect of the existing image matching algorithm. Traditional evolutionary algorithm trapped into the local minimum easily. Therefore, based on a simple evolutionary algorithm and combine the base ideology of orthogonal test then applied it to the population initialization, to prevent local convergence to form a new evolutionary algorithm. Compared the traditional evolutionary algorithm, the new algorithm enlarges the searching space and the complexity is not high. We use this new algorithm in image matching; from the results we reach the conclusion: in the optimization precision and the optimization speed, the new algorithm is efficiency for the image match problem.

The global k-means algorithm is an incremental approach to clustering that dynamically adds one cluster center at a time through a deterministic global search procedure from suitable initial positions, and employs k-means to minimize the sum of the intra-cluster variances. However the global k-means algorithm sometimes results singleton clusters and the initial positions sometimes are bad, after a bad initialization, poor local optimal can be easily obtained by k-means algorithm. In this paper, we modified the global k-means algorithm to eliminate the singleton clusters at first, and then we apply MinMax k-means clustering error method to global k-means algorithm to overcome the effect of bad initialization, proposed the global Minmax k-means algorithm. The proposed clustering method is tested on some popular data sets and compared to the k-means algorithm, the global k-means algorithm and the MinMax k-means algorithm. The experiment results show our proposed algorithm outperforms other algorithms mentioned in the paper.

This paper describes an algorithm for maximising a conditional likelihood function when the corresponding unconditional likelihood function is more easily maximised. The algorithm is similar to the EM algorithm but different as the parameters rather than the data are augmented and the conditional rather than the marginal likelihood function is maximised. In exponential families the algorithm takes a particular simple form which is computationally very close to the EM algorithm. The algorithm alternates between a T-step which calculates a tilted version of the unconditional likelihood function and an M-step which maximises it. The algorithm applies to mixed graphical chain models (Lauritzen and Wermuth, 1989) and their generalisations (Edwards, 1990), and was developed with these in mind, but it may have applications beyond these. The algorithm has been implemented in the most recent version of the MINI software (Edwards, 2000), where it was named the ME algorithm. The name has been changed to avoid confusion with the algorithm described by Marschner (2001). © 2001 Biometrika Trust.

The Group Steiner Tree (GST) problem is a fundamental problem in database area that has been successfully applied to keyword search in relational databases and team search in social networks. The state-of-the-art algorithm for the GST problem is a parameterized dynamic programming (DP) algorithm, which finds the optimal tree in O(3kn+2k(n log n + m)) time, where k is the number of given groups, m and n are the number of the edges and nodes of the graph respectively. The major limitations of the parameterized DP algorithm are twofold: (i) it is intractable even for very small values of k (e.g., k=8) in large graphs due to its exponential complexity, and (ii) it cannot generate a solution until the algorithm has completed its entire execution. To overcome these limitations, we propose an efficient and progressive GST algorithm in this paper, called PrunedDP. It is based on newly-developed optimal-tree decomposition and conditional tree merging techniques. The proposed algorithm not only drastically reduces the search space of the parameterized DP algorithm, but it also produces progressively-refined feasible solutions during algorithm execution. To further speed up the PrunedDP algorithm, we propose a progressive A*-search algorithm, based on several carefully-designed lower-bounding techniques. We conduct extensive experiments to evaluate our algorithms on several large scale real-world graphs. The results show that our best algorithm is not only able to generate progressively-refined feasible solutions, but it also finds the optimal solution with at least two orders of magnitude acceleration over the state-of-the-art algorithm, using much less memory.

Finding minimum spanning trees (MST) in various types of networks is a well-studied problem in theory and practical applications. A number of efficient algorithms have been already developed for this problem. In this paper we present an efficient algorithm, namely Edge Pruned Minimum Spanning Tree (EPMST) algorithm, which combines ideas from randomized selection, Kruskal's algorithm and Prim's algorithm. The algorithm has a superior performance relative to the best-known algorithms especially when the graph is not very sparse. Specifically, EPMST outperforms a recently devised efficient algorithm on a wide range of input graphs.

The Vortex Search (VS) algorithm is one of the recently proposed metaheuristic algorithms which was inspired from the vortical flow of the stirred fluids. Although the VS algorithm is shown to be a good candidate for the solution of certain optimization problems, it also has some drawbacks. In the VS algorithm, candidate solutions are generated around the current best solution by using a Gaussian distribution at each iteration pass. This provides simplicity to the algorithm but it also leads to some problems along. Especially, for the functions those have a number of local minimum points, to select a single point to generate candidate solutions leads the algorithm to being trapped into a local minimum point. Due to the adaptive step-size adjustment scheme used in the VS algorithm, the locality of the created candidate solutions is increased at each iteration pass. Therefore, if the algorithm cannot escape a local point as quickly as possible, it becomes much more difficult for the algorithm to escape from that point in the latter iterations. In this study, a modified Vortex Search algorithm (MVS) is proposed to overcome above mentioned drawback of the existing VS algorithm. In the MVS algorithm, the candidate solutions are generated around a number of points at each iteration pass. Computational results showed that with the help of this modification the global search ability of the existing VS algorithm is improved and the MVS algorithm outperformed the existing VS algorithm, PSO2011 and ABC algorithms for the benchmark numerical function set.