Search

This paper presents a new Reinforcement Learning (RL)-based control approach that uses the Policy Iteration (PI) and a metaheuristic Grey Wolf Optimizer (GWO) algorithm to train the Neural Networks (NNs). Due to an efficient tradeoff to exploration and exploitation, the GWO algorithm shows good results in NN training and solving complex optimization problems. The proposed approach is compared to the classical PI RL-based control approach using the Gradient Descent (GD) algorithm, and with the RL-based control approach which uses the metaheuristic Particle Swarm Optimization (PSO) algorithm. The experiments are conducted using a nonlinear servo system laboratory equipment. Each approach evaluated on how well it solves the optimal reference tracking control for an experimental servo system position control system. The policy NNs specific to all three approaches are implemented as state feedback with integrator controllers to remove the steady-state control errors and thus ensure the convergence of the objective function. Because of the random nature of metaheuristic algorithms, the experiments for GWO and PSO algorithms are run multiple times and the results are averaged before the conclusions are presented. The experimental results shows that for the control objective considered in this paper, the GWO algorithm represents a better solution compared to GD and PSO algorithms.

The paper deals with online distributed convex optimization over a multi-agent network, where a swarm of agents attempts to minimize coordinately a sequence of time-changing global loss functions. At per time slot, the global loss function is decomposed into a sum of several local loss functions, each of which is available sequentially and held privately by one agent over the network, and individual agent does not possess prior knowledge of the future global loss function. We are interested in a bandit setting, where only the values of local loss functions at queried points are disclosed to individual agent. Meanwhile, we consider a general multi-agent network where the agents’ communication is represented as a sequence of time-changing unbalanced digraphs and the corresponding weight matrices are only row stochastic. We investigate two bandit scenarios including one- and two-point bandit feedback, and then design two corresponding online distributed bandit algorithms for such a class of problems by exploiting the classical dual averaging method. We show that both algorithms can achieve the sub-linear expected dynamic regret provided that the accumulative variation of the benchmark sequence grows sub-linearly with time. In particular, the bounds of the expected static regret obtained in this paper can reduce the relevant results when restricted to the centralized bandit online convex optimization. In contrast to existing algorithms with gradient feedback, numerical tests verify the competitive performances of the proposed algorithms.

Temporal graphs are graphs whose edges are associated with timestamps. Subgraph matching on temporal graphs retrieves temporal subgraphs whose edge timestamps satisfy user-specified temporal orders. In this paper, a temporal query pattern is composed of a query graph with an arbitrary structure and a partial order on the edge set of the query graph. Moreover, the time span between the minimum and the maximum edge timestamps in a matching is required to be less than or equal to a specified threshold. This paper proposes a temporal subgraph matching algorithm based on two key techniques. First, a memory-efficient index structure called TO-tree is designed to compactly store all necessary information required for finding all temporal subgraph matchings. The TO-tree constructed for a temporal query pattern is much smaller than the temporal graph because unnecessary information is mostly excluded from the TO-tree by three powerful filters. The second technique is a temporal subgraph matching enumeration method that runs on the TO-tree instead of on the temporal graph. This enumeration method expands temporal subgraph matchings in an edge-by-edge manner. Since the TO-tree can fit into the main memory, the enumeration method runs very fast on the TO-tree. An extensive experimental evaluation has been carried out. The experimental results show that the TO-tree index structure is memory-efficient, which in turn enables a fast temporal subgraph matching enumeration. Overall, our algorithm is at least 3X faster than the baseline algorithm adapted from the state-of-the-art non-temporal subgraph matching algorithm CECI and is at least 4X faster than the temporal subgraph matching algorithm HASSE .

Evolving fuzzy systems (EFSs) are a type of adaptive fuzzy rule-based systems which can self-adapt both their structures and parameters simultaneously. However, the existing EFSs suffer from two drawbacks: 1) classical EFSs usually use all input features to model systems, resulting in lengthy fuzzy rules; 2) some redundant information in fuzzy rules may hinder high generalization . To address these two issues, a promising method is proposed in this paper by combining very sparse random projection (VSRP) with a class of EFSs based-on data clouds, called VSRP-AnYa-EFS. The proposed method introduces: 1) a random sparse-Bernoulli (RSB) matrix based-on VSRP is utilized to compress the lengthy antecedent part into a tighter form, triggering a feature-reduction mechanism. By employing VSRP in RSB matrix, some redundant information in fuzzy rules can be filtered; 2) Local learning is used for consequent parameter optimization to suit decoupled behavior of rules after redundant information between rules is deleted. By adopting VSRP and local learning, the proposed VSRP-AnYa-EFS owns a compact structure and fast learning speed. Numerical examples presented in this paper demonstrate that the proposed method can significantly reduce training time from hours to minutes while the accuracy can be improved up to 5%.

In recent years, many graph compression methods have been introduced. One successful category of them is based on local decompression designed to answer neighborhood queries. These techniques mainly rely on local similarities of vertices. Besides, their performance is usually a function of graph sparsity. The proposed approach, in this paper, is a lossy compression technique used to answer neighborhood queries with a more general precondition, called transitivity. The output of this method is a sparse graph optimized to keep original adjacent vertices, in at most 2-distance from each other and vice versa. In other words, by traversing a compressed graph by depth of 2, from any desired vertex, its original adjacency list is reconstructed, with an acceptable error. This paper models an optimization problem to solve the inverse problem of finding the best compressed graph in order to minimize the reconstruction error. Then, this NP problem is approximated by a heuristic with a low degree polynomial time-complexity near to the complexity of the forward problem. The results of applying the proposed method on toy and real datasets are compared with the state of the art that improves compression ratio and performance with an acceptable query response time.

This paper proposes a new tent-multi dynamic piecewise coupled mapping lattice (TMDPCML). Through the comprehensive analysis of the performance test results, TMDPCML system increases the correlation dynamics of spatiotemporal behavior and improves the efficiency of energy diffusion between lattices. Moreover, TMDPCML system has larger parameter space, better chaos and outstanding cryptographic characteristics. Therefore, this paper proposes a privacy image encryption algorithm combined with TMDPCML system. The application of TMDPCML system in private images encryption further proves that TMDPCML system has good chaotic behavior and meets the requirements of cryptography.

