Search

This paper proposes an innovative image encryption algorithm that leverages partial face recognition and DNA diffusion, building upon advancements in chaotic image encryption and face recognition technologies. The key is generated using the secure SHA-512 algorithm, while the chaotic sequence for the encryption algorithm is derived from the Chen system. This paper begins by employing facial recognition technology to identify and focus on specific facial regions within an image. Subsequently, DNA diffusion is applied to encrypt the recognized face segment via a diffusion-scrambling encryption process. Following this, the entire image undergoes diffusion utilizing the devised diffusion rules. Finally, a uniform scrambling operation based on Fisher-Yeats is performed on the entire image. The experimental results and safety analysis show that the information entropy can exceed 7.9973, the average correlation can be less than 0.0023, the pixel number change rate can reach 99.6052%, and the unified average change intensity can reach 33.4644%, all of which are very close to the ideal values. [ABSTRACT FROM AUTHOR]

In this work, an improved moth-flame optimization algorithm is proposed to alleviate the problems of premature convergence and convergence to local minima. From the perspective of diversity, an inertia weight of diversity feedback control is introduced in the moth-flame optimization to balance the algorithm's exploitation and global search abilities. Furthermore, a small probability mutation after the position update stage is added to improve the optimization performance. The performance of the proposed algorithm is extensively evaluated on a suite of CEC'2014 series benchmark functions and four constrained engineering optimization problems. The results of the proposed algorithm are compared with the ones of other improved algorithms presented in literatures. It is observed that the proposed method has a superior performance to improve the convergence ability of the algorithm. In addition, the proposed algorithm assists in escaping the local minima. [ABSTRACT FROM AUTHOR]

Privacy preserving utility mining (PPUM) aims to solve the problem of sensitive information leakage in utility pattern mining. In recent years, researchers have proposed algorithms to solve the privacy-preserving problem. However, these algorithms have high side effects, long sanitization time, and computational complexity. Although the FPUTT algorithm reduces the number of database scans, tree construction and traversal still take much time. The paper proposes a fast utility-list dictionary algorithm (FULD). The utility-list dictionary consists of all sensitive items. Through dictionary lookup, sensitive items can be found and modified. In addition, the novel concepts of SINS and tns are proposed to reduce the side effects of the algorithm. In this paper, the experiments show that the FULD algorithm has good performance, such as running time and side effects. The running time of the FULD is 15–20 times shorter than the FPUTT algorithm. It performs well both on sparse and dense datasets. [ABSTRACT FROM AUTHOR]

Identifying discriminative patterns, known as shapelets, within time series is a critical step in many time series classification tasks. A major limitation of shapelet learning is that often hindered by their unsupervised methods, treating shapelet learning as an unsupervised subsequence clustering process and discovery based on pre-defined metric, which performed sequentially. This sequential procedure presents challenges, as it fails to establish a direct connection between shapelets and samples, and lacks the capacity to explicitly incorporate label information. In this paper, we proposed a novel shapelet learning algorithm called Discriminative Shapelet Learning via Temporal Clustering and Matrix Factorization (DSLMF). DSLMF introduced a joint framework that combines matrix factorization and coherent temporal clustering to discovery salient and coherent feature subsets. To further enhance discriminability and prevent arbitrary shapelet shapes, DSLMF integrates a label-specific shapelet regularization as a guiding mechanism enabling the learning of shapelets optimized for higher classification performance. The proposed algorithm has shown to be effective for capturing the temporal cluster structure and interpretability of shapelet-based method. The results of experiments showcased in this paper highlight DSLMF's effectiveness in capturing temporal cluster structures and learning meaningful shapelets, ultimately leading to promising performance on benchmark datasets. [ABSTRACT FROM AUTHOR]

Deep Cross-Modal Hashing (DCMH) has garnered significant attention in the field of cross-modal retrieval due to its advantages such as high computational efficiency and small storage space. However, existing DCMH methods still face certain limitations: (1) they neglect the correlation between labels, while label features exhibit high sparsity; (2) they lack fine-grained semantic alignment; (3) they fail to effectively address data imbalance. In order to tackle these issues, this paper introduces a framework named Semantic-Alignment Transformer and Adversary Hashing for Cross-modal Retrieval (SATAH). To the best of our knowledge, this is the first attempt at the Semantic-Alignment Transformer algorithm. Specifically, this paper first designs a label learning network that utilizes a crafted transformer module to extract label information, guiding adversarial learning and hash function learning accordingly. Subsequently, a Balanced Conditional Generative Adversarial Network (BCGAN) is constructed, marking the first instance of adversarial training guided by label information. Furthermore, a Weighted Semi-Hard Cosine Triplet Constraint is proposed to better ensure high-ranking similarity relationships among all items. Lastly, considering the correlation between labels, a semantic-alignment constraint is introduced to handle label correlation from a fine-grained perspective, capturing similarity on a global scale more effectively. Extensive experiments are conducted on multiple representative cross-modal datasets. In experiments with 64-bit hash code length, SATAH achieves average mAP values of 84.75%, 68.87%, and 68.73% on MIR Flickr, NUS-WIDE, and MS COCO datasets, respectively, outperforming state-of-the-art methods. The code is available at https://github.com/Daydaylight/SATAH. [ABSTRACT FROM AUTHOR]

This paper introduces an algorithm to construct a bidirectional causal graph using an augmented graph. The algorithm decomposes the augmented graph, significantly reducing the size of the variable set required for conditional independence testing. Simultaneously, it preserves the fundamental structure of the augmented graph after decomposition, saving time and cost in constructing a global skeleton graph. Through experiments on discrete and continuous datasets, the algorithm demonstrates a clear advantage in runtime compared to traditional methods. In large-scale sparse networks, the training time is only about one-tenth of traditional methods. Additionally, the algorithm shows improvement in terms of construction error. Since the input to the algorithm is an augmented graph, this paper also discusses the impact on construction error when using both real and generated augmented graphs as input. Furthermore, the concept of markov blanket is extended to multivariate regression chain graphs, providing a method for rapidly constructing augmented graphs given certain prior knowledge. [ABSTRACT FROM AUTHOR]

Aiming at the intrinsic aspect that the weak features of the early fault information of rolling bearings are not easy to extract, a parameter Adaptive Variational Modal Decomposition (AVMD) based algorithm is proposed for bearing fault signal feature extraction. Since the number of Variational Modal Decomposition (VMD) decomposition and penalty factor play an important role in VMD decomposition effect, the irregularities in the selection of these two influencing parameters are analyzed. We exploit the stronger global search capability of the improved sparrow search algorithm (LSSA) for adaptive parameter selection of the VMD algorithm. In this paper, the Levy flight algorithm is introduced and chaos is added to initialize sparrow population position to prevent sparrow from falling into the disadvantage of local optimum in the search process. In addition, this paper also combines the maximum kurtosis index, the minimum envelope entropy index and the number of iterations of VMD to form the objective function of the LSSA. The VMD algorithm with optimized parameters decomposes the signal to be measured, the decomposed IMFs was reconstructed, finally the validity of the model was verified by calculating 20 features (time domain and frequency domain) of the reconstructed signal as the input vector of the SVM classifier. Finally, the feasibility of this model in fault diagnosis of rolling bearing is verified using simulation and example experiments. [ABSTRACT FROM AUTHOR]

