Search

• The event-triggered mechanism in this paper does not involve any global information. • This paper assumes this sub-graph is directed and the control input of leader is derivable. The main obstacle is the complex and unpleasant interaction between the non-linear function used to handle the control input of leader and the directed sub-graph between followers. • This paper designs a fully distributed event-triggered controller, thus, only needs discrete communication among neighboring agents, which is more suitable to be utilized in practical engineering. This article studies the design method of fully distributed event-triggered output feedback protocol for linear multi-agent systems with a derivable leader under directed graphs. The research on such problems in the existing reference needs complete state feedback and the state variable of the adaptive law used in the literature controller is monotonic and non decreasing, which may make the control input exceed the actual saturation value, thus affecting the practical application. How to design a completely distributed event-triggered adaptive output feedback protocols for multi-agent systems with a derivable leader on directed graphs is an open issue. In order to overcome the difficulty of designing the controller without complete state information, we design the event-triggered output feedback protocol, and with the help of σ -modification techniques, ensure the boundedness of the adaptive law in proposed protocol. The conditions to ensure the realization of the leader–follower consensus and exclude Zeno behavior are given. Furthermore, the control protocol is fully distributed and uses neighbor information only when the event trigger conditions are satisfied. Finally, the availability of the proposed protocol is illustrated by simulation examples. [ABSTRACT FROM AUTHOR]

In this paper, the communication network associated with a linear heterogeneous clustered multi-agent system (CMAS) contains several clusters, each of which is modeled by a strongly connected digraph and contains a leader. Intra-cluster agents are in continuous communication with their intra-cluster neighbors during the communication period, whereas only the inter-cluster leaders can communicate with other leaders at a series of discrete moments, known as the reset time. Thus this paper presents a reduced-order observer-based reset output feedback controller that solves the output synchronization problem associated with the linear heterogeneous CMAS. Specifically, a reset internal model is built to solve the output synchronization for each agent, and a reset reduced-order observer is developed to estimate the unavailable states of each agent. Then an output feedback control strategy is developed using only the output information, internal model state, and observer state. Sufficient conditions are also obtained for ensuring the output synchronization of the CMAS. Finally, an illustrative example is provided to demonstrate the efficiency of the proposed control strategy. [ABSTRACT FROM AUTHOR]

• This paper presents MASK, a multilevel algorithm for approximate similarity search • MASK can distribute the index over as many computing nodes as we can afford. • Experimental results show the applicability of this novel indexing method • MASK achieves superior performance with high-dimensional and high-sparsity datasets. Similarity search based on a distance function in metric spaces is a fundamental problem for many applications. Queries for similar objects lead to the well-known machine learning task of nearest-neighbours identification. Many data indexing strategies, collectively known as Metric Access Methods (MAM), have been proposed to speed up these queries. Moreover, since exact approaches to solving similarity queries can be complex and time-consuming, alternative options have emerged to reduce query execution time, such as returning approximate results or resorting to distributed computing platforms. In this paper, we introduce MASK (Multilevel Approximate Similarity search with k -means), an unconventional application of the k -means algorithm as the foundation of a multilevel index structure for approximate similarity search suitable for metric spaces. We show that this method leverages inherent properties of k -means for this purpose, like representing high-density data areas with fewer prototypes. An implementation of this new indexing procedure is evaluated using a synthetic dataset and two real-world datasets in high-dimensional and high-sparsity spaces. Experimental tests show that MASK performs better than alternative algorithms for approximate similarity search. Results are promising and underpin the applicability of this novel indexing method in multiple domains. [ABSTRACT FROM AUTHOR]

• Temporal information is effectively integrated into the representation model. • A new prediction model is proposed to gain temporal information. • The memory capacity of the replay buffer is smaller than the existing methods. • The policy lag is proven to be decreased quickly in maximum-entropy reinforcement learning. Partially observable Markov decision process (POMDP) is a key challenging problem in the application of reinforcement learning since it comprehensively describes real agent-environment interactions. Recent works mainly utilize conventional reward signals to train a representation that converts POMDPs to MDPs. However, rewards alone are not enough for a good representation without temporal information. In this paper, we first introduce a novel Recurrent Prediction Model to integrate temporal information into the representation that solves POMDP problems by training three additional unsupervised prediction models, named transition model, reward recovery model, and observation recovery model. This paper secondly makes a modification of the data structure of vanilla replay buffer to reduce the memory usage and thirdly proposes an off-policy correction algorithm to decrease the policy lag in POMDPs. The experiments show that our model achieves better performance in partially observable environments on both stand-alone and distributed training systems. [ABSTRACT FROM AUTHOR]

• An adaptive random competition strategy is designed. • A neighborhood crossover strategy to improve convergence speed and accuracy. • A more robust multi-strategy competitive-cooperative co-evolution algorithm is presented. In order to effectively solve multi-objective optimization problems (MOPs) and fully balance uniformity and convergence, a multi-strategy competitive-cooperative co-evolutionary algorithm based on adaptive random competition and neighborhood crossover, namely MSCOEA is developed in this paper. In the MSCOEA, a new adaptive random competition strategy is designed to determine whether one sub-population loses diversity through the performance. A random competition process is executed to increase the sub-population diversity in order to compete for participation opportunities in the next iteration. And the extra population is employed to store the found non-dominated solutions. A new neighborhood crossover strategy is designed to enhance the local search ability. Finally, three different types of multi-objective benchmark functions are selected to verify the effectiveness of the MSCOEA. The experiment results show that the MSCOEA can effectively balance convergence and uniformity, and obtains better optimization performance and robustness by comparing with other algorithms. The convergence performance of the adaptive random competition and the neighborhood crossover strategies are also analyzed in detail. [ABSTRACT FROM AUTHOR]

In numerous binary classification tasks, the two groups of instances are not equally represented, which often implies that the training data lack sufficient information to model the minority class correctly. Furthermore, many traditional classification models make arbitrarily overconfident predictions outside the range of the training data. These issues severely impact the deployment and usefulness of these models in real life. In this paper, we propose the boundary regularizing out-of-distribution (BROOD) sampler, which adds artificial data points on the edge of the training data. By exploiting these artificial samples, we are able to regularize the decision surface of discriminative machine learning models and make more prudent predictions. Next, it is crucial to correctly classify many positive instances in a limited pool of instances that can be investigated with the available resources. By smartly assigning predetermined nonuniform class probabilities outside the training data, we can emphasize certain data regions and improve classifier performance on various material classification metrics. The good performance of the proposed methodology is illustrated in a case study that consists of both benchmark balanced and imbalanced classification data sets. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNNs) are vulnerable to adversarial examples generated by adding subtle perturbations to benign inputs. While these perturbations are somewhat small due to the L p norm constraint, they are still easily spotted by human eyes. This paper proposes Perceptual Sensitive Attack (PS Attack) to address this flaw with a perceptually adaptive scheme. We add Just Noticeable Difference (JND) as prior information into adversarial attacks, making image changes in areas that are insensitive to the human eyes. By integrating the JND matrix into the L p norm, PS Attack projects perturbations onto the JND space around clean data, resulting in more imperceivable adversarial perturbations. PS Attack also mitigates the trade-off between the imperceptibility and transferability of adversarial images by adjusting a visual coefficient. Extensive experiments manifest that combining PS attacks with state-of-the-art black-box approaches can significantly promote the naturalness of adversarial examples while maintaining their attack ability. Compared to the state-of-the-art transferable attacks, our attacks reduce LPIPS by 8% on average when attacking typically-trained and defense models. [ABSTRACT FROM AUTHOR]

