Search

Many real-life problems can be formulated as numerical optimization problems. Such problems pose a challenge for researchers when designing efficient techniques that are capable of finding the desired solution without suffering from premature convergence. This paper proposes a novel evolutionary algorithm that blends the exploitative and explorative merits of two main evolutionary algorithms, namely the Stochastic Fractal Search (SFS) and a Differential Evolution (DE) variant. This amalgam has an effective interaction and cooperation of an ensemble of diverse strategies to derive a single framework called Iterative Cyclic Tri-strategy with adaptive Differential Stochastic Fractal Evolutionary Algorithm (Ic3-aDSF-EA). The component algorithms cooperate and compete to enhance the quality of the generated solutions and complement each other. The iterative cycles in the proposed algorithm consist of three consecutive phases. The main idea behind the cyclic nature of Ic3-aDSF-EA is to gradually emphasize the work of the best-performing algorithm without ignoring the effects of the other inferior algorithm during the search process. The cooperation of component algorithms takes place at the end of each cycle for information sharing and the quality of solutions for the next cycle. The algorithm's performance is evaluated on 43 problems from three different benchmark suites. The paper also investigates the application to a set of real-life problems. The overall results show that the proposed Ic3-aDSF-EA has a propitious performance and a reliable scalability behavior compared to other state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

• A many objective evolutionary algorithm based on decomposition and moth search is proposed. • Random and adaptive weights is used to break the limitation of uniform distribution weights. • Mutual evaluation value is used to evaluate the optimal individual in the neighborhood. • Improving scale factor α in MSA is to improve the performance of the proposed algorithm. Decomposition-based multi-objective evolutionary algorithm (MOEA/D) has good performance in solving multi-objective problems (MOPs) but poor performance in solving many-objective optimization problems (MaOPs). The weight vectors in MOEA/D are relatively fixed, which results in poor performance when dealing with complex MaOPs. In this paper, random and adaptive weights are introduced into MOEA/D to break the limitation of fixed weight vectors. And the moth search algorithm (MSA) is used as an operator to improve global search ability. The updating strategies in MSA are more consistent with the neighborhood learning strategy adopted in MOEA/D. In addition, to enable MSA to find the optimal solution in the neighborhood on the MaOPs to update other individuals. This paper introduces mutual evaluation value for evaluating the optimal individual in the neighborhood, and the proposed algorithm is abbreviated as MOEA/DMS. In comparative experiments on the MaF test suite, hypervolume (HV) and inverted generational distance (IGD) are used to measure MOEA/DMS and other many-objective evolutionary algorithms (MaOEAs). The results show that MOEA/DMS has an excellent performance in dealing with MaOPs. Besides, MOEA/DMS is compared with other state-of-the-art MaOEAs on two combinatorial MaOPs. The results show that MOEA/DMS also has significant advantages in dealing with combinatorial MaOPs. [ABSTRACT FROM AUTHOR]

• A parallel EA with primary-auxiliary knowledge is proposed, using our improved cuckoo search algorithm as the primary knowledge and integrating several other EA as the auxiliary knowledge, which fundamentally enables the hybrid of different algorithms and greatly enhances algorithmic diversity. • The way of acting of the evolutionary strategy change from the traditional greedy selection to the balanced primary-auxiliary approach in this paper, which allows each strategy to work in iteration. • A novel knowledge migration strategy is proposed, which allows the primary knowledge to learn the excellent knowledge from the auxiliary knowledge through a novel topology designed for knowledge migration. The development of hybrid algorithms or the application of multiple strategies is one of the focal points for research on improving evolutionary algorithms. However, since most of the evolutionary equations of different algorithms can be transformed into each other, it is difficult to change the established properties of the given algorithm through a hybrid algorithm based on a mixture of evolutionary equations. In addition, multi-strategy methods tend to adopt the best strategy for the current local domain through greedy strategies in the solution process, which does not ensure validity in the global domain. Recently, Federated Learning has achieved remarkable results in machine learning, where the idea of model independence, parallelism and data sharing can essentially compensate for the weaknesses of hybrid and multi-strategy algorithms. Inspired by the idea of Federated Learning, this paper proposes an evolutionary algorithm named as parallel based Evolutionary Algorithm with primary-auxiliary knowledge. Specifically, a Spark-based primary-auxiliary knowledge model is developed, with different evolutionary algorithms used on each parallel sub-model. Then, an effective topological knowledge (individual) migration method is devised, which enables the primary knowledge model to learn the best knowledge from different auxiliary knowledge models through a topological structure. In this way, the best knowledge on the auxiliary knowledge models can be transferred to the primary knowledge model. Through a test conducted on the CEC2013 test set, it can be found out that the proposed algorithm clearly outperforms the 10 algorithms compared, which demonstrates the excellent performance of our proposed parallel based evolutionary algorithm with primary-auxiliary knowledge. [ABSTRACT FROM AUTHOR]