Deep reinforcement learning (DRL) has made significant strides in addressing tasks with high-dimensional continuous action spaces. However, the field still faces the challenges of low sample utilization and insufficient exploration-exploitation balance, limiting the generalizability of algorithms across different environments. To effectively improve sample utilization, optimize the exploration-exploitation balance, and achieve higher rewards in tasks, this paper designs the selective priority timing entropy (SPTE) algorithm. Subsequently, selective prioritized experience replay (SPER) is proposed, which employs frequent replay of multiframe memories to enhance sample utilization and improve the stability of policy updates. Additionally, the temporal advantage with decay (TAD) method introduces a decay factor to help adjust the weights of the variance and bias, thereby reducing estimation errors. The reward mechanism is augmented with multientropy (ME) for entropy-regularized training, achieving a balance between information exploration and exploitation. Finally, experimental testing on the challenging Arcade platform demonstrated that the SPTE algorithm surpasses the average testing level of human players by 104.936%. Furthermore, compared to other algorithms, SPTE achieves an average score increase of over 32.75%, and it consistently outperforms the compared methods in more than 60% of tasks, indicating its strong adaptability and robustness. The implementation process of SPTE [ABSTRACT FROM AUTHOR]

Experience replay (ER) used in (deep) reinforcement learning is considered to be applicable only to off-policy algorithms. However, there have been some cases in which ER has been applied for on-policy algorithms, suggesting that off-policyness might be a sufficient condition for applying ER. This paper reconsiders more strict "experience replayable conditions" (ERC) and proposes the way of modifying the existing algorithms to satisfy ERC. In light of this, it is postulated that the instability of policy improvements represents a pivotal factor in ERC. The instability factors are revealed from the viewpoint of metric learning as i) repulsive forces from negative samples and ii) replays of inappropriate experiences. Accordingly, the corresponding stabilization tricks are derived. As a result, it is confirmed through numerical simulations that the proposed stabilization tricks make ER applicable to an advantage actor-critic, an on-policy algorithm. Moreover, its learning performance is comparable to that of a soft actor-critic, a state-of-the-art off-policy algorithm. [ABSTRACT FROM AUTHOR]

Co-training is a semi-supervised algorithm that aims to improve prediction effects by exchanging confident instances and pseudo-labels among multiple learners. One central issue is how to mitigate the negative impact of low-quality pseudo-labels during training. In this paper, we propose semi-supervised regression via embedding space mapping and pseudo-label smearing (S2RMS) to ensure that unlabeled data contribute positively to the prediction process. First, a Triplet neural network is trained using pairwise data generated from labeled data. This network maps the training data to the embedding space to better separate dissimilar instances. Second, the embedded data are randomly partitioned into different subsets to train corresponding regression models (a.k.a. regressors). These regressors are integrated into the prediction process. Third, these subsets are augmented using unlabeled data with high similarity to the labeled data and high-confidence pseudo-labels. Here, the similarity and confidence are calculated using the network and the smearing technique, respectively. Experiments are conducted on fourteen datasets, and the results are compared to those of three excellent algorithms. The results show that S2RMS outperforms other co-training and metric semi-supervised regression algorithms. The source code is available at: https://github.com/BetaCatPro/S2RMS. [ABSTRACT FROM AUTHOR]

Label distribution learning, where deal with label ambiguity by describing the degree of relevance of each label to a specific instance. As a novel machine learning paradigm, the curse of dimensionality is one of the prominent problems. Feature selection is a vital preprocessing step to reduce the high dimensionality of data. However, most existing label distribution feature selection methods focus on selecting a feature subset that has relevant capabilities for all labels, ignoring label-specific features with the maximum discriminatory power for each label. To tackle this issue, a label distribution feature selection algorithm based on label-specific features is proposed in this paper. Initially, we introduce a feature selection optimization model for label distribution data that simultaneously considers common and label-specific features, leveraging sparse learning to further investigate the intricate relationships between features and labels. Subsequently, the label correlation coefficient is employed to enhance the collaborative learning effect of labels. Finally, the relevance between features and labels is taken into account to guide the feature selection process, which can effectively eliminate the redundant features. Comprehensive experiments demonstrate the advantage of our proposed method over other well-established feature selection algorithms for selecting label-specific features to label distribution data. [ABSTRACT FROM AUTHOR]

The design of effective multi-view clustering algorithms has recently garnered significant research attention. In this paper, we develop a robust multi-view clustering via collaborative constraints and multi-layer concept factorization (RMCCMCF) model to enhance clustering power and robustness by mining low-rank data and hierarchical information from multiple views. This approach adopts low rank (LR) and deep learning (DL) to explore the essential and hierarchical structures hidden in various views. The proposed approach applies multiple graph regularization constraints to extract manifold information from different views. Kernel technology is used for concept factorization in high-dimensional space to effectively distinguish different classes of sample points. Additionally, we design an alternating iterative optimization approach to solve the RMCCMCF algorithm and discuss its convergence. The results of extensive experiments conducted on the CBSR, HW2Sources, 3Sources, CUB, and UCI datasets demonstrate that the RMCCMCF approach outperforms several other recent multi-view clustering approaches in terms of recognition performance. [ABSTRACT FROM AUTHOR]

The k-means algorithm is a prevalent clustering method due to its simplicity, effectiveness, and speed. However, its main disadvantage is its high sensitivity to the initial positions of the cluster centers. The global k-means is a deterministic algorithm proposed to tackle the random initialization problem of k-means but its well-known that requires high computational cost. It partitions the data to K clusters by solving all k-means sub-problems incrementally for all k = 1 , ... , K . For each k cluster problem, the method executes the k-means algorithm N times, where N is the number of datapoints. In this paper, we propose the globalk-means++ clustering algorithm, which is an effective way of acquiring quality clustering solutions akin to those of global k-means with a reduced computational load. This is achieved by exploiting the center selection probability that is effectively used in the k-means++ algorithm. The proposed method has been tested and compared in various benchmark datasets yielding very satisfactory results in terms of clustering quality and execution speed. [ABSTRACT FROM AUTHOR]