Time series forecasting has wide applications in our daily lives, such as meteorological warnings and decision-making. However, traditional supervised models do not perform well on forecasting tasks due to the lack of annotated training data available in real time series. Recently, researchers have proposed self-supervised methods, especially the contrastive learning approach, which can build a universal representation framework for downstream tasks through data augmentation and contrastive loss. Although it alleviates the shortage of labeled data, the data augmentation approach directly transferred from computer vision is not appropriate for the time domain due to noise vectors and unrelated variables that may interfere with the accuracy of representation. In this paper, we propose a novel time series forecasting model based on disentangled seasonal-trend representation named ACST. It employs an improved cycle generative adversarial data augmentation method to generate samples close to real data for contrastive loss. Moreover, we apply gated residual networks and a noise decomposition module to reduce the impact of different noise vectors and feature variable weights on the results. Extensive experiments show that ACST achieves an average improvement of 26.8% on six benchmarks. [ABSTRACT FROM AUTHOR]

In this paper, we consider the resilient distributed estimation in adversarial networks, compromised by false data injection (FDI) attacks. We develop a two-step distributed approach to detect the compromised nodes of the network due to FDI attacks. Specifically, we develop a distributed correntropy-based discrimination scheme followed by a distributed state perception scheme to distinguish the compromised nodes from the secure, such that the reliable neighbors could be determined by each node in a distributed manner. We further theoretically analyze the mean and mean-square performance of the resilient distributed estimation, and propose an optimal reference neighbor selection scheme to enhance the network estimation performance. Illustrative simulations validate the superior performance of both FDI attack detection and parameter estimation of the proposed distributed algorithms for adversarial networks, especially under heavy and time-varying FDI attacks. The theoretical results well predict the performance of the proposed resilient distributed estimation algorithm. [ABSTRACT FROM AUTHOR]

In this paper, we present a new classifier called elastic net twin support vector machine (ETSVM). It resolves two smaller-sized quadratic programming problems (QPPs) similarly to the twin support vector machine (TSVM). The key difference between them is that ETSVM uses elastic net penalty for slack variables, which enhances classification performance. The dual QPPs of the ETSVM do not involve matrix inversion, in contrast to conventional TSVM. As a result, ETSVM can avoid the ill-conditioning case. We theoretically discuss its properties, including exploring the violation tolerance upper bound for the two QPPs of ETSVM. In order to increase computing efficiency, we derive a sequence of safe screening rules using variational inequalities to quicken the ETSVM parameter tuning process (SSR-ETSVM). We compare the proposed algorithm with SVM, nonparallel hyperplane SVM, elastic net SVM, and elastic net nonparallel hyperplane SVM. Numerical experiments confirm the rationality and effectiveness of the proposed methods. [ABSTRACT FROM AUTHOR]

The dynamic event-triggered control issue for multi-input and multi-output (MIMO) nonlinear systems with unknown dynamics and disturbances is solved in this paper. It is the first attempt to design a dynamic event-triggered mechanism for MIMO systems, which can further reduce data transmission from controller to actuator, thereby producing fewer event. The unknown dynamics of MIMO systems are approximated by neural networks. Then, the composited uncertain vector consisting of approximated terms and disturbances is transformed into a linear parameterized form with a single parameter. In addition, the dynamic event-triggered adaptive control method is designed under the backstepping framework. Finally, the validity of the presented dynamic event-triggered control solution is verified by simulation example and comparison analysis. [ABSTRACT FROM AUTHOR]

Code clones are comparable code snippets that are introduced into software by developers in order to increase software development productivity. A change to code clone may result in a consistent-defect if the developers forget to verify the consistency of the code after the change. To reduce such change-related maintenance costs, researchers have proposed a number of methods for predicting clone consistency in advance. Unfortunately, the effectiveness of these cross-project models is unsatisfactory, and performing such predictions with insufficient data remains a challenge. Meanwhile, cross-project defect prediction via transfer learning method is prevalent in the software engineering community. Consequently, we first construct an empirical study to explore whether transfer-learning techniques could well be utilized for clone cross-project consistent-defect prediction in the initial stages of software development. In this paper, we employ transfer-learning techniques to predict clone consistency at both the time of clone creating and clone changing in order to avoid clone consistent-defects and maintenance. We conduct an experiment on open-source projects to evaluate the effectiveness of various transfer-learning methods. Our investigation demonstrates that transfer-learning techniques have a beneficial impact on predicting cross-project clone consistent-defect, and that the size of the dataset also has a positive effect on prediction. In order to promote software safety and security, we recommend that developers leverage transfer-learning to enhance the capability for clone cross-project consistent-defect prediction early in the software development phase. • We propose a clone cross-project consistent-defect prediction approach for both clone-creating and changing times. • We first investigate the effectiveness of transfer-learning methods for clone cross-project consistent-defect prediction. • We conduct experiments on eight open-source projects to address our research problem. [ABSTRACT FROM AUTHOR]

The best-worst method is a recently developed approach that addresses pairwise comparisons in multi-criteria decision-making problems. It has attracted the interest of many academics and has gained widespread use for its effectiveness in reducing the pairwise comparison times and for its strong performance in maintaining consistency between judgments. However, the existing best-worst method prioritization methods don't consider the indirect judgments and fail to meet several established criteria. This paper first develops two prioritization methods, the approximate eigenvalue method and the logarithmic least squares method, to the best-worst method. Then, the inconsistency thresholds of the proposed prioritization methods are given. Finally, the Monte Carlo simulation method is applied to generate random matrixes to compare the performance of various prioritization methods on the selected criteria and analyze the relationship between different consistency measures. It is found that the logarithmic least squares method is the best prioritization method because of simple calculation, considering indirect judgments and containing minimum violations. [ABSTRACT FROM AUTHOR]