In this paper, a novel approach of supervised feature selection is proposed based on the principle of dense subgraph discovery. To exploit dense subgraph discovery for the purpose of feature selection, the dataset is initially mapped to an equivalent weighted graph notation by considering the set of all features as its vertex set and the mutual dependency between each pair of features as the weight of the corresponding edge. The proposed feature selection algorithm proceeds in a two-phase manner. In the first stage, a dense sub-graph is first discovered so that the features within it become maximally non-redundant among each other and the averaged class relevance as well as averaged standard deviation of all these features are obtained as maximal as possible. In this regard, a novel induced degree is also defined for each feature by incorporating the aforesaid three important objectives of feature selection. In this phase, a modified version of an existing approximation algorithm is also used to find dense subgraph module. Finally, in the second stage, a floating forward–backward search is performed on the dense subgraph so obtained to reveal a better feature subset. In both stages, an existing version of the normalized mutual information score is employed to compute both the class relevance and redundancy. The main contribution of this paper is proposing a feature selection strategy by which the reduced features have the characteristics like maximal average class relevance, minimal average pairwise redundancy, and good discriminating power. The experimental results demonstrate that the proposed approach is competent with several conventional as well as state-of-art algorithms of supervised feature selection.

The influence of each entity in a network is a crucial index of the network information dissemination. Greedy influence maximization algorithms suffer from time efficiency and scalability issues. In contrast, heuristic influence maximization algorithms improve efficiency, but they cannot guarantee accurate results. Considering this, this paper proposes a Gaussian propagation model based on the social networks. Multi-dimensional space modeling is constructed by offset, motif, and degree dimensions for propagation simulation. This space’s circumstances are controlled by some influence diffusion parameters. An influence maximization algorithm is proposed under this model, and this paper uses an improved CELF algorithm to accelerate the influence maximization algorithm. Further, the paper evaluates the effectiveness of the influence maximization algorithm based on the Gaussian propagation model supported by theoretical proofs. Extensive experiments are conducted to compare the effectiveness and efficiency of a series of influence maximization algorithms. The results of the experiments demonstrate that the proposed algorithm shows significant improvement in both effectiveness and efficiency.

This paper proposes an adaptive denoising methodology for noisy signals that employs a novel adaptive complementary ensemble empirical mode decomposition (NACEEMD) and a peak statistics (PS)-based thresholding technique. The key idea in this paper is the peak statistics (PS)-based thresholding technique,which breaks the traditional strategy with respect to selecting more accurate and more adaptive thresholds. The NACEEMD algorithm is proposed to decompose the noisy signal into a series of intrinsic mode functions (IMFs). At the same time, NACEEMD is also used to verify the applicability of the PS-based thresholding technique in different decomposition algorithms. The PS-based threshold is used to remove the noise inherent in noise-dominant IMFs, and the denoised signal is reconstructed by combining the denoised noise-dominant IMFs and the signal-dominant IMFs. This paper uses a various of simulated signals in various noisy environments for experiments, the experimental results indicate that the proposed algorithm outperforms traditional threshold denoising methodologies in terms of signal-to-noise ratio, root mean square error, and percent root distortion. Moreover, through real ECG signal and multi-sensor data fusion experiments, the application of the proposed algorithm in the field of engineering is explored and expanded.

Velocity planning serves as an important issue in motion planning for autonomous vehicles. The presented paper proposes a novel velocity planning method with minimum moving time on the basis of path curvature which is accomplished in three steps. First, the assigned path is divided into some elementary parts based on the path curvature. Second, the velocity planning is transformed into an unconstrained optimization problem by assuming the velocity of vehicle to be a specific cubic polynomial on every elementary part to avoid a sudden acceleration in path switching. Finally, we use a modified projection particle swarm optimization (PPSO) algorithm to obtain the time-optimal velocity profile . The proposed method can generate a smooth time-optimal velocity profile while considering all possible relevant constraints. Three examples are provided on different types of path to demonstrate that the final velocity profile is efficient to avoid the sudden acceleration change. Furthermore, the modified PPSO algorithm in this paper is used to solve the optimization problem with high dimensional variables when its upper bound is known, which can not be achieved by the general PPSO algorithm.

Non-singleton fuzzification is used in rule-based fuzzy systems when the measurements that activate them are imperfect or uncertain or when their inputs are words. It models such measurements or words as fuzzy numbers or more general fuzzy sets so that, regardless of the cause of a measurement’s or word’s imperfections or uncertainties, they are treated within the framework of fuzzy sets and systems. Since 2011, there has been a resurgence of interest in both type-1 and interval type-2 non-singleton fuzzy systems. This paper removes a computational bottleneck associated with computing the firing level or firing interval for such fuzzy systems, by providing closed-form formulas for them when the involved fuzzy sets are trapezoidal or triangular, which are widely used fuzzy sets. This is done for both the minimum and product t-norms. It is also demonstrated that a non-singleton fuzzy system that uses the product t-norm has the potential to outperform a non-singleton fuzzy system that uses the minimum t-norm. The results in this paper greatly simplify non-singleton fuzzy systems, which should make them much more popular.

This paper mainly proposes a new method for clustering linear ordinal ranking (LOR) information by agglomerative hierarchical clustering (AHC) algorithm. Considering that the cores of the AHC algorithm for LOR clustering are the difference measure among different LORs and the aggregation of the individual LORs, we firstly systematically analyze the existing studies for LOR distance measure, based on which we extend the method to depict LORs. Subsequently, the corresponding new distance measure is proposed starting from the perspective of utilizing the rankings’ position information and relationship information together. In addition, we simplify the dominating index and dominated index-based aggregation method for LORs fusion. Further, we present a numerical case on online financial product recommendation to illustrate the usage of the algorithm and also try to provide a feasible way for online financial product recommendation. Then, we make some discussions on the proposed distance measure and the aggregation method under the framework of the AHC algorithm to show the features of the algorithm proposed in this paper.