• A parallel EA with primary-auxiliary knowledge is proposed, using our improved cuckoo search algorithm as the primary knowledge and integrating several other EA as the auxiliary knowledge, which fundamentally enables the hybrid of different algorithms and greatly enhances algorithmic diversity. • The way of acting of the evolutionary strategy change from the traditional greedy selection to the balanced primary-auxiliary approach in this paper, which allows each strategy to work in iteration. • A novel knowledge migration strategy is proposed, which allows the primary knowledge to learn the excellent knowledge from the auxiliary knowledge through a novel topology designed for knowledge migration. The development of hybrid algorithms or the application of multiple strategies is one of the focal points for research on improving evolutionary algorithms. However, since most of the evolutionary equations of different algorithms can be transformed into each other, it is difficult to change the established properties of the given algorithm through a hybrid algorithm based on a mixture of evolutionary equations. In addition, multi-strategy methods tend to adopt the best strategy for the current local domain through greedy strategies in the solution process, which does not ensure validity in the global domain. Recently, Federated Learning has achieved remarkable results in machine learning, where the idea of model independence, parallelism and data sharing can essentially compensate for the weaknesses of hybrid and multi-strategy algorithms. Inspired by the idea of Federated Learning, this paper proposes an evolutionary algorithm named as parallel based Evolutionary Algorithm with primary-auxiliary knowledge. Specifically, a Spark-based primary-auxiliary knowledge model is developed, with different evolutionary algorithms used on each parallel sub-model. Then, an effective topological knowledge (individual) migration method is devised, which enables the primary knowledge model to learn the best knowledge from different auxiliary knowledge models through a topological structure. In this way, the best knowledge on the auxiliary knowledge models can be transferred to the primary knowledge model. Through a test conducted on the CEC2013 test set, it can be found out that the proposed algorithm clearly outperforms the 10 algorithms compared, which demonstrates the excellent performance of our proposed parallel based evolutionary algorithm with primary-auxiliary knowledge. [ABSTRACT FROM AUTHOR]

Surrogate-assisted evolutionary algorithms (SAEAs) with prescreening model management strategies show great potential in handling expensive optimization problems (EOPs). However, their performance is highly dependent on the search strategy and surrogate model. This paper proposes an evolutionary algorithm called IDRCEA, which utilizes an individual-distribution search strategy (IDS) and a regression-classification-based prescreening mechanism (RCP) to improve the ability to solve various complex and high-dimensional EOPs. Specifically, IDRCEA first combines an individual-based search strategy and a distribution-based search strategy to enrich offspring generation. Then, a regression model and a classification model are cooperatively used to prescreen the high-level offspring. Finally, both performance-based and distribution-based infill criteria are utilized to determine the most promising offspring from the high-level group for expensive evaluation. Experimental results validate the advantages of IDRCEA over some state-of-the-art SAEAs on many complex benchmark problems and an oil reservoir production optimization problem. • Individual-based and distribution-based search strategies are used to enrich offspring generation. • Regression and classification models are cooperatively employed to pre-screen the high-level offspring. • Performance-based and distribution-based infill criteria are adopted to choose the final offspring for function evaluation. [ABSTRACT FROM AUTHOR]