Quantum algorithms can enhance machine learning in different aspects. The quantum support vector machine was proposed to improve the performance, in which the Swap Test plays a crucial role in realizing the classification. However, as the Swap Test is destructive, the quantum support vector machine must be repeated in preparing qubits and manipulating operations. This paper proposes a quantum support vector machine based on the amplitude estimation (AE-QSVM) which gets rid of the constraint of repetitive process and saves the quantum resources. At first, a generalized quantum amplitude estimation is introduced in which the initial state can be arbitrary instead of being |0〉. Then, AE-QSVM is trained by the quantum singular value decomposition and a query sample is classified by the generalized quantum amplitude estimation. In AE-QSVM, a high accuracy can be achieved by adding auxiliary qubits instead of repeating the algorithm. The time and space complexity of AE-QSVM are reduced compared with other algorithms. Finally, we ran experiments on the IBM's quantum computer and experimental results demonstrate that classification with a 95% probability of success only uses 12 qubits. • The generalized amplitude estimation can take any quantum state as the initial state. • The complexity of AE-QSVM is reduced compared with other algorithms. • Quantum support vector machine gets rid of the constraint of repetitive process. • Classification is realized on the IBM quantum computer using 12 qubits. [ABSTRACT FROM AUTHOR]

Multi-source information fusion is a sophisticated estimation process that generates a unified profile to assess complex situations. Dempster–Shafer evidence theory (DSET) is a practical theory in handling uncertain information in multi-source information fusion. However, highly-conflicting evidence may cause the combination rule of Dempster to provide counterintuitive results. Thus, how to effectively reconcile highly-conflicting evidence in DSET is still an open issue. To address this problem, in this paper, a novel belief divergence, higher order belief Jensen-Shannon divergence is proposed to measure the discrepancy between BPAs in DSET. The proposed higher order belief Jensen-Shannon divergence is the first method to dynamically measure the discrepancy between BPAs over the time evolution, i.e., to measure the discrepancy between BPAs with different time scale in the future. Besides, the proposed higher order belief Jensen-Shannon divergence has benefits from the perspective of measurement. It satisfies the properties of nonnegativeness and nondegeneracy, symmetry, and the triangle inequality of root form. Based on the proposed higher order belief Jensen-Shannon divergence, a novel multi-source information fusion algorithm is proposed. Eventually, the proposed algorithm is applied to a pattern classification experiment with real-world datasets. [ABSTRACT FROM AUTHOR]

This paper focuses on the problem of the fully distributed fixed-time bipartite output consensus tracking for nonlinear heterogeneous multiagent systems (MASs) via both state-feedback and output-feedback methods under the switching topology. The adaptive fully distributed state observers with quantization information are designed to eliminate the dependence on the Laplacian matrix. For MASs with unknown model matrix, a novel fixed-time observer-based regulator equation is employed, which avoids repeatedly getting the unnecessary solution of universal one in the time-varying cooperation-competition communication topology. In this case, both fixed-time state-feedback and output-feedback controllers are constructed such that the output consensus tracking is achieved regardless the state value is available or unavailable. Besides, the upper bound of convergence time can be adjusted only by parameters without initial states. Lyapunov functions are established to derive conditions of achieving consensus tracking by mathematical analysis. Eventually, the effectiveness of the proposed control strategy is manifested by simulation results. [ABSTRACT FROM AUTHOR]

Ensemble classifier is a well-known method that has been used to solve several machine learning problems. To have reliable results, one should ensure the build of a good ensemble. In order to do so, researchers have proposed some heuristics like Random Subspace Ensemble (RSM), Rough set bas. The drawback of these mentioned approaches is their disability to handle uncertain data especially when uncertainty is represented by the evidence theory. The aim of this paper is to adapt both RSM and Rough set based ensemble in order to let them working in the context of evidential data. Three ensemble classifier approaches based on the rough set theory have been proposed and have been compared with each other. For the comparison purpose, we have relied on Ensemble Enhanced Evidential k Nearest Neighbor (EE k -NN) classifier, real world datasets from the UCI repository as well as synthetic databases. [ABSTRACT FROM AUTHOR]

This paper deals with the stability analysis of generalized neural networks (GNN) with interval time-varying delay, and some novel stability criteria are proposed. For the delay, more specifically, it has the upper and lower bounds but the information of its derivative is unknown (or itself is not differentiable). Firstly, a new Lyapunov-Krasovskii functional (LKF) is constructed to obtain the less conservative stability results. Then, in order to cooperate with the introduced LKF, the free-matrix-based inequality (FMBI) is employed in the estimation of the integral quadratic terms. Next, by utilizing the new negative definite conditions (NDCs) of matrix-valued cubic polynomials, the stability conditions are presented in the form of linear matrix inequalities (LMIs). In addition, in the establishment process of these conditions, no extra state variables are introduced and the positive definite constraint on the LKF is relaxed. Finally, the validity and advantages of the proposed conditions are illustrated with some classical numerical examples. [ABSTRACT FROM AUTHOR]

Ensemble learning is a widely used methodology to build powerful predictors from multiple individual weaker ones. However, the vast majority of ensemble learning models are designed for offline application scenarios, the use of evolving fuzzy systems in ensemble learning for online learning from data streams has not been sufficiently explored, yet. In this paper, a novel self-adaptive fuzzy learning ensemble system is introduced for data stream prediction. The proposed ensemble system employs the very sparse random projection technique to compress the consequent parts of the learned fuzzy rules by individual base models to a more compressed form, thereby reducing redundant information and improving computational efficiency. To improve the overall prediction performance, a dynamical base model pruning scheme is introduced to the proposed ensemble system together with a novel inferencing scheme, such that less accurate base models will be removed from the ensemble structure at each learning cycle automatically and only these more accurate ones will be involved in joint decision-making. Numerical examples based on a wide range of benchmark datasets demonstrate the stronger prediction performance of the proposed ensemble system over the state-of-the-art alternatives. [ABSTRACT FROM AUTHOR]

This paper addresses the design problem of the discrete-time stable unknown input estimator (UIE) based on parameter optimization using the artificial bee colony (ABC) algorithm. First, a stability-guaranteed design method for UIEs is presented, and a sufficient condition for the applicability of the design is provided. Next, to design UIEs with good disturbance rejection properties, a new objective function is developed that incorporates both waveform-based and norm-based performance criteria to allow direct evaluation of the adverse effects of disturbances on system performance. Finally, the proposed design method is compared with the previous one using an objective function based on the estimated disturbance to confirm the improvement of the disturbance rejection properties at the plant output. Furthermore, as another approach to the design of stable UIEs, the original UIE design method is combined with a constrained ABC algorithm, and the approach is compared with the proposed one in terms of disturbance rejection properties. • A design of the discrete-time stable unknown input estimator is addressed. • A new objective function is developed that incorporates both waveform-based and norm-based performance criteria. • The artificial bee colony algorithm is used to efficiently optimize the design parameters in the unknown input estimator. • A comparison with the constrained optimization approach is made. [ABSTRACT FROM AUTHOR]