In unsupervised domain adaptation (UDA), knowledge is transferred from label-rich source domains to relevant but unlabeled target domains. Current most popular state-of-the-art works suggest that performing domain alignment from the class perspective can alleviate domain shift. However, most of them based on domain adversarial which is hard to train and converge. In this paper, we propose a novel contrastive learning to improve diversity and discriminability for domain adaptation, dubbed as IDD_ICL, which improve the discriminativeness of the model while increasing the sample diversity. To be precise, we first design a novel implicits contrastive learning loss at sample-level by implicit augment sample of the source domain. While augmenting the diversity of the source domain, we can cluster the samples of the same category in the source domain together, and disperse the samples of different categories, thereby improving the discriminative ability of the model. Furthermore, we show that our algorithm is effective by implicitly learning an infinite number of similar samples. Our results demonstrate that our method doesn't require complex technologies or specialized equipment, making it readily adoptable and applicable in practical scenarios. [ABSTRACT FROM AUTHOR]

Inferring multilayer diffusion networks from observed cascades is both crucial and realistic. To infer multilayer diffusion networks, constructing continuous-time diffusion models that capture diffusion dynamics is a prerequisite. However, developing such models faces two main challenges: (1) reducing the number of learnable parameters for precise optimization with limited cascades while effectively modeling for accurate inference and (2) adapting the models to more realistic scenarios. In this paper, we propose a novel continuous-time diffusion model, namely the Embedding-based Continuous-time Diffusion (ECD) model, which employs an embedding method while modeling symmetric relationship strength, asymmetric relationship strength, and trust strength. Specifically, by leveraging the embedding method, the number of learnable parameters is significantly reduced compared with previous models. Then, by modeling symmetric relationship strength, our model can be used in scenarios where the relationships between nodes are symmetric. Subsequently, the trust strength can be inferred by our proposed efficient heuristic algorithm, making our model suitable for scenarios where time information is unavailable. Furthermore, we develop an optimization algorithm to optimize the proposed model and infer multilayer diffusion networks. The experimental results on synthetic and real datasets show that our model and algorithms outperform the comparison methods. [ABSTRACT FROM AUTHOR]

Unmanned Aircraft Systems (UASs) offer many benefits in wildfire monitoring when compared to traditional wildfire monitoring technologies. Wildfire spread is a highly heterogeneous process with non-uniform spreading speed and fireline intensity in both space and time. When planning the path of a UAS for monitoring a dynamically growing wildfire, it is important to consider the uneven spread of the fire because different parts of the fire boundary would need different levels of monitoring attention. This paper presents a real-time path planning algorithm for a UAS to autonomously monitor the perimeter of a spreading wildfire with uneven importance. The proposed path planning algorithm allows the UAS to focus on the most active regions of a wildfire while still covering the entire fire perimeter. The design and implementation of this algorithm are described, and an in-depth analysis of the performance of the algorithm is provided. Experiment results based on simulated wildfires demonstrate the effectiveness and robustness of the proposed algorithm for monitoring dynamic wildfires. [ABSTRACT FROM AUTHOR]

Discovering new relationships between entities in networked data is essential in various applications such as sociology, security, physics, and biology. This paper introduces a novel approach to directed link prediction, filling a notable research gap by acknowledging the importance of the directionality of relationships often overlooked in traditional methods. We present three algorithms: an asymmetric similarity-popularity algorithm, which applies the similarity-popularity paradigm specifically to directed networks; a path exploration algorithm, which utilizes path patterns, closure probabilities, and paths' exploratory potential to predict new links formation; and a hybrid algorithm that merges the strengths of both approaches. The effectiveness of these methods is rigorously evaluated on real-life networks, demonstrating their robust performance across various types and sizes of networked data. In addition to predictive power and runtime performance assessments, we study the impact of predicted links on network spreading capacity. This perspective provides invaluable insights into the broader implications of our algorithms on network behavior and dynamics. [ABSTRACT FROM AUTHOR]

Maximizing the advantages of different views and mitigating their respective disadvantages in fine-grained segmentation tasks are an important challenge in the field of point cloud multi-view fusion. Traditional multi-view fusion methods ignore two fatal problems: 1. the loss of depth and quantization information due to mapping and voxelization operations, resulting in "anomalies" in the extracted features; 2. how to pay attention to the large differences in object sizes among different views during point cloud learning, and fine-tune the fusion efficiency in order to improve the performance of network. In this paper, we propose a new algorithm that uses channel self-attention to fuse range-point-voxel, abbreviated as RPV-CASNet. RPV-CASNet integrates the three different views: range, point and voxel in a more subtle way through an interactive structure (range-point-voxel cross-adaptive layer known as RPVLayer for short), to take full advantage of the differences among them. The RPVLayer contains two key designs: the Feature Refinement Module (FRM) and the Multi-Fine-Grained Feature Self-Attention Module(MFGFSAM). Specifically, the FRM allows for a re-inference representation of points with entrained anomalous features, correcting the features. The MFGFSAM addresses two challenges: efficiently aggregating tokens from distant regions and preserving multiscale features within a single attention layer. In addition, we design a Dynamic Feature Pyramid Extractor (DFPE) for network deployment, which is used to extract rich features from spherical range images. Our method achieves impressive mIoU scores of 69.8% and 77.1% on the SemanticKITTI and nuScenes datasets, respectively. [ABSTRACT FROM AUTHOR]

Because of the open-collaborative of online encyclopedias, a large number of knowledge triples are improperly classified in online encyclopedia systems, and it is necessary to denoise and refine the open-domain encyclopedia Knowledge Bases (KBs) to improve the quality and precision. However, the lack and inaccuracy of triple semantic features lead to a poor refinement effect. In addition, considering large-scale encyclopedia KBs, the processing of massive knowledge will lead to too much computing time and poor scalability of the algorithm. To solve the problems of knowledge denoising in the Chinese encyclopedia system, first, based on data field theory, this paper proposes a new Cartesian product mapping-based method (TripleES) to calculate the semantic similarity of entity triples, based on which a method for quantifying the quality of entry tags is proposed. Second, to further improve the denoising effect on KBs, this paper proposes a new method (TriplePV) to compute the potential value of triple based on multi-feature fusion strategy to calculate the semantic distance between the "out-of-vocabulary" entry tags and embeds it into the potential function. Third, to ensure our algorithms have good scalability, the proposed denoising algorithms are implemented and optimized in parallel based on the Spark cluster-computing framework. Specifically, Spark-based TripleES (ES_Spark) and Spark-based TriplePV (PV_Spark) algorithms are proposed to calculate the semantic similarity and potential value of triples respectively. Finally, a comprehensive comparative analysis is performed on the denoising effect and time efficiency with the state-of-the-art distributed Chinese encyclopedia knowledge denoising algorithm. The experimental results on real-world datasets show that the parallel denoising algorithm proposed in this paper can improve the efficiency of knowledge denoising and the accuracy of KBs, which outperforms the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In mobile crowdsensing, the sensing platform recruits users to complete large-scale sensing tasks cooperatively. In order to guarantee the quality of sensing tasks, the platform needs to recommend suitable tasks to users. Existing task recommendation methods typically focus on unilateral factors, such as user preferences or task quality, leading to low platform utility and task acceptance rate respectively. To solve the above issue, this paper proposes a task recommendation method which takes both user preferences and user-task matching into consideration. Firstly, we apply the Deep Interest Network (DIN) in the context of mobile crowdsensing to recommend tasks according to user preferences. Secondly, the concept of user-task matching is introduced, in which both the task difficulty and the user reliability are taken into account. Finally, we propose task recommendation algorithms and conduct extensive experiments on a real dataset. The experimental results show that the proposed method can not only improve the utility of the platform significantly, but also improve the recommendation accuracy slightly under longer recommendation list. [ABSTRACT FROM AUTHOR]