• Novel application of local optima network analysis to neural architecture landscapes. • Findings on a reduced search space were found to still hold on a larger search space. • Neural architecture search landscapes have relatively few local optima. The global structure of the hyperparameter spaces of neural networks is not well understood and it is therefore not clear which hyperparameter search algorithm will be most effective. In this paper we analyze the landscapes of convolutional neural network architecture search spaces to provide insight into appropriate search algorithms for these spaces. Using a classical fitness landscape analysis approach (fitness distance correlation) and a more recent tool (local optima networks) we study the global structure of these spaces. Our analysis on six image classification datasets reveals that the landscapes are multi-modal, but with relatively few local optima from which it is not hard to escape with a simple perturbation operator. This led us to explore the performance of iterated local search, which we found to more effectively search the training landscapes than three evolutionary algorithm variants. Evolutionary algorithms, however, outperformed iterated local search in terms of generalization on problems with larger discrepancies between the training and testing landscapes. [ABSTRACT FROM AUTHOR]

Penalty function is well-known for constrained evolutionary optimization. An open question in the penalty function is how to tune the penalty coefficient. This paper proposes an adaptive fuzzy penalty method to address this issue, where the coefficient is adjusted at both the individual level and the population level. At the individual level, each individual chooses a penalty coefficient from a predefined domain according to some fuzzy rules. At the population level, the domain of the crisp output is adjusted adaptively by using population information. To enhance the population diversity, an effective mutation scheme is developed. Due to its numerous merits, differential evolution is used to design a search algorithm. By the above processes, a constrained optimization evolutionary algorithm called AFPDE is proposed. Since the objective function value and the degree of constraint violation are normalized, AFPDE is less problem-dependent than the seminal work of the fuzzy penalty method. AFPDE introduces a lower penalty value in the early stage of AFPDE while a higher one in the later stage. Thus, it can escape local optima in the infeasible region. Experiments on three well-known benchmark test sets and two mechanical design problems validate that AFPDE is competitive. [ABSTRACT FROM AUTHOR]

The paper describes the problem of bias in parameter adaptation in Differential Evolution and other Evolutionary Algorithms. Based on the newly proposed Expected Fitness Improvement metric, the shift towards exploitation is demonstrated. The generalized Lehmer mean and Linear Bias Reduction are for the first time proposed to control the parameter adaptation bias for the fitness improvement based L-SHADE and distance based Db-L-SHADE algorithms. The experiments are performed on the benchmark functions of the Institute of Electrical and Electronics Engineers (IEEE) Congress on Evolutionary Computation (CEC) 2017 competition on real-parameter optimization. The influence of the modified scaling factor and crossover rate adaptation is evaluated using Friedman ranking procedure and Mann–Whitney statistical test. The usage of Lehmer mean and Linear Bias Reduction is shown to deliver statistically better results for high-dimensional functions and improve the exploration properties of the search algorithm in the long-term perspective. To support the statements made, the search process is additionally analyzed using diversity measures and cluster analysis techniques. [ABSTRACT FROM AUTHOR]

The main objective of this paper is to propose an effective evolutionary method for solving the problem of working paths optimization in survivable MPLS network. The paper focuses on existing network, in which only network flow can be optimized to provide network survivability using the local repair strategy. The problem is NP-complete, the solution space of the test cases is large and many genes are required to code the potential solution. Recently, the MuPPetS method (Multi-Population Pattern Searching Algorithm for Flow Assignment) was proposed and seems to be a promising tool for tackling high-dimensional, hard optimization problems. The MuPPetS is a linkage learning method that minimizes the negative effects of typical EA bottlenecks, e.g., preconvergence and significant effectiveness dropdown caused by an increasing number of genes in the chromosome. In comparison to other evolutionary methods, the MuPPetS was shown to be effective and capable of solving GA-hard problems. Therefore, the proposed MuPPetS-FuN method (Multi-Population Pattern Searching Algorithm for Flow Assignment in Non-bifurcated Commodity Flow) is based on MuPPetS. The additional objective of this paper is to propose changes to general MuPPetS framework to increase its effectiveness via better subpopulation number control strategy. [ABSTRACT FROM AUTHOR]