Many issues require the application of forecasting models applied to spatiotemporal data in Geographic Information Systems (GIS) to predict the spatial distribution and evolution of a specific feature. The use of soft computing techniques in the development of these forecasting models makes it possible to detect non-linear trends but has the disadvantage of increasing the computational complexity of the model. In this paper we present a GIS-based framework in which a fast soft computing forecasting model based on the multidimensional Fuzzy Transform (for short, MF-transform) is applied to evaluate the spatial distribution and the time evolution in a study area of a measurable entity (the feature). The study area is divided into homogeneous zones (the subzones) in which the feature was measured in each time frame. The time series of the feature are analyzed to assess the trend of the feature in subsequent time frames; furthermore, those sub-areas are detected in which the feature is higher than a maximum threshold (hot spots) or lower than a minimum threshold (cold spots) in this time range. A process of fuzzifying the values of the feature is carried out in order to facilitate the interpretation of the results by expert users. The framework was tested on a study area provided by the province of Naples (Italy) to predict and analyze the spatial distribution and temporal trend of the monthly rate of births compared to deaths. Furthermore, the thematic map of the hot and cold spots detected in the three months following the time period of measurements was built. The results show that our method provides reliable results both in terms of forecast error and similarity between the detected hot and cold spots and those who have really formed. [ABSTRACT FROM AUTHOR]

Network pruning has been widely adopted for reducing computational costs and memory consumption in low-resource devices. Recent studies show that the lottery ticket hypothesis achieves high accuracy under high compression ratios (i.e. , 80–90% of the parameters in original networks are removed). Nevertheless, finding well-trainable networks with sparse parameters (i.e. , < 10% of the parameters remaining) is still a challenging task, commonly believed by the lack of model capacity. This paper revisits the training process of existing pruning methods and observe that dead connections , which do not contribute to model capacity, occur in existing pruning methods. To this end, we propose a novel pruning method, namely all-alive pruning (AAP), which produces the pruned networks with only trainable weights with no dead connections. Notably, AAP is broadly applicable to various pruning methods and model architectures. We demonstrate that AAP equipped with the existing pruning methods (e.g. , iterative pruning, one-shot pruning, and dynamic pruning) consistently improves the accuracy of the original methods at high compression ratios on the various image- and language-based tasks. [ABSTRACT FROM AUTHOR]

The existence of a large amount of partially labelled data is due to the high cost of labelled data. For this type of data, traditional rough set models cannot well represent the distribution of objects in real data. This limits effective decision-making and classification of the data. Replacing the missing labels with prediction labels can compensate for the absence of labels to some extent. Then, multiset-valued data with prediction labels is considered. This paper studies semi-supervised attribute reduction for partially labelled multiset-valued data via a prediction label strategy. First, the distance between information values in a multiset-valued decision information system (MSVDIS) is constructed, and the tolerance relation on the object set of an MSVDIS is introduced. Then, a partially labelled multiset-valued decision information system (p-MSVDIS) is defined. Next, a prediction label strategy (i.e., the existing labels in a p-MSVDIS remain the same, and the missing labels are replaced with the prediction label) is proposed. Incidentally, a new MSVDIS is obtained, and the object set is reclassified by the decision attribute. Moreover, the dependence and conditional entropy in the obtained MSVDIS are proposed, and some properties are discussed. Semi-supervised attribute reduction algorithms in a p-MSVDIS via dependence and conditional entropy are designed. Finally, experiments on real datasets show that the prediction label strategy outperforms the traditional rough set approach and that the designed algorithms have better classification and outlier detection performance than the existing algorithms. [ABSTRACT FROM AUTHOR]

Social networks have become an important aspect of our modern times and are gradually becoming an integral means of communication worldwide. Overwhelming amounts of data are being transferred over social networks every day. Hence ensuring security becomes a necessity. Suspicious users or spammers may pose a threat to the information and data shared by users over the network. With this in mind, outliers detection is a crucial aspect of network communication. In our paper, a new technique is proposed to identify the anomalies in a network from a global perspective by using the network community structure. In general, state-of-the-art outliers detection algorithms mainly focus on the individual nodes and their direct neighborhood. But our technique considers only those nodes which tend to belong to multiple communities or whose neighbors belong to the same community or do not belong to any community. Results after experiments on synthetic and real-world networks show an improvement of 7-10% and 29% in F-Score and Jaccard similarity, respectively, compared to the state-of-the-art algorithms. Furthermore, we achieve almost 1.83 times speedup compared to the state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Priority rules are one of the most common and popular approaches to real-time scheduling. Over the last decades, several methods have been developed to generate rules automatically. In addition, it has been shown that combining rules into ensembles is better than using a single rule in many cases. In this paper, we analyze different ways to create and use ensembles previously developed through genetic programming. In our study, we classify ensembles as either collaborative or coordinated, depending on how the rules are used. In the first case, all the rules contribute to the creation of the same solution, while in the second case, each rule works independently on its own solution, and the best of them is selected as the solution of the ensemble. We found that each method has its own strengths and weaknesses, which leads us to use them in combination. Based on this hypothesis, we developed new methods to design and combine collaborative and coordinated ensembles and evaluated these methods for the One Machine Scheduling Problem with time-varying capacity and minimization of total tardiness. The results of the experimental study provided interesting insights into the use of ensembles and showed that our proposals outperform previous methods. • Adaptation of existing methods for constructing collaborative ensembles. • Combination of collaborative and constructive ensembles to improve their efficiency. • Analysis of ensemble cardinality and combination methods on performance. [ABSTRACT FROM AUTHOR]

This paper studies predefined-time leader-follower consensus issue for second-order multi-agent system (MAS) with communication noises over digraph. Crosstalk resistant adaptive noise canceller is utilized to filter out communication noise. Next, with the filtered states received from its neighbors, a prescribed-time distributed observer is first proposed, which requires no global information and provides prescribed-time leader's states estimation under noisy communication environment and directed communication graph. Then, an adaptive predefined-time control strategy is presented to counteract the mismatched disturbance caused by estimation error and achieve practical predefined-time leader-follower output consensus of the studied MAS. Lyapunov stability analysis proves that the developed distributed observer achieves prescribed-time convergence and the developed tracking controller achieves predefined-time convergence. Finally, the developed control strategy is applied to consensus tracking of networked single-link robotic manipulators and simulation results show its effectiveness. • Solve predefined-time consensus tracking of second-order MAS with communication noise. • Propose prescribed-time distributed observer under communication noise over digraph. • Propose a new predefined-time consensus control protocol. • Propose an adaptive noise canceller to eliminate communication noise. [ABSTRACT FROM AUTHOR]