Identifying frequent item-sets is a popular data-mining task. It consists of finding sets of items frequently appearing in data. Yet, finding all frequent item-sets in large or dense datasets may be time-consuming, and a user may be interested merely in some specific item-sets rather than all of them. Recently, methods have been proposed for targeted item-set mining; that is to calculate the support of some item-sets of interest. Though this approach is often more suitable for real applications than traditional item-set mining approaches, performance remains an issue. To address that issue, this paper presents a novel algorithm for multitude-targeted mining, named Guided Frequent Pattern-Growth (GFP-Growth). The GFP-Growth algorithm is designed to quickly mine a given set of item-sets using a small amount of memory. This paper proves that GFP-Growth yields the exact frequency-counts for each item-set of interest. It further shows that GFP-Growth can boost the performance for several problems requiring item-set mining. We specifically study the problem of generating minority-class rules from imbalanced data and develop the Minority-Report Algorithm (MRA) that uses GFP-Growth to solve this problem efficiently. We prove several theoretical properties of MRA and present experimental results showing substantial performance gain.

Periodic frequent patterns are sets of events or items that periodically appear in a sequence of events or transactions. Many algorithms have been designed to identify periodic frequent patterns in data. However, most assume that the periodic behavior of a pattern does not change much over time. To address this limitation, this paper proposes to discover a novel type of periodic patterns in a sequence of events or transactions, called Local Periodic Patterns (LPPs) which are patterns (sets of events) that have a periodic behavior in some non predefined time-intervals. A pattern is said to be a local periodic pattern if it appears regularly and continuously in some time-interval(s). Two novel measures are proposed to assess the periodicity and frequency of patterns in time-intervals. The maxSoPer (maximal period of spillovers) measure allows detecting time-intervals of variable lengths where a pattern is continuously periodic, while the minDur (minimal duration) measure ensures that those time-intervals have a minimum duration. To discover all LPPs, the paper presents three efficient algorithms. An experimental evaluation on real datasets shows that the proposed algorithms are efficient and can provide useful patterns that cannot be found using traditional periodic pattern mining algorithms.

In this paper, an improved cross coupled map lattice (CCML) which is derived from CCML is proposed. The experimental analysis of CCML shows that it exhibits weak or even lose chaotic behaviors in a large range of parameters. To enhance its chaotic performance, the new model changes the internal chaotic map of CCML from the logistics map to the tent map and introduces the module operations to limit the obtained value. The theoretical analysis and experimental results prove that the proposed model has a broad chaotic regime over an extensive range of system parameters, positive Lyapunov exponents, higher information entropies and lower mutual information values when compared to CCML, which are highly suitable for chaos-based image encryption. Moreover, this paper applies that model on a bit-level image encryption scheme. The algorithm consists of three main operations: the generation of secret keys, confusion and diffusion operations that can break the limitation of row and column. To confirm the novel algorithm's effectiveness and safety against the common types of attacks, many experiments are done and security analyses are discussed. It demonstrates the good chaos of the proposed model from the aspect of encryption performance.

In this paper, a chaotic image encryption algorithm based on the matrix semi-tensor product (STP) with a compound secret key is designed. First, a new scrambling method is designed. The pixels of the initial plaintext image are randomly divided into four blocks. The pixels in each block are then subjected to different numbers of rounds of Arnold transformation, and the four blocks are combined to generate a scrambled image. Then, a compound secret key is designed. A set of pseudosecret keys is given and filtered through a synchronously updating Boolean network to generate the real secret key. This secret key is used as the initial value of the mixed linear-nonlinear coupled map lattice (MLNCML) system to generate a chaotic sequence. Finally, the STP operation is applied to the chaotic sequences and the scrambled image to generate an encrypted image. Compared with other encryption algorithms, the algorithm proposed in this paper is more secure and effective, and it is also suitable for color image encryption.

There are three indispensable, yet contrasting requirements for a watermarking scheme: imperceptibility, robustness, and payload. Therefore, it is a challenge to obtain a tradeoff among above requirements in image watermark detection. In this paper, we propose a new statistical watermark decoder in discrete non-separable Shearlet transform (DNST) domain using singular values and Gaussian-Cauchy mixture-based vector hidden Markov tree (HMT). Our method can obtain great performance in imperceptibility, robustness and payload, and it is necessary for image copyright protection. We first perform DNST on the host image, and apply singular value decomposition (SVD) to the significant DNST domain high entropy blocks. We then embed the digital watermark into the DNST high entropy blocks by modifying the robust singular values. At the receiver, by combining the Gaussian-Cauchy mixture-based vector HMT and maximum likelihood (ML) decision, we propose a new blind image watermark decoder in DNST domain. Here, robust DNST domain singular values are firstly modeled by using Gaussian-Cauchy mixture-based vector HMT, where the Gaussian-Cauchy mixture marginal distribution and various strong dependencies of DNST domain singular values are incorporated. Then the statistical model parameters of Gaussian-Cauchy mixture-based vector HMT are estimated using parameter-expanded expectation–maximization (PXEM) approach. And finally, a blind image watermark decoder is developed using Gaussian-Cauchy mixture-based vector HMT and ML decision rule. The major contribution of this paper is the use of singular value, Gaussian-Cauchy mixture-based vector HMT and PXEM algorithms, which enhances the performance of watermarking scheme. Experimental results on some test images and comparison with well-known existing methods demonstrate the efficacy and superiority of the proposed method.

Despite decades of intensive effort, the current solutions for efficiently searching high-dimensional data spaces are not entirely satisfactory. This paper proposes a more efficient global low-density locality sensitive hashing search algorithm (GLDH) based on the minimal cut hyperplane and ensemble learning. The innovation is that a novel global low-density hyperplane candidate set is constructed by the graph cut method, the minimum information gain method and random maximum entropy method are used to greedily select the hyperplane, and the ensemble learning method is used to query the global approximate nearest-neighbors data. This paper proves that the GLDH algorithm produces a low error hyperplane partition. The results of extensive experiments show that the proposed GLDH method performs better than the latest methods when using the same hash coding length for datasets from different fields.