Event sequences are common types of data. Several episode mining algorithms have been developed to find episodes (subsequences of events) that appear frequently in an event sequence, with the aim of discovering useful knowledge for decision-making and predictions. However, most of these algorithms can only process simple event sequences (where, at most, one event occurs at each timestamp). In contrast, in many real-life applications, multiple events may occur at the same timestamp, resulting in complex event sequences. Moreover, numerous episode mining algorithms overestimate the frequency of episodes by counting the same events multiple times. As a solution, some algorithms have been designed to count only non-overlapping occurrences. Yet, it can be argued that this definition is too strict and discards many important events. To address these limitations, this paper presents an algorithm named EMDO (Episode Mining under Distinct Occurrences (EMDO) to find frequent episodes in a complex sequence by counting distinct occurrences. The proposed concept of distinct occurrences ensures that each event is not counted more than once but allows distinct occurrences to overlap. A second algorithm, called EMDO-P, is also presented in this paper to derive strong episode rules in event sequences from episodes found by EMDO. To the best of our knowledge, this is the first study on mining frequent episodes using a frequency definition based on distinct occurrences. The experimental results confirm that the proposed algorithms are efficient. [ABSTRACT FROM AUTHOR]

Multi-label feature selection serves an effective dimensionality reduction technique in the high-dimensional multi-label data. However, most feature selection methods regard the label as complete. In fact, in real-world applications, labels in a multi-label dataset may be missing due to various difficulties in collecting sufficient labels, which enables some valuable information to be overlooked and leads to an inaccurate prediction in the classification. To address these issues, a feature selection algorithm based on the granular-ball based mutual information is proposed for the multi-label data with missing labels in this paper. At first, to improve the classification ability, a label recovery model is proposed to calculate some labels, which utilizes the correlation between labels, the properties of label specific features and global common features. Secondly, to avoid computing the neighborhood radius, a granular-ball based mutual information metric for evaluating candidate features is proposed, which well fits the data distribution. Finally, the corresponding feature selection algorithm is developed for selecting a subset from the multi-label data with missing labels. Experiments on the different datasets demonstrate that compared with the state-of-the-art algorithms the proposed algorithm considerably improves the classification accuracy. The code is publicly available online at https://github.com/skylark-leo/MLMLFS.git [ABSTRACT FROM AUTHOR]

This paper proposes an integrated approach to group decision-making (GDM) by using stochastic preference analysis (SPA) and best-worst method (BWM). BWM preparation algorithm is proposed to obtain the best and worst of experts and the relative importance degree. Meanwhile, expert weights model and expert's priority vector model are proposed. Furthermore, the stochastic composite rank acceptability index, stochastic composite expected priority vector, stochastic composite expected rank and stochastic composite confidence factor are developed based on SPA to describe the ranks of alternatives based on SPA. Finally, a group stochastic preference analysis-best worst method (GSPA-BWM) algorithm is developed by analyzing the judgments space through Monte Carlo simulation. The experts can use this method to choose some of the outcomes which they find most useful to make reliable decisions. Examples and comparison analyses show that the proposed method is effective. [ABSTRACT FROM AUTHOR]

Surrogate-assisted evolutionary algorithms have been widely employed to solve data-driven optimization problems. However, for offline data-driven optimization, it is very challenging to perform evolutionary search efficiently as well as accurately since no new data is available during the optimization process. To mitigate this issue, a multifidelity surrogates-assisted multi-objective particle swarm optimization (MFSa-PSO) algorithm is proposed in this paper. First, two low-fidelity models with convergence and diversity characteristics separately and a high-fidelity model are constructed to assemble multifidelity surrogate models. Second, by adopting the knowledge transfer strategy, the multifidelity surrogates-assisted two-archive multi-objective particle swarm optimization is conducted to search optimal solutions more exactly and effectively. Third, the output solution set is achieved by associating the solutions of two archives with reference vectors. Finally, the proposed MFSa-PSO is compared with some popular surrogate-assisted evolutionary algorithms on benchmark problems to verify its effectiveness and outperformance. Additionally, a real-world application of the municipal solid waste incineration process is carried out to verify the engineering applicability of MFSa-PSO. [ABSTRACT FROM AUTHOR]

Neighborhood rough set theory is an important computational model in granular computing and has been successfully applied in many areas. One of its most prominent applications is in attribute reduction. However, most current attribute reduction methods for neighborhood rough sets are supervised or semi-supervised, which makes them unable to handle datasets without decision information. To address this, we propose an unsupervised attribute reduction strategy based on neighborhood dependency. First, a neighborhood rough set model based on conditional attribute sets is constructed. Then, based on all individual attribute subsets in the datasets, the importance of the attributes is defined to indicate the significance of the candidate attributes. Furthermore, a neighborhood dependency-based unsupervised attribute reduction (NDUAR) algorithm is designed. Finally, NDUAR is compared with existing algorithms on publicly available datasets. The experimental results show that NDUAR can select fewer attributes to maintain or improve the performance of the clustering algorithm. The effectiveness of the algorithm proposed in this paper is thereby confirmed. [ABSTRACT FROM AUTHOR]

The goal of the influence maximization problem is to identify a seed set that maximizes the number of users who will be impacted. This issue is critical to the spread of information in social networks. The greedy and heuristic algorithms are two prominent algorithms for this problem. The greedy algorithms are usually effective but they have high computational complexity. The heuristic algorithms are tendly efficient while the accuracy is low. The major goal of this study is to strike a compromise between efficacy and efficiency in selecting k influential nodes to maximize influence propagation while keeping the expense of doing so to a minimal. This paper proposes an improved heuristic algorithm, called the Best Hop for Independent Cascade Model (BHICM). This approach presents a dynamic relationship strategy between propagation probability and hop count that avoids processing the seed node's neighbors. Afterwards, the proposed algorithm defines the diffusion score as the criterion for selecting seed nodes, which ensures its accuracy and efficiency. According to experimental results on four genuine social networks, the proposed approach exceeds all comparison algorithms in terms of accuracy while maintaining an acceptable running time. [ABSTRACT FROM AUTHOR]