Recent researches show that deep learning model is susceptible to backdoor attacks. Many defenses against backdoor attacks have been proposed. However, existing defense works require high computational overhead or backdoor attack information such as the trigger size. In this paper, we propose a novel backdoor detection method based on intentional adversarial perturbations. The proposed method leverages intentional adversarial perturbation to detect whether an image contains a trigger, which can be applied in both the training stage and the inference stage (sanitize the training set in training stage or detect the backdoor instances in inference stage). Specifically, given an untrusted image, the adversarial perturbation is added to the image intentionally. If the prediction of the model on the perturbed image is consistent with that on the unperturbed image, the input image will be considered as a backdoor instance. Compared with most existing defense works, the proposed method is faster and introduces less computational overhead during backdoor detection process. Moreover, the proposed method maintains the visual quality of the image (as the ℓ 2 norm of the added perturbation is as low as 2.8715, 3.0513 and 2.4362 on Fashion-MNIST, CIFAR-10 and GTSRB datasets, respectively). Experimental results show that, for general backdoor attack, the backdoor detection rate of the proposed defense method is 99.63%, 99.76% and 99.91% on Fashion-MNIST, CIFAR-10 and GTSRB datasets, respectively. For invisible backdoor attack, the backdoor detection rate of the proposed defense method is 99.75% against blended backdoor attack and 98.00% against sample-specific backdoor attack. It is also demonstrated that the proposed method can achieve high defense performance against backdoor attacks under different attack settings (trigger transparency, trigger size and trigger pattern). In addition, the experimental comparison with related work demonstrates that the proposed method has better detection performance and higher detection efficiency. [ABSTRACT FROM AUTHOR]

Batch-constrained reinforcement learning constrains the learned policy to be close to the behavior policy, which holds a tremendous promise for alleviating the distributional shift in offline reinforcement learning. Existing batch-constrained techniques rely on perturbation models to adjust the actions generated from the generative model to maximize the estimated value function. However, the perturbation model deviates from the distribution of the offline datasets and introduces a new distribution drift problem, which affects the performance of the learned policies. In addition, since offline reinforcement learning cannot learn by trial and error, the final policies are often prone to failure in reality or make unsafe decisions when trained with a noisy or small size dataset. To address the above issues, this paper employs constrained generative adversarial network to generate actions with given states. Specifically, we train the generator to maximize the estimated value and constrain the state-action pairs to follow the dataset distribution. The perturbation model is trained to maximize the probability of the perturbed actions belonging to the dataset and minimize the likelihood of taking dangerous actions. Moreover, we utilize safety critics to predict the risk of the actions under a state. Experimental results show that the proposed method is effective and can choose safe actions while maintaining a high performance in offline settings. [ABSTRACT FROM AUTHOR]

Classification with incomplete data is practically challenging since missing data hinders classification. The two-stage strategy of filling missing values first and then classifying complete data degenerates performance due to the separate optimization of classification and imputation. While efforts have been made to classify incomplete data directly without imputation, the benefits of imputation on classification have been sacrificed. To this end, this paper proposes a Category-Aware Optimal Transport Neural Network (CAOT-NN) for incomplete data classification. Specifically, we design a category-aware optimal transport method conditioned on intra-category data distribution, which explicitly employs the category information of each sample and fills in discriminative values for the downstream classification module. Moreover, we reconstruct the observed values during the imputation process when performing classification learning, thus implicitly utilizing the category information of the classifier for missing values estimation. Extensive experiments conducted on 16 real-world and 30 synthetic datasets from the UCI repository demonstrate the superiority of CAOT-NN against existing state-of-the-art methods. Also, the visualization analysis reveals that CAOT-NN enables a tighter intra-category distribution of the imputed data, implying that the missing mechanism may be utilized as a data augmentation strategy to boost accuracy. • CAOT-NN can make the imputed values more favorable for classification. • A category-aware imputation loss function is defined based on optimal transport. • Good imputation performance doesn't improve incomplete data classification results. [ABSTRACT FROM AUTHOR]

Usually, manual labeling of large-scale electrocardiograms (ECGs) for deep learning is always expensive, as it requires considerable effort and time from cardiologists. Currently, contrastive learning can utilize unlabeled ECGs to improve deep learning using insufficient labels. However, this field may still lack instructive literature. In this paper, a dense lead contrast (DLC) is proposed for effective contrastive learning on multilead ECGs. It develops contrastive learning between any two leads from different views to explore intralead and interlead invariance. A joint loss combining intralead and interlead contrastive loss guides the DLC pretraining. Moreover, DLC introduces a multibranch network (MBN) for contrastive learning, generates a representation for each lead, and fuses all the leads for a global representation. In the downstream tasks, DLC outperforms the standard contrastive learning paradigm of multilead ECGs by 2.78%∼6.59% in AUROC (linear probe). Using only 10% of the labeled training data, it still outperforms standard contrastive learning by a significant margin in AUROC. Compared with existing methods, DLC shows obvious advantages in all the experiments. Therefore, DLC may be more suitable for the contrastive learning of multilead ECGs. Its good performance based on insufficient labels can alleviate the cardiologists' burden from data labeling. • Dense lead contrast (DLC) conducts an effective self-supervised pre-training. • DLC jointly explores intra- and inter-lead invariance of electrocardiograms (ECGs). • Multiple branch network (MBN) architecture organizes all the leads. • Raw ECGs from a real-world hospital are used to validate algorithm robustness. [ABSTRACT FROM AUTHOR]

Dynamic multiobjective optimization problems (DMOPs) vary over time, requiring an optimization algorithm to track the position of Pareto-optimal front (PF) in a dynamic environment. To achieve that, a novel co-evolutionary algorithm based on problem analysis (CAPA) is proposed in this paper. CAPA is designed to solve DMOPs from decision space and objective space simultaneously, which is achieved by the combination of adjustable prediction (AP) and precise mapping strategy (PM). In decision space, the proposed multi-model prediction can estimate the location of new population based on the historical median points. In objective space, a novel sampling method is developed to search for sample points with better convergence or diversity. Then, mapping these sample points back to decision space based on inverse model. Through the problem analysis mechanism, the proportion of the new solutions produced by each strategy changes adaptively. CAPA is incorporated into the dynamic multiobjective evolutionary algorithm (DMOEA) based on decomposition (MOEA/D) to construct a novel algorithm. The efficacy of CAPA is validated by comparison with five state-of-the-art algorithms on 28 benchmarks. Experimental results show that CAPA has the ability to generate high quality population uniformly along PF. [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]