This paper tackles the problem of community detection in large-scale graphs. In the literature devoted to this topic, an iterative algorithm, called Louvain Method (LM), stands out as an effective and fast solution for this problem. However, the first iterations of the LM are the most costly. To overcome this issue, this paper introduces CoVeC, a Coarse-grained Vertex Clustering for efficient community detection in sparse complex networks. CoVeC pre-processes the original graph in order to forward a graph of reduced size to the LM. The subsequent group formation, including the maximization of group quality, as per the modularity metric, is left to the LM. We evaluate our proposal using real-world and synthetic networks, presenting distinct sizes and sparsity levels. Overall, our experimental results show that CoVeC can be a way faster option than the first iterations of the LM, yet similarly effective. In fact, for sparser graphs, the combo CoVeC+LM outperforms the standalone LM and its variations, attaining a mean processing time reduction of 47% and a mean modularity reduction of only 0.4%.

Metaheuristic algorithms provide a practical tool for optimization in a high-dimensional search space. Some mimic phenomenons of nature such as swarms and flocks. Prominent one is the Gravitational Search Algorithm (GSA) inspired by Newton’s law of gravity to manipulate agents modeled as point masses in the search space. The law of gravity states that interaction forces are inversely proportional to the squared distance in the Euclidean space between two objects. In this paper we claim that when the set of solutions lies in a lower-dimensional manifold, the Euclidean distance would yield unfitted forces and bias in the results, thus causing suboptimal and slower convergence. We propose to modify the algorithm and utilize geodesic distances gained through manifold learning via diffusion maps. In addition, we incorporate elitism by storing exploration data. We show the high performance of this approach in terms of the final solution value and the rate of convergence compared to other meta-heuristic algorithms including the original GSA. In this paper we also provide a comparative analysis of the state-of-the-art optimization algorithms on a large set of standard benchmark functions .

An advanced predictor offers the most effective way for performance enhancement of prediction-error expansion (PEE) based reversible data hiding. Unlike conventional predictors which exploit the correlation among adjacent pixels, pixel-value-ordering (PVO) exploits the one between the largest/smallest two pixels within block regardless of where they are located. Later, the spatial location received the attention from Peng et al. and led to better expansion bins selection. In this paper, a flexible spatial location (FSL) strategy is proposed to optimize the utilization of spatial correlation in PVO prediction. With FSL, pixels within block are no longer collected in a fixed way (e.g., from top to bottom, left to right). Specifically, eight modes of defining spatial location are designed for pixel collection. After pixel sorting, the inverse number of the location sequence is used to evaluate all modes and determine the best one. By determining the optimal mode for each block, the prediction becomes adaptive and thus a prediction-error histogram with higher expansion bin will be obtained. Numerous PVO-based schemes will benefit from FSL. In this paper, how FSL is applied to conventional PEE and high-dimensional PEE is also introduced. Experimental results demonstrate that FSL-PVO is of great significance to better capacity-distortion trade-off.

3D local descriptors are the fundamental and essential elements that have been commonly applied in 3D computer vision. This paper proposes a novel and effective 3D local descriptor for describing the 3D local shape. The research focuses on accelerating the descriptor generation by simplifying the Local Reference Frame (LRF) and optimizing the feature space through a Weighted Height Image (WHI). An in-depth theoretical analysis of the LRF is conducted. Then, this study proposes a simplified LRF to reduce the redundant computations of the covariance matrix and share the calculations with the 3D information coding. Besides, the feature space is modeled and analyzed in this paper. Based on the analysis, we propose a weighting function to strengthen the abilities of the feature representation. The experimental results indicate that the proposed WHI descriptor outperforms the state-of-the-art (SOTA) algorithms in terms of accuracy and efficiency. Meanwhile, the compactness of the WHI is about six times more than that of the SOTA algorithms. Moreover, for the application of point cloud registration, the proposed WHI exhibits high effectiveness in terms of both accuracy and real-time capability.

Identifying high utility sequences in a quantitative sequence database is an important data mining task. However, a key problem of current approaches is that extensions of a high utility sequence often have a high utility. Hence, traditional techniques are often biased toward finding long patterns. To circumvent this problem, this paper proposes techniques for the problem of high average-utility sequence (HAUS) mining ( H A U S M ). HAUSs are more meaningful than high utility sequences because the former are found using the average-utility measure, which consider the length of patterns in utility calculations. H A U S M is more general than high average-utility itemset mining but it is also a more difficult problem because the downward-closure property used for search space reduction does not hold for the average-utility. To overcome that challenge, this paper introduces two upper bounds and a weak upper bound on the average-utility measure, and four width pruning, depth pruning, reducing, and tightening strategies. These strategies are designed to eliminate candidate HAUSs to speed up HAUS mining. Based on these theoretical results, a novel algorithm named EHAUSM is proposed for H A U S M . Experiments on both real-life and synthetic quantitative sequence databases have confirmed its efficiency in terms of memory consumption and runtime.

The file attached to this record is the author's final peer reviewed version. The Publisher's final version can be found by following the DOI link. In general, for the iteration process of an evolutionary algorithm (EA), there exists the problem of uneven distribution of individuals in the target space for both multi-objective and single-objective optimization problems. This uneven distribution significantly degrades the population diversity and convergence speed. This paper proposes an adaptive hybrid evolutionary immune algorithm based on a uniform distribution selection mechanism (AUDHEIA) for solving MOPs efficiently. In AUDHEIA, the individuals in the population are mapped to a hyperplane, which is correlated with the objective space and are clustered to increase the diversity of solutions. To improve the distribution of the solutions, the mapped hyperplane is evenly sectioned. With the constantly changing distribution during the iteration, a threshold as a standard for judging the distribution level is adjusted adaptively. When the threshold is not satisfied in the corresponding interval, the distribution enhancement module is activated. Then, the same number of individuals should be selected in each interval. However, sometimes, there are insufficient or no individuals in the interval during the iterative process. To obtain sufficient individuals, the limit optimization variation strategy of the best individual is adopted. Experiments show that this algorithm can escape from local optima and has a high convergence speed. Moreover, the distribution and convergence of this algorithm are superior to the peer algorithms tested in this paper.