In literature, Clustered Shortest-Path Tree Problem (CluSPT) is an NP-hard problem. Previous studies focus on approximation algorithms which search for an optimal solution in relatively large space. Thus, these algorithms consume a large amount of computational resources while the quality of obtained results is lower than expected. In order to enhance the performance of the search process, this paper proposes two different approaches which are inspired by two perspectives of analyzing the CluSPT. The first approach intuition is to narrow down the search space by reducing the original graph into a multi-graph with fewer nodes while maintaining the ability to find the optimal solution. The problem is then solved by a proposed evolutionary algorithm. This approach performs well on those datasets having small number of edges between clusters. However, the increase in the size of the datasets would cause the excessive redundant edges in multi-graph that pressurize searching for potential solutions. The second approach overcomes this limitation by breaking down the multi-graph into a set of simple graphs. Every graph in this set is corresponding to a mutually exclusive search space. From this point of view, the problem could be modeled into a bi-level optimization problem in which the search space includes two nested search spaces. Accordingly, the Nested Local Search Evolutionary Algorithm (N-LSEA) is introduced to search for the optimal solution of glscluspt, the upper level uses a simple Local Search algorithm while the lower level uses the Genetic Algorithm. Due to the neighboring characteristics of the local search step in the upper level, the lower level reduced graphs share the common traits among each others. Thus, the Multi-tasking Local Search Evolutionary Algorithm (MLSEA) is proposed to take advantages of these underlying commonalities by exploiting the implicit transfer across similar tasks of multi-tasking schemes. The improvement in experimental results over N-LSEA via this multi-tasking scheme inspires the future works to apply M-LSEA in graph-based problems, especially for those could be modeled into bi-level optimization. [ABSTRACT FROM AUTHOR]

• Feature selection (FS) of high-dimensional data is reformulated as a many-objective optimization problem (MaOP) , consisting of three objectives to be minimized simultaneously. • To solve the formulated problem, we developed a PSO-based algorithm to search for the Pareto optimal solutions with two-level particle cooperation under the MOEA/D framework. • We made a systematical comparison between these the proposed methods and some state-of-the-art single objective and other MaOP FS approaches and the results demonstrate the efficacy of our proposed methods both in classification accuracy on the test data and the size of the feature subset. Feature selection (FS) plays a crucial role in classification, which aims to remove redundant and irrelevant data features.unknown However, for high-dimensional FS problems, Pareto optimal solutions are usually sparse, signifying that most of the decision variables are zero. Solving such problems using most existing evolutionary algorithms is difficult. In this paper, we reformulate FS as a many-objective optimization problem comprising three objectives to be minimized. To solve this problem, we proposed a binary particle swarm optimization with a two-level particle cooperation strategy. In the first level, to maintain rapid convergence, randomly generated ordinary particles and strict particles filtered by ReliefF are combined as the initialized particles. In the second level, under the decomposition multiobjective optimization framework, cooperation between particles is conducted during the update process to search for Pareto solutions more efficiently. In addition, a many-objective reset operation is proposed to enable the algorithm to jump out of the local optimum. Experimental studies on 10 real-world benchmark data sets revealed that our proposed algorithm could effectively reduce the number of features and achieve a competitive classification accuracy compared with some state-of-the-art evolutionary FS methods and non-evolutionary approaches. [ABSTRACT FROM AUTHOR]

Visualising a solution set is of high importance in many-objective optimisation. It can help algorithm designers understand the performance of search algorithms and decision makers select their preferred solution(s). In this paper, an objective reduction-based visualisation method (ORV) is proposed to view many-objective solution sets. ORV attempts to map a solution set from a high-dimensional objective space into a low-dimensional space while preserving the distribution and the Pareto dominance relation between solutions in the set. Specifically, ORV sequentially decomposes objective vectors which can be linearly represented by their positively correlated objective vectors until the expected number of preserved objective vectors is reached. ORV formulates the objective reduction as a solvable convex problem. Extensive experiments on both synthetic and real-world problems have verified the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