This paper addresses the challenging problem of developing the automatic algorithm for solving direct current circuit problem. Leveraging on the innovated methods it proposes a high-performance relation based algorithm, called RaDCC. The challenges of the problem lie in relation acquisition and relation inference presentation after adopting the newly-established relation principle of solving problems. A high-performance procedure is developed for the challenging task of relation acquisition by leveraging on three innovated methods. Three methods are an enhanced schematics understanding method that can understand complicated structures of schematics, an extended syntax-semantics model method and a unit-theorem inference method to acquire schematic relations, explicit text relations and implicit text relations, respectively. To address another challenging problem of readable solution generation an action-schema presentation method is proposed to convert relation inference actions into relation inference presentations. The experimental results show that the proposed algorithm is high-performance since it achieves an accuracy of over 83.2% for solving problems from textbooks and 70.6% for solving problems from examination papers on a dataset that contains 1012 direct current circuit problems collected from the authority sources, much higher than the performance of the baseline algorithm. [ABSTRACT FROM AUTHOR]

Neural network architectures have been used to address multiple real-world problems with high success. Their extension to graph-structured data started recently to be explored. Graph-neural network (GNN) achieved state-of-the-art performance in multiple problems. In highly imbalanced application domains, such as network intrusion problems, GNN was used to model the network topology. However, in this scenario, the class imbalance problem still affects the performance. Another graph-based solution, the graph attention network (GAT) has also been applied to multiple predictive tasks. Although being a promising avenue, graph-based solutions are still under-explored in imbalanced scenarios. This paper proposes two novel graph-based algorithms, the E-ResSAGE and E-ResGAT algorithms, which build on top of the established GraphSAGE and GAT algorithms, respectively. The key idea is to integrate residual learning into the GNN leveraging the available graph information. Residual connections are added as a strategy to deal with the high class imbalance, aiming at retaining the original information and improving the minority classes' performance. A case study on intrusion detection is provided. Extensive experiments on four recent intrusion detection datasets show the excellent performance of our proposed approaches, especially when predicting minority classes. We demonstrate that embedding residuals in graph-based algorithms presents a strong advantage when learning under imbalanced domains. [ABSTRACT FROM AUTHOR]

Identifying all frequent high utility occupancy itemsets (FHUOIs) in a quantitative transaction dataset is a new trend in data mining. By combining both factors of frequency and utility occupancy, these patterns are more suitable for several applications in the real world. These patterns not only reflect the interests of most users but also contribute a high proportion of the utility in supporting transactions. Nonetheless, the set of all discovered FHUOIs may be very large, especially for large and dense datasets or for low values of predefined minimum thresholds. For this reason, it is often quite challenging for users to analyze and use the obtained patterns. To address this issue, this paper proposes two novel algorithms named MaxCloFHUOIM and CloFHUOIM to extract compact representations of FHUOIs. The former is designed to simultaneously mine two concise representations of FHUOIs that consist of all closed FHUOIs and all maximal FHUOIs, whereas the latter only discovers the closed FHUOIs, which provide a lossless summary of all FHUOIs. The proposed algorithms rely on a novel weak upper bound on utility occupancy, to reduce the search space by quickly pruning itemsets with low utility occupancy. Especially, the algorithms integrate two new efficient strategies to prune non-closed FHUOI candidate branches early in the prefix search tree. Results from an in-depth experimental evaluation conducted on several benchmark real-life and synthetic quantitative datasets demonstrate that MaxCloFHUOIM and CloFHUOIM have excellent performance in terms of runtime, memory usage, and scalability. In particular, the proposed algorithms are up to two orders of magnitude faster than a baseline algorithm. [ABSTRACT FROM AUTHOR]

In large-scale unmanned aerial vehicle (UAV) swarm confrontation scenarios, the design of decision-making and coordination strategies becomes extremely difficult. Multi-Agent Reinforcement Learning (MARL), as a novel decision-making approach to address this issue, faces challenges such as poor scalability and the curse of dimensionality. To overcome these challenges, the paper proposes a Hierarchical Attention Actor-Critic (HAAC) algorithm. The HAAC algorithm includes a centralized critic network based on a Hierarchical Two-stage Attention Network (H2ANet), along with a hierarchical actor policy network that combines rules and reinforcement learning approaches. H2ANet is specifically designed to model the relationships between UAVs and extract crucial information from neighboring UAVs, enabling the generation of advanced cooperative and competitive strategies. The HAAC algorithm effectively reduces the dimensionality of both action and state spaces. Experimental results conducted demonstrate that the HAAC algorithm outperforms existing methods and is able to extend its learned policies to large-scale scenarios. [ABSTRACT FROM AUTHOR]

The theory of Formal Concept Analysis (FCA) finds diverse applications in fields like knowledge extraction, cognitive concept learning and data mining. The construction of a concept lattice significantly influences the effectiveness of formal concept analysis; hence, the development of high-performance algorithms for concept construction is crucial. In this paper, we introduce a novel algorithm called "Bit-Close" for formal concept construction. Bit-Close leverages bit representation and operations, fundamental to computer science, to enhance the In-Close algorithm. Furthermore, we explore the parallel method of Bit-Close. Our experimental results, obtained from multiple public and random datasets, demonstrate that Bit-Close outperforms In-Close by approximately 20% and is significantly better than other competing algorithms. [ABSTRACT FROM AUTHOR]

Surrogate-assisted evolutionary algorithms (SAEAs) are popular for solving expensive optimization problems. However, most existing SAEAs are designed for solving single-objective or multiobjective optimization problems with two or three objectives. Few works had been reported to deal with expensive many-objective optimization problems with more than three objectives because of two difficulties. One is the curse of dimensionality caused by many-objective problems, and the other is the fewer computational resources available in a limited time for expensive optimization problems. Since an effective selection method can better solve the many-objective optimization problems, high-efficiency search and accurate model can save computational resources for expensive optimization problems, this paper proposes a diverse/converged individual competition algorithm, which owns a novel diverse/converged individual competition selection mechanism, a hybrid search mechanism, and a segmentation approach. The diverse/converged individual competition selection mechanism maintains a good balance between the convergence and diversity of the selected solutions for solving many-objective optimization problems. The hybrid search mechanism performs a memetic search and genetic search at different stages of the evolution process to further generate superior solutions. The segmentation approach uses two different populations with small numbers to build two surrogate models which will predict different areas, and it can improve the accuracy of the prediction. The proposed algorithm is compared with several state-of-art algorithms on widely used benchmark functions. The experimental results show that the proposed algorithm performs significantly better than the compared algorithms. [ABSTRACT FROM AUTHOR]

Improving classification performance is an essential goal for various practical applications. Feature selection has become an important data preprocessing step in machine learning systems. However, many effective methods based on heuristic search strategies have the problem of high running costs. This paper proposes an efficient multiobjective feature selection method based on artificial immune algorithm optimization. It introduces a clone selection algorithm to explore the search space of optimal feature subsets. According to the target requirements of feature selection, combined with biological research results, this method introduces genome shuffling technology and a conditional lethal mutation mechanism to improve the search performance of the algorithm. Experimental comparisons are conducted on 21 benchmark datasets with 17 advanced feature selection methods in terms of classification accuracy, the number of feature subsets, and computational cost. The results show that the algorithm achieves the smallest number of selected features (only 3.26% compared to the lowest) and better average classification accuracy with a much lower average computational cost than others (only 3.62% compared to the lowest). [ABSTRACT FROM AUTHOR]