With the proliferation of mobile technology and IT development, people can use social network services anywhere and anytime. Among many social network mining problems, identifying cohesive subgraphs has attracted extensive attention from different fields due to its numerous applications. The most widely used among many cohesive subgraph models is k -core, because of its simple and intuitive structure. In this paper, we formulate (p , n) -core in signed networks by extending the k -core model. (p , n) -core is designed to guarantee sufficient internal positive edges and deficient internal negative edges. We formally prove that finding an exact (p , n) -core is NP-hard. Hence, we propose three efficient and effective algorithms to find a solution. We demonstrate the superiority of our proposed algorithms using real-world and synthetic networks. [ABSTRACT FROM AUTHOR]

Real-world multi-objective optimization problems encounter different types of uncertainty that may affect the quality of solutions. One common type is the stochastic noise that contaminates the objective functions. Another type of uncertainty is the different forms of dynamism including changes in the objective functions. Although related work in the literature targets only a single type, in this paper, we study Dynamic Multi-objective Optimization problems (DMOPs) contaminated with stochastic noises by dealing with the two types of uncertainty simultaneously. In such problems, handling uncertainty becomes a critical issue since the evolutionary process should be able to distinguish between changes that come from noise and real environmental changes that resulted from different forms of dynamism. To study both noisy and dynamic environments, we propose a flexible mechanism to incorporate noise into the DMOPs. Two novel techniques called Multi-Sensor Detection Mechanism (MSD) and Welford-Based Detection Mechanism (WBD) are proposed to differentiate between real change points and noise points. The proposed techniques are incorporated into a set of Dynamic Multi-objective Evolutionary Algorithms (DMOEAs) to analyze their impact. Our empirical study reveals the effectiveness of the proposed techniques for isolating noise from real dynamic changes and diminishing the noise effect on performance. [ABSTRACT FROM AUTHOR]

This paper is concerned with the communication protocol-based zonotopic state and fault estimation issue for a class of discrete time-varying industrial cyber-physical systems under bounded disturbances and constrained communication resources. First, in order to alleviate resource consumption, a round-robin protocol is developed to schedule the data transmission between the local sensor and the remote estimator. Second, a zonotopic set-membership estimation algorithm is developed to guarantee that the unavailable system state and fault signals could be simultaneously estimated and enclosed in a zonotope at every instant of time. Subsequently, the size of such a zonotope is minimized at each iteration by appropriately selecting a correlation matrix. Furthermore, the rigorous stability and performance analysis for the developed zonotopic estimation algorithm is provided. Finally, two simulation examples are provided to verify the obtained theoretical results. • Novel zonotopic set-membership estimation algorithm, where the zonotope guarantees to contain unknown system state and fault a rigorous stability analysis for the developed estimation scheme is provided by analyzing the boundedness of zonotope. • Finally two simulation examples are provided to show the effectiveness of the proposed estimation scheme. [ABSTRACT FROM AUTHOR]

This paper addresses the finite-time synchronization problem of complex networks under a new communication constraint. The situation that some communication channels used to transmit sub-information between two nodes may fail is considered, which causes the receiving node to be unable to obtain complete state information. Moreover, due to less information transmission between nodes, the analysis of finite-time synchronization problem becomes more difficult. Therefore, several questions naturally arise here: Under this communication restriction, how to remodel the complex network, and can the network achieve finite-time synchronization? To better solve these questions, the complex network model with partial communication channels failure is established first. Then, based on the finite-time stability theory and state layered method, two finite-time synchronization criteria for complex networks are provided. Finally, the effectiveness of obtained results are verified by numerical simulations. [ABSTRACT FROM AUTHOR]

Multi-agent reinforcement learning (MARL) is a promising algorithm for traffic signal control (TSC), and graph neural networks make a further improvement on its learning capacity. However, these state-of-the-art algorithms adopt fine-granularity information (i.e., current step-states) to get traffic-network embeddings, but ignore great-granularity information (e.g., previous multiple step-states); In addition, these algorithms are optimized by loss functions in the viewpoint of MARL, but ignore the correlation between input information and output embeddings. This paper proposes a Hierarchical Graph Multi-agent Mutual Information (HG-M2I) algorithm for TSC. Specifically, HG-M2I algorithm is developed based on: 1) Multi-granularity fusion employs a proposed Hierarchical Graph representation learning (HG) algorithm, which fuses both current step-states of multiple agents and previous multiple step-states of each agent , thus facilitates to produce optimal final embeddings; 2) Joint optimization employs a proposed Multi-agent Mutual Information (M2I) framework, which measures the correlation between input step-states and output embeddings by maximizing mutual information, the corresponding mutual-information and MARL loss jointly optimize the whole algorithm, thus facilitates to derive optimal traffic-signal policies. HG-M2I is compared against these state-of-the-art algorithms on synthetic-road and real-road datasets. Experimental results not only show the superior performance of HG-M2I over other compared algorithms, but also illustrate its advantageous transferability. [ABSTRACT FROM AUTHOR]

This paper proposes a resources-aware pattern of finite-time controller design for discrete-time Markov jump systems (DTMJSs) Firstly, the performance evaluation of DTMJSs during a given interval is performed using a self-triggered mechanism that decides the subsequent execution only on the basis of historical data. Also, an adaptive self-triggered mechanism with customizable thresholds is proposed for more communication resource conservation. To examine the impact of the association between mode transitions and triggering signals on finite-time performance, two instances are explored regardless of whether it activated within the working period. Additionally, a co-design framework that calculates the adaptive self-triggered parameters and the finite-time controller gains synchronously is developed, which increases the efficiency of the resource conservation as well as facilitates its adaption. [ABSTRACT FROM AUTHOR]

• Proposing a new deep learning model for accurately detecting cyber threats to extract a feature representation and learn temporal information from cybersecurity data sequences. • Decreasing the learning rate of the Adam optimizer exponentially to improve its exploitation operator when it cannot improve in the validation loss within several epochs. • Evaluating the proposed model's performance, DeepAK-IoT uses three datasets, two of which are aggregated using heterogeneous data sources, to assess its ability to detect the malicious code that can threaten IoT sources. • Refining DeepAK-IoT provides greater accuracy for the validated datasets, superior to three contemporary models. Our daily lives have been profoundly changed over the past few years owing to the growing presence of the Internet of Things (IoT). Importantly, IoT makes our lives more convenient, simpler, and more efficient; however, gadgets are vulnerable to a wide variety of cyberattacks due to the lack of robust security mechanisms and hardware security support. This paper presents an alternative deep learning model known as DeepAK-IoT to detect cyberattacks against IoT devices. DeepAK-IoT uses three blocks as its foundation: the residual-based-spatial representation (RSR) block, the temporal representation block (TRB), and the detection block (DB). The RSR block uses five residual blocks to extract a feature representation from the output of the preceding layer. The four convolutional layers are connected in parallel with a skip connection within each block to avoid vanishing or exploding gradients. Then, the second block uses the extracted spatial representation to learn a temporal representation to detect cyber threats. The final block decides how to classify the input record. We evaluated the accuracy and generalization ability of DeepAK-IoT using three well-known public datasets: TON-IoT, Edge-IIoTset, and UNSW-NB15. The proposed model was compared to three state-of-the-art deep learning models to demonstrate its effectiveness in detecting cyber threats in IoT systems. According to the experimental results, DeepAK-IoT was found to be a powerful alternative model for managing cyber threats in IoT networks, as it provided 90.57% accuracy for TON IoT, 94.96% for Edge-IIoTset, and 98.41% for UNSW NB15. [ABSTRACT FROM AUTHOR]