This paper presents an online-learning ensemble framework for nonstationary time series prediction . Optimal granular fuzzy rule-based models with different objective functions and constraints are evolved from data streams. Evolving optimal granular systems (eOGS) consider multiobjective optimization , the specificity of information, model compactness , and variability and coverage of the data within the process of modeling data streams. Forecasts of individual base eOGS models are combined using averaging aggregation functions: ordered weighted averaging (OWA), weighted arithmetic mean, median, and linear non-inclusive centered OWA. Some aggregation functions use specific weights for the relevance of the base models and exclude extreme values and outliers. The weights of other aggregation functions are adapted over time based on a quadratic programming problem and the data within a sliding window. This paper investigates whether an online-learning ensemble can outperform individual eOGS models, and which aggregation function provides the most accurate forecasts. Real multivariate weather time series, particularly time series of daily mean temperature, air humidity, and wind speed from different weather stations, such as Paris–Orly, Frankfurt–Main, Reykjavik, and Oslo–Blindern, are used for evaluation. The results show that ensemble schemes outperform individual models. The proposed linear non-inclusive centered OWA function provided the most accurate numerical predictions .

This paper presents an investigation into the short-term versus long-term performance of various particle swarm optimization (PSO) control parameter configurations. While evidence suggests that the best PSO parameter values to employ are time-dependent, this paper provides an in-depth examination of a small set of parameter values to provide a more concrete quantification of the performance degradation observed with specific control parameter configurations over time. Given that the short-term performance is not necessarily indicative of long-term performance, this paper proposes that machine learning techniques be used to build predictive models based on two easily-observable landscape characteristics. Finally, using the predictive models as a basis, this paper also proposes a parameter-free PSO algorithm, which performs on par with other top-performing PSO variants, namely the three best performing static PSO configurations, particle swarm optimization with time-varying acceleration coefficients (PSO-TVAC), and particle swarm optimization with improved random constants (PSO-iRC).

Target tracking is a difficult estimation problem due to target motion uncertainty and measurement origin uncertainty. In this paper, we consider the target tracking problem in the presence of only target motion uncertainty. The traditional approaches to address this uncertainty, such as multiple model approaches, can suffer performance degradation when there is a model mismatch. The statistical accuracy of conventional model-based methods is also usually limited because of the measurement errors and insufficient data for the estimation. In this paper, deep neural network-based methods are proposed to handle target motion uncertainty due to their strong capability of fitting any mapping as long as there are sufficient training data. Specifically, a recurrent neural network-based structure is proposed to estimate the true states that is consistent with the sequential manner of target tracking. In addition, it is expected that better performance will be achieved due to access to true states during the training of the networks. We propose two networks that are based on different principles and are capable of real-time tracking. An approach to further reduce the computational load is also introduced. Simulation results show that the proposed methods can handle the target motion uncertainty well and provide better estimation accuracy.

The concept of SCN offers a fast framework with universal approximation guarantee for lifelong learning of non-stationary data streams. Its adaptive scope selection property enables for proper random generation of hidden unit parameters advancing conventional randomized approaches constrained with a fixed scope of random parameters. This paper proposes deep stacked stochastic configuration network (DSSCN) for continual learning of non-stationary data streams which contributes two major aspects: 1) DSSCN features a self-constructing methodology of deep stacked network structure where hidden unit and hidden layer are extracted automatically from continuously generated data streams; 2) the concept of SCN is developed to randomly assign inverse covariance matrix of multivariate Gaussian function in the hidden node addition step bypassing its computationally prohibitive tuning phase. Numerical evaluation and comparison with prominent data stream algorithms under two procedures: periodic hold-out and prequential test-then-train processes demonstrate the advantage of proposed methodology., This paper has been published in Information Sciences

This paper is concerned with the problem of distributed Nash equilibrium computation in aggregative games. Note that the traditional computation algorithms are designed based on time-scheduled communication strategy, which may lead to high communication consumption of the whole network. To reduce the consumption, this paper proposes a novel distributed algorithm with an event-triggered mechanism, where the communication between any two agents is only carried out when an edge-based event condition is triggered. In the convergence analysis of the proposed algorithm, an important event-related error variable is firstly defined. Then, based on a zero-sum property of this event-related error, two key relations on the agents’ estimates in the proposed algorithm are provided. Further, by using these relations, it is proven that the agents’ estimates can achieve a Nash equilibrium under a proper event-triggering condition. Finally, examples on the demand response of power systems are presented to verify the theoretical findings.

To ensure security, image encryption algorithms generally include two stages: permutation and diffusion. The traditional image permutation algorithms include the sort-based permutation algorithm, Arnold-based permutation algorithm, Baker-based permutation algorithm and the cyclic shift permutation algorithm, etc. However, these algorithms have the disadvantages of either high time complexity or poor permutation performance. Therefore, in combination with cyclic shift and sorting, this paper proposes a permutation algorithm that can not only guarantees good permutation performance but also guarantee low time and space complexity. Most importantly, this paper proposes a parallel diffusion method. This method ensures the parallelism of diffusion to the utmost extent and achieves a qualitative improvement in efficiency over traditional streaming diffusion methods. Finally, combined with the proposed permutation and diffusion, the paper proposes a computational model for parallel image encryption algorithms.

A certificateless aggregate signature (CLAS) scheme employing a reduced signature size eliminates the complexity of certificate management in a traditional public key cryptosystem . A compact aggregate signature, where the size does not depend on the aggregate number, is desirable when the objective is to reduce storage cost and bandwidth. However, in conventional compact CLAS schemes, aggregation of signatures is restricted. The state information is meant to be used only once for security; it is generally used to restrict aggregation to signatures generated with the same state information. This paper proposes the first unrestricted and compact CLAS scheme where the signature size is constant and any combination of signatures can be aggregated. Aside from convenient storage and communication costs, our scheme is also equipped with a secure system against realistic adversaries. Moreover, we evaluate the performance of our CLAS scheme and demonstrate its effectiveness. Finally, this paper reveals that it is not possible to construct an unrestricted and compact CLAS scheme for a widely used structure with constant pairing computation.