The tradeoff between exploration and exploitation is critical to the performance of an evolutionary algorithm. Different levels of exploration-exploitation tradeoff are required at different evolutionary stages for achieving a satisfactory performance of an evolutionary algorithm. In this paper, we propose a novel survival analysis method to intelligently guide the maintenance of the exploration-exploitation tradeoff in multiobjective evolutionary algorithms. The survival analysis stems from a deep understanding of the evolutionary search procedure. Through survival analysis, an indicator is derived, which is used to guide the adoption of appropriate recombination operators, based on the assumption that the roles of these operators in terms of their capabilities on exploration-exploitation can be asserted. In the developed algorithm, a differential evolution recombination operator and a new sampling strategy are hybridized. Empirical comparison with five well-known multiobjective evolutionary algorithms on a number of test instances with complex Pareto sets and Pareto fronts indicates the effectiveness and superiority of the developed algorithm in terms of commonly-used performance metrics on these test instances. [ABSTRACT FROM AUTHOR]

Surrogate models have been radically used in metaheuristic algorithms owing to their capacity in solving computationally expensive problems. However, despite the promising performance of surrogate-assisted metaheuristic algorithms in coping with low-dimensional problems, they failed to tackle high-dimensional problems efficiently. Thus, a multi-surrogate-assisted stochastic fractal search algorithm (MSASFS) is proposed in this paper. Several improvements are integrated into the algorithm design: (1) By combining the original coordinate system with the eigencoordinate system, an improved surrogate-assisted differential evolution (SDE) updating mechanism is proposed to ameliorate the generalization ability of the algorithm and extend the scope of exploration. (2) A new expected improvement (EI) pre-screening strategy based on the Gaussian process (GP) model is employed to select promising candidate solutions. (3) Two different surrogate models are applied to enhance the robustness of the proposed algorithm. The effectiveness of MSASFS is further demonstrated by numerical experiments on some widely used benchmark problems with dimensions ranging from 30 to 200 and parameter estimation problem of fractional-order chaotic systems. The results reveal that, compared with state-of-the-art surrogate-assisted evolutionary algorithms (SAEAs), the proposed algorithm can effectively solve high-dimensional expensive problems. Furthermore, MSASFS shows a more significant efficiency when the dimension of problems becomes higher. [ABSTRACT FROM AUTHOR]

Classical Markowitz mean-variance model is widely used for portfolio assets selection and allocation, which aims at simultaneously maximizing the expected return of the portfolio and minimizing portfolio variance. Many numerical approaches and metaheuristic algorithms have been proposed to effectively solve this portfolio optimization problem under an ideal condition. However, introducing various realistic constraints inadvertently leads to a non-convex search space, which has hindered the application of many classic, exact algorithms such as quadratic programming (QP). The increasing size of available assets and complex constraints has made the effectiveness of metaheuristic algorithms deteriorated. This paper proposes a hybrid bi-objective algorithm combining with the respective advantages of local search algorithm, evolutionary algorithm and QP with a pre-selection strategy. The algorithm first down select the assets that have greater contribution to the Pareto frontier by applying the pre-selection strategy. Then local search and evolutionary algorithm combined with QP are employed to fully exploit the useful assets combination modes to lead the search process toward the frontier direction quickly. The experimental study demonstrates that the proposed hybrid approach can obtain faster and better convergence compared with eight state-of the-art multi-objective evolutionary algorithms. The results also show that the proposed method with the pre-selection strategy always displays a closer proximity to the Pareto frontier compared with k -means strategy. [ABSTRACT FROM AUTHOR]

Biclustering has become a popular technique to analyse gene expression datasets and extract valuable information by clustering rows and columns of a dataset simultaneously. Using a good merit function together with a suitable local search can lead to the detection of interesting biclusters. In this paper, a multi-objective evolutionary algorithm with local search is proposed to search for multiple biclusters concurrently in a single run of the evolutionary algorithm. We call our method PBD-SPEA2 (Parallel Biclustering Detection using Strength Pareto front Evolutionary Algorithm 2). In our algorithm, a new dynamic encoding scheme is used to encode multiple biclusters in each individual. Our multi-objective function consists of three objectives that simultaneously optimizes the homogeneity of the elements in the bicluster, the size of the bicluster, and the variance of the column in the bicluster with respect to the entire dataset. Crossover is done by selecting and combining the best biclusters among the encoded biclusters from both parents through a strategy of exploration and exploitation. Finally, a sequential selection procedure is used to select the final set of biclusters from individuals that constitute the Pareto front. Experimental results are presented to compare the performance and biological enrichment of detected biclusters with several existing algorithms. [ABSTRACT FROM AUTHOR]