The training effect of the neural network model is directly influenced by optimization algorithms. The Barzilai-Borwein(BB) method is used in the stochastic gradient descent (SGD) and other deep learning optimization algorithms because of its outstanding performance in terms of convergence speed. In order to improve the stability of BB step size and the output quality of network model in deep learning, this paper presents two improved optimization algorithms based on the BB method: BBbound and AdaBBbound. BBbound reduces the floating BB step size by generating the upper bound of the step in the current iteration. It avoids the occasional occurrence of long steps; AdaBBbound is an adaptive gradient method based on BB method with dynamic bound. The algorithm has a fast convergence effect in the early stage, then calculates learning rate and smoothly transitions to SGD by setting the better conditions of the BB method. We analyzed the performance of the two algorithms based on the initial conditions and the learning rate change curve. Meanwhile, we tested our algorithms on popular network models such as ResNet and DenseNet. The results showed that the new optimization algorithms have achieved high stability and significantly improved the performance of the network model. [ABSTRACT FROM AUTHOR]

To further enhance the reliability of Machine Learning (ML) systems, considerable efforts have been dedicated to developing privacy protection techniques. Recently, membership privacy has gained increasing attention, with a focus on determining whether a specific data point is present in the confidential training set of an ML model. However, most current attacks only prioritize attack accuracy and fail to extend their range to the evidence that contributes to the member/non-member classification. This limitation greatly reduces the practicality of Membership Inference Attacks (MIA), as real-world data typically includes multiple features, making it challenging to identify which features are involved in the sensitive training set. Therefore, this paper targets one of the fundamental challenges in membership inference attack: measuring the distance between an attack sample and a member sample. Specifically, we propose a novel threat model called Membership Reconstruction Attack (MRA), which aims to reconstruct the exact distribution of the target training set. MRA achieves this by marking each input dimension (e.g., pixels) according to its similarity to the target dataset in feature space. Our attack demonstrates its effectiveness across various settings, including different major datasets (MNIST, CIFAR-10, CIFAR-100) and different model architectures (AlexNet, ResNet, DenseNet, and generative models). Additionally, we evaluate MRA from the defenders' perspective and test several defense approaches against our attack. [ABSTRACT FROM AUTHOR]

• Introduce the role of phrases in text representation and semantic disambiguation. • Propose ParsingPhrase to extracts combination phrases and improves phrase quality. • Mining candidate phrases efficiently by traversing the parse tree hierarchically. • Put forward a new phrase evaluation indicator to enhance phrase evaluation accuracy. • We suggest a joint scheme to solve sentence disambiguation and phrase optimization. Phrases represent independent semantics in natural language but usually have indeterminate lengths and different combinations. So, extracting meaningful phrases from unstructured texts will substantially reduce semantic ambiguity and lay the foundation for downstream natural language tasks. Most existing research obtains candidate phrases by N-grams, which includes meaningless word sequences and degrades algorithm performance. In this paper, we propose a novel phrase-mining algorithm, called ParsingPhrase, which effectively extracts combination phrases from text and improves phrase quality by syntactic features. It consists of three stages. Firstly, all sentences in texts are represented as parsing trees by PCFG (Probabilistic Context-Free Grammar). We propose PBMP (Parsing-Based Phrase mining) to obtain candidate phrases from those parsing trees. Then, we introduce a new phrase evaluation indicator, called Significance, that relies on the role of phrases to measure their importance. We integrate the Significance with conventional evaluation indexes for a more reasonable phrase evaluation. Finally, we optimize the phrase quality again by exploiting the optimal phrase composition features of sentences. To the best of our knowledge, it is the first work to employ parsing for combination phrase mining and evaluation, meanwhile offering a solution for syntactic disambiguation. Experiments on three real corpora demonstrate that the ParsingPhrase exceeds state-of-the-art baselines, is 7% higher in the candidate phrase conversion rate, and is 6% better in terms of Precision. [ABSTRACT FROM AUTHOR]

In this paper, we improve the robustness of Machine Learning (ML) classifiers against training-time attacks by linking the risk of training data being tampered with to the redundancy in the ML model's design needed to prevent it. Our defense mechanism is directly applicable to classifiers' training data, without any knowledge of the specific ML model to be hardened. First, we compute the training data proximity to class separation surfaces, identified via a reference linear model. Each data point is associated with a risk index, which is used to partition the training set by an unsupervised technique. Then, we train a learner for each partition and combine the learners' output in an ensemble. Our method treats the protected ML classifier as a black box and is inherently robust to transfer attacks. Experiments show that, for data poisoning rates between 6 and 25 percent of the training set, our method is more robust compared to benchmarks and to a monolithic version of the model trained on the whole training set. Our results make a convincing case for adopting training set partitioning and ensemble generation as a stage of ML models' development and deployment lifecycle. [ABSTRACT FROM AUTHOR]

This issue of DeepFake technique that may cause great threat to privacy, democracy, and national security has attracted the attention of deep learning researchers. DeepFake detection, therefore, has been a very hot issue in deep learning research. The face landmark feature maps are often used by many DeepFake approaches in generating fake faces. It also provides key information to help to detect manipulated face images. In this paper, we propose a detection approach for manipulated face images. To make full use of face landmark information, we propose a facial geometry prior module (FGPM) to extract facial geometry feature maps. Then the facial geometry feature maps are embedded into upsampling feature maps generated by the CNN-LSTM network. The proposed FGPM exploits facial maps and frequency domain correlation to analyze the discriminative characteristics between manipulated and non-manipulated regions by incorporating the CNN-LSTM network. Finally, a decoder is used to learn the mapping from low-resolution feature maps to pixel-wise to predict manipulation localization. Or a softmax classifier is used to predict real or fake face images. By experimenting on several popular datasets, the proposed detection model has demonstrated the capability of localizing manipulation at the pixel level and with a high prediction on real or fake face images. • Propose a manipulated face detection model which predicts fake face in pix-lever. • Facial geometry feature maps are used in predicting fake face. • Propose a CNN-LSTM network to discriminate manipulated face images. [ABSTRACT FROM AUTHOR]