DNA sequence reconstruction is a challenging research problem in the computational biology field. The evolution of the DNA is too complex to be characterized by a few parameters. Therefore, there is a need for a modeling approach for analyzing DNA patterns. In this paper, we proposed a novel framework for DNA pattern analysis. The proposed framework consists of two main stages. The first stage is for analyzing the DNA sequences evolution, whereas the other stage is for the reconstruction process. We utilized cellular automata (CA) rules for analyzing and predicting the DNA sequence. Then, a modified procedure for the reconstruction process is introduced, which is based on the Probabilistic Cellular Automata (PCA) integrated with Particle Swarm Optimization (PSO) algorithm. This integration makes the proposed framework more efficient and achieves optimum transition rules. Our innovated model leans on the hypothesis that mutations are probabilistic events. As a result, their evolution can be simulated as a PCA model. The main objective of this paper is to analyze various DNA sequences to predict the changes that occur in DNA during evolution (mutations). We used a similarity score as a fitness measure to detect symmetry relations, which is appropriate for numerous extremely long sequences. Results are given for the CpG-methylation-deamination processes, which are regions of DNA where a guanine nucleotide follows a cytosine nucleotide in the linear sequence of bases. The DNA evolution is handled as the evolved colored paradigms. Therefore, incorporating probabilistic components help to produce a tool capable of foretelling the likelihood of specific mutations. Besides, it shows their capabilities in dealing with complex relations.

Many real-world decision problems consider more than one objective. These objectives usually conflict with each other. In the problem environments, many data and parameters are usually uncertain at the time of planning. So multi-objective optimization combined with uncertainty is a challenging research topic. To deal with the basic search process in such problems, in this paper, a new approach has been proposed that combines multiple population based algorithms under a single algorithm structure. To deal with the uncertainty, a dynamic scenario-based approach is developed and integrated with the search process, in which the number of scenarios is dynamically set during the solution process. This is a new way of solving multi-objective optimization problems under uncertainty. To judge the performance of the proposed approach, we have solved a set of standard test problems without uncertainty and a number of practical problems with uncertainty. The practical problems are the well-known dynamic economic and emission dispatch problems, with different combinations of energy sources. The experimental studies demonstrated that the proposed approach performs better than other well-known algorithms compared in this paper.

This paper presents the unification and generalization of different evolving clustering methods based on Cauchy density. This can be done by introducing different inner matrix norms to obtain different functionalities of the algorithm. This unified approach with a general inner matrix norm is called eCauchy recursive clustering. The well-known possibilistic c-means clustering (PCM) can be seen as a special example of the proposed eCauchy algorithm. The main motivation of the proposed method is to solve and overcome the problems of modelling the nonlinear data streams in highly noisy environments with frequently appearing outliers. By introducing the different inner matrix into the density metric, the algorithm can be modified in different ways to deal with different clustering problems, from classical classification to the preprocessing for solving regression problems. The evolving nature of the algorithm and simple computation also make it appropriate for dealing with big-data problems. The described eCauchy algorithm needs just a few initial parameters such as minimal and maximal density. The algorithm incrementally changes the structure of the model based on the flow of samples from the data stream, more specifically it evolves the structure of the model during the operation by adding, merging, splitting and removing the clusters. This approach allows the identification of very different clusters in size and shape and is also quite insensitive to the outliers and significant noise. In the paper, the universality of the proposed algorithm is shown on various examples.

In this paper we study the reachable region queries for objects moving in obstacle space. Loosely speaking, given any one of moving objects and many other information, we are asked to return the whole region in which each location has the non-zero probability to be reachable for the moving object. To solve this problem, this paper presents an efficient algorithm that runs in O(nlog n) time and consumes O(nlog n) space, where n is the number of vertices of obstacles. Moreover, this paper establishes the lower bound for the proposed problem, showing that our solution essentially is optimal in terms of random access memory (RAM) model. Empirical study based on both real and synthetic datasets demonstrates the effectiveness and efficiency of our solution.

This paper presents an efficient use of the Ordered Fuzzy Numbers model in the description of processes undergoing dynamic changes. The purpose of this model is to provide a representation of inaccurate quantitative data, and is an alternative to the standard fuzzy numbers model proposed by Zadeh. Ordered Fuzzy Numbers have an additional feature—a direction. Giving them a new potential in applications to represent trends in imprecise values. This new feature is presented and explained via a simplified model of the volume of liquid in a reservoir. The environment is changing dynamically and is described by the changes of inflow and outflow. The proposed example explains how to interpret the direction of Ordered Fuzzy Numbers and how to understand the results of calculations and its influence. Additionally, this paper presents a comparison of the results for Ordered Fuzzy Numbers with the results of Zadeh’s classical fuzzy numbers applied to the same exemplary process.

Approximate nearest neighbor (ANN) search in high dimensional space is a fundamental operation in many applications. During the past decades, many indexing methods have been proposed to solve ANN search, including inverted index based methods, tree structures, locality sensitive hashing based methods and proximity graphs. In this paper, we explore lower bound techniques to speed up ANN search over the existing indexing methods. Since traditional lower bounds cannot reach a satisfying tradeoff between efficiency and tightness, we examine the concept of the existing lower bounds and define two new types of lower bounds, i.e., progressive lower bound (PLB) and statistical lower bound (SLB), which expand the concept of lower bound. Moreover, we propose a good instance of PLB, i.e. progressive partial distance (PPD). As the first attempt, we propose the idea of SLB and demonstrate asymmetric quantizer distance (AQD) as a good instance of SLB. Notably, AQD was only treated as an estimation of the distance in previous works, while we discover its functionality of being a lower bound in this paper. According to the extensive experiments on real data sets , we demonstrate that our lower bounds are able to obviously accelerate ANN search of the existing indexing methods, and our lower bounds outperform the existing lower bounds by a significant margin, due to their strong pruning powers. Especially, AQD speedups the state-of-the-art indexing method HNSW (Malkov and Yashunin, 2016) 1.9 times for a high recall 0.95. As to other indexing methods, the speedups of AQD are even higher, because they need to access more candidates than HNSW.