In Machine Learning (ML), Feature Selection (FS) plays a crucial part in reducing data's dimensionality and enhancing any proposed framework's performance. However, in real-world applications, FS work suffers from high dimensionality, computational and storage complexity, noisy or ambiguous nature, high performance, etc. The area of FS is very vast and challenging in its nature. There are lots of work that have been reported on FS over the various area of applications. This paper has discussed FS's framework and the multiple models of FS with detailed descriptions. We have also classified the various FS algorithms with respect to the data, i.e., structured or labeled data and unstructured data for the different applications of ML. We have also discussed what essential features are, the commonly used FS methods, the widely used datasets, and the widely used work done in the various ML fields for the FS task. Here we try to view the multiple comparison experimental results of FS work in different result discussions. This paper draws a descriptive survey on FS with the associated area of real-world problem domains. This paper's main objective is to understand the main idea of FS work and identify the core idea of how FS will be applicable in various problem domains. [ABSTRACT FROM AUTHOR]

Fuzzy neural network (FNN) is the product of the combination of fuzzy theory and neural network. It combines the advantages of neural network and fuzzy theory, which has achieved great success in various fields. However, on the one hand, when the practical input is intricate and high dimensional data, the existing FNN can't achieve a good modeling effect in the speed of convergence, the accuracy of modeling and generalization ability. On the other hand, the number of rules in FNN is fixed, which will also lead to the above problems in nonlinear system modeling. In this paper, a self-organizing fuzzy neural network based on Particle Swarm Optimization with improved Levenberg-Marquardt learning algorithm (SOFNN-PSO-ILM) is proposed for nonlinear system modeling. First, SOFNN based on PSO-ILM is built online by a method of constantly learning parameters and structures. In the process of structures learning, the number of fuzzy rules that have been set can be self-designed with the growing and pruning algorithm, which is based on the size of the singular value. In the process of parameters learning, PSO algorithm combined with ILM algorithm is used to update parameters. Then, the convergence and stability of SOFNN based on PSO-ILM are analyzed. Finally, the proposed method is used to model in the nonlinear system by three examples. The modeling results demonstrate that the proposed SOFNN based on PSO-ILM can model in nonlinear systems effectively. [ABSTRACT FROM AUTHOR]

Ontology, presentation of the world clearly, is a conceptualized and formalized model of knowledge. Knowledge interpretation is subjective, organizations or individuals have different expressions the same knowledge, i.e., ontology heterogeneity. It has arisen as a challenge on the Semantic Web (SW). Ontology meta-matching (OMM) technology can map two body of knowledge in the same domain effectively, and then ontology heterogeneity is addressed. The OMM is generally considered as a nonlinear global optimization problem to achieve alignment. As one of the meta-heuristic algorithms, fireworks algorithm (FA) has a distinct advantage in convergence speed and optimization precision. However, the population size of FA is usually not fixed during the exploitation and exploration process. The sparks are generated by the interaction and cooperation between fireworks, and are larger than the number of fireworks. This search behavior leads that the computational memory is costly. In paper, a novel compact fireworks algorithm (CFA) is proposed to reduce the memory consumption of FA. In CFA the distribution of populations is represented by a probabilistic model, i.e., virtual populations, and two operators are proposed to simulate the process of fireworks explosion and update disturbance vector respectively. The initialization has a great influence on the exploitation and exploration of meta-heuristic algorithm. In addition, a group of high quality initialization can often improve the optimization ability of the algorithm. In paper a novel heuristic initialization strategy (HIS) is designed to obtain a better set of initial population by introducing concept of accurateness, pseudo-weight and barycenter. In addition, HIS is more suitable for higher dimensional decision space of ontology meta-matching by a more convenient coding scheme. Finally, a non-reference matching optimization model, maximal confidence alignment model (MCAM), is proposed to evaluate the quality of alignment. In the experiment, the benchmark track of Ontology Alignment Evaluation Initiative (OAEI) is used. Experimental results demonstrate the effectiveness of the proposed methods, and the memory consumption of CFA reduces by 52.5% than FA. [ABSTRACT FROM AUTHOR]

Chameleon algorithm is a hierarchical clustering based on dynamic modeling. It can find high-quality clusters with different shapes, sizes and densities. However, Chameleon algorithm requires user-specifiedkwhen constructing sparse graph, which directly influences the clustering performance. In addition, the graph-partitioning technology used in the original algorithm, hMetis algorithm, is hard to build operation environment, and the number of partitions needs to be specified. These problems are arduous to determine without prior knowledge. In order to overcome the first problem, this paper introduces an improved natural neighbor method to construct a sparse graph, which can reflect the initial sparseness of the data. To address the second problems, this paper proposes a new method of generating sub-clusters in sparse graphs, which is simple and objective. In summary, this paper proposes Chameleon Algorithm Based on Improved Natural Neighbor Graph Generating Sub-clusters (INNGS-Chameleon). This algorithm is tested on 8 synthetic data sets and 10 UCI data sets. The results are compared with the Chameleon algorithm, its improved algorithm and several classic algorithms. The experimental results show that the INNGS-Chameleon algorithm is feasible and effective. [ABSTRACT FROM AUTHOR]

In the era of big data, the demand for multivariate time series prediction has surged, drawing increased attention to feature selection and neural networks in machine learning. However, certain feature selection methods neglect the alignment between actual data sample differences and clustering results, while neural networks lack automatic parameter adjustment in response to changing target features. This paper presents the MBSSA-Bi-AESN model, a Bi-directional Adaptive Echo State Network that utilizes the modified salp swarm algorithm (MBSSA) and feature selection to address the limitations of manually set parameters. Initial feature subset selection involves assigning weights based on the consistency of clustering results with differences. Subsequently, the four critical parameters in the Bi-AESN model are optimized using MBSSA. The optimized Bi-AESN model and selected feature subset are then integrated for simultaneous model learning and optimal feature subset selection. Experimental analysis on eight datasets demonstrates the superior prediction accuracy of the MBSSA-Bi-AESN model compared to benchmark models, underscoring its feasibility, validity, and universality. [ABSTRACT FROM AUTHOR]