The backtracking search algorithm (BSA) is a recently proposed evolutionary algorithm (EA) that has been used for solving optimisation problems. The structure of the algorithm is simple and has only a single control parameter that should be determined. To improve the convergence performance and extend its application domain, a new algorithm called the learning BSA (LBSA) is proposed in this paper. In this method, the globally best information of the current generation and historical information in the BSA are combined to renew individuals according to a random probability, and the remaining individuals have their positions renewed by learning knowledge from the best individual, the worst individual, and another random individual of the current generation. There are two main advantages of the algorithm. First, some individuals update their positions with the guidance of the best individual (the teacher), which makes the convergence faster, and second, learning from different individuals, especially when avoiding the worst individual, increases the diversity of the population. To test the performance of the LBSA, benchmark functions in CEC2005 and CEC2014 were tested, and the algorithm was also used to train artificial neural networks for chaotic time series prediction and nonlinear system modelling problems. To evaluate the performance of LBSA with some other EAs, several comparisons between LBSA and other classical algorithms were conducted. The results indicate that LBSA performs well with respect to other algorithms and improves the performance of BSA. [ABSTRACT FROM AUTHOR]

The aim of hybridization in the context of evolutionary computation is to combine appropriate operators from different evolutionary computation paradigms to form a single technique that enjoys a statistically superior performance over a wide range of optimization problems. This paper introduces a novel hybridization between differential evolution and update processes of the stochastic fractal search algorithm. The diffusion property of the fractal search algorithm is applied in random fractals followed by two novel update processes to explore the search space more efficiently. In this algorithm, a diffusion process based on differential evolution algorithm is used instead of random fractals in the original stochastic fractal search algorithm. A new success-based scheme is used to utilize the update processes and to solve the burden of extra computations during the search. This new algorithm captures the strengths of both component algorithms and produces a greater explorative power as compared to the original algorithms. To verify the performance of our algorithm, a challenging test suite of 30 benchmark functions from the IEEE CEC2014 real parameter single objective competition is used. The results affirm the effectiveness and robustness of the proposed approach compared to the original stochastic fractal search and other recent state-of-the-art algorithms. The proposed algorithm enjoys a statistically superior performance over most of the tested benchmarks, especially hybrid and composition test functions compared to the other contestant algorithms. [ABSTRACT FROM AUTHOR]

In this paper we propose a specially designed memetic algorithm for multimodal optimisation problems. The proposal uses a niching strategy, called region-based niching strategy, that divides the search space in predefined and indexable hypercubes with decreasing size, called regions. This niching technique allows our proposal to keep high diversity in the population, and to keep the most promising regions in an external archive. The most promising solutions are improved with a local search method and also stored in the archive. The archive is used as an index to effiently prevent further exploration of these areas with the evolutionary algorithm. The resulting algorithm, called Region-based Memetic Algorithm with Archive, is tested on the benchmark proposed in the special session and competition on niching methods for multimodal function optimisation of the Congress on Evolutionary Computation in 2013. The results obtained show that the region-based niching strategy is more efficient than the classical niching strategy called clearing and that the use of the archive as restrictive index significantly improves the exploration efficiency of the algorithm. The proposal achieves better exploration and accuracy than other existing techniques. [ABSTRACT FROM AUTHOR]