The Shortest Path in networks is an important problem incombinatorial optimization and has many applications in areas like telecommunications and transportation. It is known that this problem is easy to solve in its classic deterministic version, but it is also known that it is an NP-Hard problem for several generalizations. The Shortest Path Problem consists in finding a simple path connecting a source node and a terminal node in an arc-weighted directed network. In some real-world situations the weights are not completely known and then this problem is transformed into an optimization one under uncertainty. It is assumed that an interval estimate is given for each arc length and no further information about the statistical distribution of the weights is known. Uncertainty has been modeled in different ways in optimization. Our aim in this paper is to study the Minmax Regret path with interval data problem by presenting a new exact branch and cut algorithm and, additionally, new heuristics. A set of difficult and large size instances are defined and computational experiments are conducted for the analysis of the different approaches designed to solve the problem. The main contribution of our paper is to provide an assessment of the performance of the proposed algorithms and an empirical evidence of the superiority of a simulated annealing approach based on a new neighborhood over the other heuristics proposed.

In this paper to adapt radial moment based watermarking for reversible cases, we have developed non-unit mapped radial moments. They have been used to extract the image invariant features to produce robust geometric invariant watermarking and reversible data hiding. This aim has been achieved through using two other newly introduced methods called reversible interval phase modulation and diagonal rotation estimation. In this paper, at first a mathematical procedure to compute these newly modified radial moments has been proposed. Then a watermarking platform has been suggested which can be used for either robust or reversible data hiding fields. The proposed robust watermarking provides robustness against signal processing and both local and global geometric attacks while at the same time the achieved robust watermarked image can remain integer as the original one. Also, the proposed radial moment computation procedure has been used to produce reversible data hiding. Reversible interval phase modulation and diagonal rotation estimation methods have been proposed to avoid any type of normalization-like degradation in the embedding and extraction. Various aspects of the proposed watermarking scheme have been thoroughly examined through several experiments and the results are highly promising.

Particle swarm optimizations (PSOs) are drawing extensive attention from both research and engineering fields due to their simplicity and powerful global search ability. However, there are two issues needing to be improved: one is that the classical PSO converges slowly; the other is that classical PSO tends to result in premature convergence, especially for multi-modal problems. This paper attempts to address these two issues. Firstly, to improve the convergent efficiency, this paper proposes an asymptotic predicting model of the globally-optimal solution, which is used to predict the global optimum based on extracting the features reflecting the evolutionary trend. The predicted global optimum is then taken as the third exemplar, in a way similar to the individual historical best solution and the swarm historical best solution in guiding the evolutionary process of other particles. To reduce the probability that the population is trapped into a local optimum due to the premature phenomenon, this paper proposes an adaptive mutation strategy, which is used to help the trapped particles to escape away from the local optimum by using the extended non-uniform mutation operator. Finally, we combine the two entities to develop a globally-optimal prediction-based adaptive mutation particle swarm optimization (GPAM-PSO). In numerical experimental parts, we compare the proposed GPAM-PSO with 11 existing PSO variants by using 22 benchmark problems of 30-dimensions and 100-dimensions, respectively. Numerical experiments demonstrate that the proposed GPAM-PSO could improve the accuracy and efficiency remarkably, which means that the combination of the globally-optimal prediction-based search and the adaptive mutation strategy could accelerate the convergence and reduce premature phenomenon effectively. Generally speaking, GPAM-PSO performs most efficiently and robustly. Moreover, the performance on an engineering problem demonstrates the practical application of the proposed GPAM-PSO algorithm.

The traditional redundancy allocation problem (RAP) aims to determine a system structure that strikes an appropriate balance between series and parallel components and that maximizes reliability by using redundant components in parallel, thus satisfying resource consumption constraints. The RAP is a well-known NP-hard problem and has been an active area of research for the past four decades. By extending the seriesparallel structure to a more general network structure, this paper proposes a novel RAP called the generalized RAP (GRAP). To reduce the computational burden in solving this intractable GRAP, a new algorithm, called the hybrid swarm optimization (HSO), is proposed. The HSO combines concepts from simplified swarm optimization (SSO), particle swarm optimization (PSO), simulated annealing, and network reliability methods. The paper presents designed experiments and computational studies that demonstrate the potential of using the proposed HSO for the GRAP instead of using the genetic algorithm, PSO, or SSO.

Support vector machine (SVM) is one of the most successful classifiers for remote sensing image classification. However, the performance of SVM is mainly dependent on its parameters; in addition, for remote sensing images with high-dimensional features, feature redundancy will have a major influence on the classification efficiency and accuracy. Feature selection and parameter optimization are two important factors for improving the performance of SVM and are traditionally solved separately. In fact, these two issues are affected by each other, so to obtain the better classification performance, selection of the optimal feature subset and tuning of SVM parameters should be considered simultaneously, as they both belong to the combinatorial optimization problem and could be handled with evolutionary algorithms or swarm intelligence algorithms. In this paper, a remote sensing image classification technique based on the optimal SVM is proposed, in which the parameters of SVM and feature selection are handled integrally by a modified coded ant colony optimization algorithm combined with genetic algorithm. The results are compared with other evolutionary algorithms and swarm intelligence algorithms, such as genetic algorithm (GA), binary-coded particle swarm optimization (BPSO) algorithm, binary-coded ant colony optimization (BACO) algorithm, binary-coded differential evolution (BDE) algorithm, and binary-coded cuckoo search (BCS) algorithm. It is demonstrated that the proposed method is robust, adaptive and exhibits the better performance than the other methods involved in the paper in terms of fitness values, so could be suitable for some practical applications.