As an effective extension of rough set theory, the variable precision neighborhood rough set model has been applied to the attribute dependency-based improvement of decision tree algorithm of the solution concerning continuous data. However, the boundary region, as an effective description of the uncertainty of knowledge, has not been taken into account in the existing algorithms. In this paper, we define a novel decision rule based on boundary region and attribute dependency, and construct a decision tree algorithm via this decision rule. First, we define a measure called boundary coefficient based on the boundary region, which can be used for comparative quantitative analysis. Second, we define the boundary mixed attribute dependency by combining the boundary coefficient and the attribute dependency, which can consider both the boundary case of the target set and the attribute dependency. Finally, a novel decision tree algorithm is proposed by using the boundary mixed attribute dependency as the decision rule. The experimental results show that with a slight increase in leaf nodes, the total running time decreases and the maximum accuracy increases to 0.9518, which indicates the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

The closed contiguous sequential pattern combines the advantages of closedness constraints and contiguity constraints and in recent years has been widely used in the fields of sequence classification, traffic trajectory visualization and football player trajectory analysis. Most of the previously developed closed contiguous sequential pattern mining algorithms pose some challenges. For instance, CCSpan, BP-CCSM, and LCCspm cannot mine the large-scale sequence database with reasonable time and memory usage, while C3Ro, which can mine patterns with multiple constraints, does not consider the specificity induced by the contiguity constraint of the pattern. To address these problems and improve the efficiency of mining closed contiguous sequential patterns, in this paper, we present an algorithm called CCSMP based on the pattern relation graph. Pattern relation graph is a novel data structure that has some key properties related to closed contiguous sequential pattern mining. In the experimental section, we not only conducted extensive experiments on real datasets to evaluate the performance and scalability of CCSMP but also analyzed the running time of each step of CCSMP to verify the effectiveness of the pattern relation graph. The experimental results show that CCSMP outperforms the existing state-of-the-art algorithm in most cases and that the use of the pattern relation graph can significantly reduce the time for closure checking. [ABSTRACT FROM AUTHOR]

Meta-heuristic algorithms have been extensively utilized in feature selection tasks because they can obtain the global optimal solution. However, the meta-heuristic algorithm will take too much time in the face of a large number of samples. Although most of the studies compromise to approximate optimal solutions for avoiding time-consuming problems, a new problem with reduced classification performance, especially classification stability, is then generated. Aiming to above problems, this paper proposes a new feature selection framework. First, this framework exploits a voting ensemble strategy to improve classification stability by reducing the impact of misclassified labels on the overall classification results. Second, the framework uses a data perturbation strategy to enhance classification accuracy. In particular, the data perturbation strategy is able to generate more neighborhood relationships in the dataset, which could reveal the distribution of various features of the samples. A voting ensemble of different feature distributions is capable of extracting more information from the dataset, then the initially misclassified samples are more likely to be returned to the correct classification. Third, the framework takes a random sampling accelerator into account to solve the problem of excessive time consumption by reducing the size of the search sample space. Finally, for the sake of verifying the effectiveness of the proposed framework, four meta-heuristic feature selection methods based on a neighborhood rough set are compared on 20 datasets. The experimental results indicate that our framework could improve classification performance and accelerate feature selection, particularly in confronting large sample sizes. [ABSTRACT FROM AUTHOR]

This study investigates the joint replenishment and delivery problem taking trade credits the delay in payment permitted by suppliers and capacity-constrained into consideration (CJRD-TC). The CJRD-TC aims to minimize the total system costs by finding a reasonable replenishment and delivery schedule policy. In this paper, an opposite-based RUNge Kutta optimization algorithm (OBRUN) is proposed to compare with other intelligent algorithms, including the differential evolution (DE), RUNge Kutta optimization algorithm (RUN), sine cosine algorithm (SCA), and adaptive differential evolution with optimal external archive (JADE). The proposed algorithm is tested on 14 optimization functions. The values of average, standard deviation, and the best are the lowest for 12 functions, which demonstrates OBRUN is superior to other algorithms. We also provide graphs of search history, 2D views of functions, trajectory curve, average fitness history, and convergence curve of OBRUN, the balance between exploration and exploitation over the course of iterations can be observed. The convergence curves of five algorithms are also depicted, which show OBRUN has a fast convergence rate and accuracy than other algorithms. We conduct numerical experiments, parameter tuning, and sensitivity analysis for each of the small-scale, medium-scale, and large-scale number of items. The OBRUN can always find better solutions with the lowest mean values and the best-found total cost results than other algorithms. Indicators of the average optimality gap and the optimality gap dominate by the value of zero. These computational results have demonstrated OBRUN is a suitable and effective candidate for managers to solve the practical CJRD-TC problems. [ABSTRACT FROM AUTHOR]

Federated medical relation extraction enables multiple clients to train a deep network collaboratively without sharing their raw medical data. To handle the heterogeneous label distribution across clients, most of the existing works enforce regularization between local and global models during updating. In this paper, we propose the concept of major classifier vectors, which are a group of classifier vectors that characterize the representation space of relation classes well. They are obtained by comparing the inter-classifier similarity between clients, which is an ensembling method that avoids the bias introduced by weighted aggregation. We propose an algorithm named FedCMC, which restricts the updating of local models by contrasting with major classifier vectors to avoid overfitting to the local label distribution by comparison with major classifier vectors. Extensive experiments show that FedCMC outperforms the other state-of-the-art federated learning (FL) algorithms on three medical relation extraction datasets. [ABSTRACT FROM AUTHOR]

Periodic frequent-pattern mining (PFPM) is a vital knowledge discovery technique that identifies periodically occurring patterns in a temporal database. Although traditional PFPM algorithms have many applications, they often produce a large set of periodic-frequent patterns (PFPs) in a database. As a result, analyzing PFPs can be very time-consuming for users. Moreover, a large set of PFPs makes PFPM algorithms less efficient regarding runtime and memory consumption. This paper handles this problem by proposing a novel model of closed periodic-frequent patterns (CPFPs) found in databases. CPFPs are less expensive to mine because they represent a concise and lossless subset uniquely describing the entire set of PFPs. We also present an efficient depth-first search algorithm, called Closed Periodic-Frequent Pattern-Miner (CPFP-Miner), to discover the patterns. The proposed algorithm utilizes the weighted ordering of the patterns concept to reduce the patterns' search space. On the other hand, the current periodicity concept is also applied to prune aperiodic patterns from the search space. Extensive experiments on both real-world and synthetic databases demonstrate that the CPFP-Miner algorithm is efficient. It outperforms the state-of-the-art algorithms regarding runtime requirements, memory consumption, and energy consumption on several real-world and synthetic databases. Additionally, the scalability of the CPFP-Miner algorithm is demonstrated to be more effective and productive than the state-of-the-art algorithms. Finally, we present two case studies to show the functionality of the proposed patterns. [ABSTRACT FROM AUTHOR]