In this paper, a novel recombination operator, called adaptive hybrid crossover operator (AHX), is designed for tackling continuous multiobjective optimization problems (MOPs), which works effectively to enhance the search capability of multiobjective evolutionary algorithms (MOEAs). Different from the existing hybrid operators that are commonly operated on chromosome level, the proposed operator is executed on gene level to combine the advantages of simulated binary crossover (SBX) with local search ability and differential evolution (DE) with strong global search capability. More opportunities are assigned to DE in the early evolutionary stage for gene-level global search in decision space; whereas, with the generation grows, more chances are gradually allocated to SBX for gene-level local search. The balance between the gene-level global and local search is well maintained by an adaptive control approach in AHX. To validate the effectiveness of AHX, it is studied by substituting the original recombination operators in the four state-of-the-art MOEAs (i.e., NSGA-II, SPEA2, SMS-EMOA, and MOEA/D), and the performance of revised algorithms is significantly improved. Furthermore, AHX is also compared to three recently proposed recombination operators, such as a newly DE inspired (DEI) recombination operator, a learning paradigm based on jumping genes (JGBL) and a bandit-based adaptive operator selection approach (FRRMAB). The experimental studies validate that AHX can be effectively integrated into different frameworks of MOEAs, and performs better than SBX, DE, DEI, JGBL and FRRMAB in solving various kinds of MOPs. [ABSTRACT FROM AUTHOR]

A keyword spotter can be considered as a binary classifier which classifies a set of uttered sentences into two groups on the basis of whether they contain target keywords or not. For this classification task, the keyword spotter needs to identify the target keywords locations based on a fast and accurate search algorithm. In our previous works, we exploited a modified Viterbi (M-Viterbi) search algorithm which has two known drawbacks. First, to locate the target keywords, it runs an exhaustive search through all possible segments of input speech. Second, while computing the start and end time-frames of each new phone, it makes the keyword spotter to trace-back and re-evaluate the timing alignments of all previous one(s), despite the fact that very limited amount of data -if any- would get updated as a result. These two pitfalls cause a dramatically enlarged search space as well as a significant increase in computational complexity. In this paper, we propose a Hierarchical Search (H-Search) algorithm which allows the system to ignore some segments of input speech at each level of hierarchy, according to their lower likelihood of containing the target keywords. In addition, unlike the M-Viterbi algorithm, the H-Search algorithm does not demand repeated evaluations when computing the current phone alignment which, in turn, results in a narrowed-down search space ( O ( TP ) versus O ( TPL max ) – where T is number of frames, P is number of keyword phones and L max is the maximum phone duration) as well as a decreased computational complexity ( O ( TPL max ) versus O ( TPL max 3 )) compared to those of the M-Viterbi algorithm. We applied the H-Search algorithm to the classification part of an Evolutionary Discriminative Keyword Spotting (EDKWS) system introduced in our previous works. The experimental results indicate that the H-Search algorithm is executed 100 times faster than the M-Viterbi algorithm while the performance of the EDKWS system degrades no more than two percent compared to that of the M-Viterbi algorithm. [ABSTRACT FROM AUTHOR]

Abstract: Artificial bee colony (ABC) algorithm is a relatively new optimization technique which has been shown to be competitive to other population-based algorithms. However, there is still an insufficiency in ABC regarding its solution search equation, which is good at exploration but poor at exploitation. To address this concerning issue, in this paper, we propose a novel ABC method called as EABC to improve the performance of ABC. In our method, in order to balance the exploration and the exploitation, two new search equations are presented to generate candidate solutions in the employed bee phase and the onlookers phase, respectively. Additionally, we use a more robust calculation to determine and compare the quality of alternative solutions. Experiments are conducted on a set of 48 benchmark functions and also two engineering optimization problems. The results show that EABC significantly improves the performance of ABC, offering faster global convergence, higher solution quality, and stronger robustness when compared with the other algorithms. [Copyright &y& Elsevier]

Abstract: Biogeography-based optimization (BBO) is an evolutionary optimization algorithm that uses migration to share information among candidate solutions. One limitation of BBO is that it changes only one independent variable at a time in each candidate solution. In this paper, a linearized version of BBO, called LBBO, is proposed to reduce rotational variance. The proposed method is combined with periodic re-initialization and local search operators to obtain an algorithm for global optimization in a continuous search space. Experiments have been conducted on 45 benchmarks from the 2005 and 2011 Congress on Evolutionary Computation, and LBBO performance is compared with the results published in those conferences. The results show that LBBO provides competitive performance with state-of-the-art evolutionary algorithms. In particular, LBBO performs particularly well for certain types of multimodal problems, including high-dimensional real-world problems. Also, LBBO is insensitive to whether or not the solution lies on the search domain boundary, in a wide or narrow basin, and within or outside the initialization domain. [Copyright &y& Elsevier]