In many traditional machine learning methods, sampling is only a process of acquiring training data. However, some studies (on sequential Markov chains and particle filters) have demonstrated that sampling can be used for solving some intractable optimization problems in classical learning methods. Along this line of thinking, the relationships between sampling and learning are theoretically exploited in this paper, wherein the key feature of the sampling process is selecting representative samples from original data that can be modeled by a probability distribution. In theory, acquiring reliable samples is not an easy task for an arbitrary probability distribution. Motivated by approaches in evolutionary computation, rejection sampling and function approximation, a novel sampling strategy, called the evolutionary sampling, is proposed in this paper, and a machine learning method, called the evolutionary sampling approach (ESA), is put forward afterwards. Within ESA, a computing model, called the support sample model (SSM), is presented as well and is used to approximate an original density function. Accordingly, a concrete implementation of an evolutionary sampling approach (ESA) is proposed to seek the optimal model parameters of the SSM. Benefiting from the combination of rejection sampling and evolutionary searching, the ESA can theoretically converge to the optimal solution by minimizing the total variation distance, and can do this with high computational efficiency. Moreover, the normalized factor of a density function can be automatically estimated with high precision within the ESA. As a result, the ESA may be suitable for machine learning problems that could be transformed into density function approximation problems within a probabilistic framework. In addition, derived from the rejection sampling strategy, the ESA can also have online learning abilities required by large-scale data stream processing tasks. Theoretical analyses and application studies are carried out in this paper, and the results demonstrate that the ESA, as a novel way of machine learning, has several prominent merits aspired by past researches in machine learning.

Challenges in many real-world optimization problems arise from limited hardware availability, particularly when the optimization must be performed on a device whose hardware is highly restricted due to cost or space. This paper proposes a new algorithm, namely Enhanced compact Artificial Bee Colony (EcABC) to address this class of optimization problems. The algorithm benefits from the search logic of the Artificial Bee Colony (ABC) algorithm, and similar to other compact algorithms, it does not store the actual population of tentative solutions. Instead, EcABC employs a novel probabilistic representation of the population that is introduced in this paper. The proposed algorithm has been tested on a set of benchmark functions from the CEC2013 benchmark suite, and compared against a number of algorithms including modern compact algorithms, recent population-based ABC variants and some advanced meta-heuristics. Numerical results demonstrate that EcABC significantly outperforms other state of the art compact algorithms. In addition, simulations also indicate that the proposed algorithm shows a comparative performance when compared against its population-based versions.

This paper develops a hybrid method for nonlinear and chaotic time series forecasting based on a local linear neuro-fuzzy model (LLNF) and optimized singular spectrum analysis (OSSA), termed OSSA–LLNF. Nonlinear and chaotic time series often exhibit complex behaviour and dynamics, turning their forecasting (particularly in multi-step ahead horizons) into a difficult task. In this paper, SSA is utilized for data processing, resulting in the elimination of noisy components and improvement of forecasting performance. The SSA parameters are fine-tuned using the particle swarm optimization algorithm. Then, the processed time series is modelled and forecasted via the LLNF model. The proposed OSSA–LLNF model is applied to forecast several well-known time series with different structures and characteristics. The comparison of the obtained results with those of several old and recently developed methods indicates the superiority and promising performance of the proposed OSSA–LLNF approach.

Criminal analysis is a very complex task requiring to process huge amounts of data coming from different sources such as billings and bank account transactions in order to gain knowledge useful for an investigator. In order to support human analytic capabilities, dedicated software tools are needed, and therefore Money Laundering Detection System (MLDS) was proposed as one of such tools in our previous paper. In this paper, the social network analysis component for this system is presented. The component makes it possible to use data from bank statements and the National Court Register and construct and analyze social networks during an investigation into money laundering cases. The system can assign roles to persons from the network and allows for analysis of connections between them. The paper also includes results of experiments aimed at investigating the performance of the implemented algorithms and the correctness of the analysis.

This paper presents a vectorized matrix parameters encoding aspect for an evolutionary computer vision approach to procedural tree modeling. A serialized fixed-size floating-point encoded tree parameter set consists of a set of auxiliary local and other global parameters. The main goal of paper is to lower problem dimensionality needed for encoding local parameters. For evolution simulation, differential evolution algorithm is used. The optimizer evolves a parameterized procedural model by fitting a set of its rendered images to a set of automatically preprocessed reference photo images. The reconstructed tree morphology is then used for reconstructed tree animation, to generate similar geometrical tree models based on similar morphology. Examples of reconstructed model animation are shown, such as simulation of its growth, sway in the wind, or adding leaves.

Formal models can be useful in computer graphics as a conceptual framework supporting representation systems. This allows to formally derive properties and algorithms and proof their correctness and validity. This paper describes a formal model based on a geometric algebra. This algebra has been used to obtain specific representation systems and study their equivalence. The representation systems derived in a natural way from this model are based on simplicial coverings and can be applied to non-manifold solids and to solids with holes. Representations have been developed for polyhedral and free-form solids. Algorithms described and proved include boolean operations and representation conversion. The paper covers the three abstraction levels: theoretical model, representations and derived algorithms. As a practical application an experimental modeller for free-form solid has been developed (ESC-MOD system: ''Extended Simplicial Chains MOdeller'').

K-nearest neighbors search (K-NNS) is a fundamental problem in many areas of machine learning and data mining. In an attempt to solve NNS problems by locality-sensitive hashing (LSH)-based algorithms and avoid the F1-trap, in this paper, we theoretically prove the necessity of maintaining close edges of each dimension, the independence of the partition location and the precision of retrieval. Based on the theories, a new method is proposed for constructing the hash function family. Then, a volume-balancing locality-sensitive hashing (VBLSH) algorithm is implemented. In this algorithm, the problem of constructing the hash function family is transformed into an optimization problem under a fixed number of hypercubes. By solving the optimization problem, different hyperplane intervals are set for each dimension to ensure that the hash functions can evenly segment the data space, where each segmented hypercube has a close volume and close side lengths. Finally, the analysis of the algorithm shows that the VBLSH algorithm has a high index coding efficiency and hash code calculation efficiency. The experimental results show that the VBLSH algorithm can simultaneously take into account the efficiency, precision and recall rate, and avoid the F1-trap; it shows excellent effectiveness on 4 datasets.