In crowdsourcing system, each instance will be usually labeled multiple times by different workers. After obtaining the multiple noise labels of data, ground truth inference algorithms are used to infer unknown true labels of instances. However, most existing ground truth inference algorithms only utilize the information in the multiple noise labels themselves while ignoring the instance similarity. This paper proposes a novel ground truth inference algorithm based on instance similarity to further improve the performance of ground truth inference. Because similar instances are more likely to be clustered in the same cluster and similar instances are more likely to have similar labels, both fuzzy c-means clustering (FCM) and k-nearest neighbors algorithm (kNN) are used to explore the instance similarity in this paper. Specifically, FCM is firstly used to adjust label distributions of instances. Then the labels of instances are inferred according to their label distributions and kNN algorithm. Based on the instance similarity, the instances with reliable label distributions will influence the instances with unreliable label distributions. The experimental results on benchmark and real-world data sets validate that using the instance similarity can effectively enhance the performance of ground truth inference. [ABSTRACT FROM AUTHOR]

Influence maximization (IM) problem is to choose k seed nodes in the social network to maximize the number of nodes that can be affected. As an essential part of the social network analysis, the influence maximization problem has been extensively investigated. Most existing algorithms usually set the activation probability to a fixed value or reciprocal of the in-degree. However, this operation is not accurate because the activation probability is more complicated than those mentioned above in the real social networks. To solve this problem, this paper proposes a new algorithm called Low-Dimensional Representation Learning for IM (LDRLIM) to address the IM problem. The LDRLIM utilizes Discount-degree Descending (DED) search strategy to generate the candidate nodes set ℂ and IC Walk to obtain the influence context of the candidate nodes. The LDRLIM algorithm learns the low-dimensional embedding vectors of influencers and susceptible nodes according to a Multi-task Neural Network Low-dimensional Representation learning model (MNNLR). Afterwards, the Similarity Influence (SI) of the node is obtained according to the representation vectors of the nodes. The LDRLIM algorithm employs SI to consider the potential influence relationship between nodes by embedding vector. Furthermore, the Cost-Effective Lazy Forward (CELF) strategy is used to accelerate the process of selecting the influential nodes, which avoids a large amount of model simulation time to improve efficiency. Therefore, the proposed LDRLIM algorithm is suitable for large-scale social networks. The experimental results on seven real-world datasets indicate that the LDRLIM significantly outperforms other comparison algorithms. [ABSTRACT FROM AUTHOR]

This paper focuses on developing an improved multi-focus image fusion (MFIF) algorithm. Existing spatial domain algorithms dependent on the obtained fusion decision map still lead to unexpected ghosting, blurred, edges as well as blocking effects such that the visual effect of image fusion is seriously degraded. To overcome these shortages, an improved MFIF algorithm is developed with the help of a novel multi-scale weighted focus measure and a decision map optimization technique. First, a novel multi-scale measurement template is designed in order to effectively extract the gradient information of rich texture regions, smooth regions as well as transitional regions between the aforementioned regions simultaneously. Then, an improved calculation scheme of the focus score matrix is designed based on the weighted sum of the focus measure maps in each region window centered on a concerned pixel, under which the advantage of pixel-by-pixel weighting is employed. In what follows, an initial decision map is obtained in light of the focus score matrix combined with threshold filtering, which is employed to eliminate the small isolated regions caused by some misclassified pixels. Furthermore, an accurate decision map is received with the help of the optimization capability of guided filtering to avoid edge unexpected artificial textures. In comparison with block-based fusion algorithms, our algorithm developed in this paper extracts the focus regions pixel-by-pixel, thereby helping to reduce the blocking effects that appear in the fusion image. Finally, some intensive comparison analysis based on common datasets is performed to verify the superiority over state-of-the-art methods in both visual qualitative and quantitative evaluations. [ABSTRACT FROM AUTHOR]

Traditional fault diagnosis methods of DC (direct current) motors require high expertise and human labor. However, the other disadvantages of these methods are low efficiency and poor accuracy. To address these problems, a new adaptive and intelligent mechanical fault diagnosis method for DC motors based on variational mode decomposition (VMD), singular value decomposition (SVD), and residual deep convolutional neural networks with wide first-layer kernels (R-WDCNN) was proposed. First, the vibration signals of a DC motor were collected by a designed acquisition system. Subsequently, VMD was employed to decompose the raw signals adaptively into several intrinsic mode functions (IMFs). Moreover, the transient frequency means method, which can quickly and accurately obtain the optimal value of K, is proposed. SVD was applied to reduce the dimensionality of the IMF matrix for further feature extraction. Finally, the reconstructed matrix containing the main fault feature information was used to train and test the R-WDCNN. Based on residual learning, identification and classification of four types of vibration signals were achieved, while the R-WDCNN was optimized by the adaptive batch normalization algorithm (AdaBN). The recognition rate and the convergence were improved by this classifier. The results show that the method proposed in this paper has better adaptability and intelligence than other methods, and the R-WDCNN can reach a 94% recognition rate on unknown samples. Therefore, the proposed method is more intelligent and accurate than other methods. [ABSTRACT FROM AUTHOR]

In the express industry, few people pay attention to the rationality of the layout of the transfer station at the end of rural express outlets. Aiming at the difficulty in selecting the location of express terminal transfer stations in rural areas, this paper presents a new transfer station location algorithm. This method is based on the reverse nearest neighbor algorithm and local density characteristics to initially screen the location of the transfer station in the area, and then determine its location and coverage area based on the density from the appropriate distance. Finally, by calculating the outlier index of each station in the distribution area, the boundary of the distribution area is redefined. Experiments show that the express transfer stations selected by this method can meet the service requirements of 98.8% of the area and 99.5% of the population in Yutian County, and the division accuracy of the distribution range has been improved to a certain extent compared with the traditional algorithm. In particular, for the selection of small-range peak points, the accuracy of the traditional density peak algorithm has been significantly improved. [ABSTRACT FROM AUTHOR]

3D point cloud registration has a wide range of applications in object shape detection, robot navigation and 3D reconstruction. Aiming at the problems of the traditional ICP registration algorithm, such as slow convergence speed and high requirements for the initial point cloud position, this paper proposes a coarse-fine point cloud registration method based on a fast and robust local point-pair feature (LPPF) and the ICP algorithm. The LPPF feature descriptor is an improved descriptor for the local application of classic point-pair features and is a histogram descriptor formed by counting the feature information of the local point cloud neighborhood. This paper completes point cloud registration through keypoint extraction, LPPF feature description, feature matching, coarse registration and fine registration. To verify the effectiveness of this method, under the evaluation indices of L1, RMSE and MAE, we analyzed the experimental results from the three aspects of descriptors, coarse registration and coarse-fine registration. Under Gaussian noise conditions, LPPF compared to the second-ranked descriptor, the L1 scores of LPPF on the Stanford, Kinect and Princeton datasets increased by 12%, 12.4% and 9.1%, respectively. The coarse registration experiment is compared with 5 classic descriptors on 3 commonly used datasets. The LPPF feature descriptor has higher registration accuracy and less registration time. Finally, the coarse-fine registration experiment shows that our method can reduce the number of iterations of the ICP algorithm by 77% under optimal conditions. [ABSTRACT FROM AUTHOR]