As a robust online learning approach, adaptive filtering based on correntropy has attracted more attentions. Recently, researchers have made efforts to develop complex correntropy to deal with adaptive filtering issues in complex domain. However, most of the literature on complex correntropy lack details regarding the processing of noncircular signals. Up to now, the maximum improper complex correntropy criterion (MICCC) algorithm for noncircular and widely-linear (WL) adaptive filtering utilizes the gradient descent method to find the maximum of the objective function. However, the MICCC algorithm fails to guarantee both low steady-state weight error power and fast convergence. To overcome this defect, this paper proposes a class of adaptive filtering algorithms which are based on fixed-point instead of gradient descent. It is worth noting that we manage to avoid the massive calculation of matrix inversion in traditional fixed-point based methods by utilizing the lemma of matrix inversion properly and further provide two alternative recursive algorithms. Finally, the local convergence analysis of mean and mean square is performed, and the simulated results offer strong evidences for the superior performance of the proposed algorithms over widely-linear complex least mean square (WL-CLMS), widely-linear recursive least square (WL-RLS) and MICCC. [ABSTRACT FROM AUTHOR]

This paper addresses the problem of secure event-triggered control is to guarantee the stability and security of a general class of industrial cyber-physical systems (CPSs) under limited resource budget and deception attacks. Note that the coupled data and the complicated structure of the CPS make it hard for the traditional control algorithms to fulfill the stability and security requirements. To this end, a neural network learning-based approximation algorithm is first proposed to separate the coupling influence, and then estimate the lumped uncertain nonlinearities. This is done by using the separated data as the input of the neural networks. Second, Nussbaum-type functions are presented to settle the unknown sign of the time-varying attack injection signal. Third, an event triggering mechanism is developed to reduce the communication load. Under the developed secure control laws, all the signals of the resulted closed-loop CPS are bounded. Finally, to validate the effectiveness of the proposed secure event-triggered control method, a benchmark control example for the one-link manipulator actuated by a DC motor is exploited. • Secure event-triggered control is proposed to ensure the security of industrial CPSs. • Neural network learning algorithm is used to estimate the uncertain nonlinearities. • Nussbaum-type functions are presented to handle the unknown sign of the time-varying attack injection signal. • Event triggering mechanism is developed to reduce the communication. [ABSTRACT FROM AUTHOR]

Spectral clustering aims to minimise inter-cluster similarity by constructing graph model, which possesses a significant effect in data of arbitrary shape. Nonetheless, there are still two limitations in the existing algorithms. First, spectral methods perform poorly if the densities of instances vary greatly. Second, they are challenging to handle the complex distribution that different objects of densities are mixed. To address the two limitations, in this paper, a novel spectral clustering algorithm is constructed to accommodate more complex multi-density data. The idea of density stratification is proposed through the probability density of Gaussian mixture model, and the number of layers can be automatically determined by the defined scatter index. Then, a new density ratio is established to effectively reduce the density imbalance of diverse instances based on mutual nearest-neighbour and density stratification, which is more advantageous in the clusters with less clear boundaries. Finally, the optimised adjacency matrix and multi-density spectral clustering algorithm are induced to improve the effect of multi-density data. As per experimental results, the proposed algorithm generally outperforms the popular representative algorithms for both synthetic and benchmark datasets, and works more effectively on multi-density as well as single-density data, which illustrates the superiority of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Recently, 3D deep learning models have been shown to be susceptible to adversarial attacks like their 2D counterparts. Most of the state-of-the-art (SOTA) 3D adversarial attacks perform perturbation to 3D point clouds. To reproduce these attacks in the physical scenario, a generated adversarial 3D point cloud needs to be reconstructed to mesh, which leads to a significant drop in its adversarial effect. In this paper, we propose a strong 3D adversarial attack named Mesh Attack to address this problem by directly performing perturbation on mesh of a 3D object. In order to take advantage of the most effective gradient-based attack, a differentiable sample module that back-propagate the gradient of point cloud to mesh is introduced. To further ensure the adversarial mesh examples without outlier and 3D printable, three mesh losses are adopted. Extensive experiments demonstrate that the proposed scheme outperforms SOTA 3D attacks by a significant margin. We also achieved SOTA performance under various defenses. Our code will be available at: https://github.com/cuge1995/Mesh-Attack. [ABSTRACT FROM AUTHOR]

In online social networks, trust prediction is crucial for applications such as user decision-making and high-impact user recommendation. In recent years, various solutions to the trust prediction problem have been proposed in many academic literatures. However, the existing methods only focus on using more available information to achieve better prediction performance, while ignoring the influence of the authenticity of user information and the user subjectivity. In this paper, we apply the sentiment labels to the trust prediction process for the first time, and propose a deep trust prediction method based on sentiment polarity perception, S-DeepTrust. We first use transfer learning to generate sentiment labels for user review data and obtain a sentiment polarity matrix. Next, we design a rating matrix generation algorithm, which weights the sentiment polarity matrix and the original rating matrix to produce a new rating matrix. Then, the Siamese network architecture is used for the first time in the field of trust prediction to judge the trust relationships based on the extracted user preference vectors, which minimizes the representation of trust user pairs. Finally, the experiment results show that our method achieves higher prediction accuracy and is robust to trust relation sparsity compared with existing methods. [ABSTRACT FROM AUTHOR]

This paper proposes an approach for the analysis and comparison of stochastic solvers based on the statistical distribution of their variables. The observed variables of the stochastic solvers are the runtime and number of function evaluations required to reach a (sub)-optimal solution. These variables were measured using a target approach. We extended the conventional approach, which usually predicts the average value of variables, by predicting their statistical distributions. If possible, we can predict, not only the average value, but also all the values of the observed variables according to the identified distribution and given probability. We can also predict the probability of reaching (sub)-optimal solutions according to the variable's values. The approach was empirically validated by comparing solvers for discrete and continuous problems. In our experiment, the differences between predicted and measured values of runtime, number of function evaluations and probability are 10% or less. This proves that the proposed approach is useful and can be used for the analysis and comparison of stochastic solvers. Although, the approach has some limitations, it can mitigate the issue of stochastic solvers being unable to provide (sub)-optimal solutions, and can be used to determine the stopping optimization criteria. • An approach to the comparison of stochastic solvers. • The predictive modeling based on optimal and sub optimal solutions. • The analysis of the distribution of runtime and number of function evaluations. • The predictive models based on distribution of solvers' variables are established. • The predictive models provide probability of reaching the optimal solutions. [ABSTRACT FROM AUTHOR]