Showing papers in "Knowledge Based Systems in 2020"
TL;DR: A well-defined “generation rate” term is proved to invigorate EO’s ability in exploration, exploitation, and local minima avoidance, and its performance is statistically similar to SHADE and LSHADE-SPACMA.
Abstract: This paper presents a novel, optimization algorithm called Equilibrium Optimizer (EO), inspired by control volume mass balance models used to estimate both dynamic and equilibrium states. In EO, each particle (solution) with its concentration (position) acts as a search agent. The search agents randomly update their concentration with respect to best-so-far solutions, namely equilibrium candidates, to finally reach to the equilibrium state (optimal result). A well-defined “generation rate” term is proved to invigorate EO’s ability in exploration, exploitation, and local minima avoidance. The proposed algorithm is benchmarked with 58 unimodal, multimodal, and composition functions and three engineering application problems. Results of EO are compared to three categories of existing optimization methods, including: (i) the most well-known meta-heuristics, including Genetic Algorithm (GA), Particle Swarm Optimization (PSO); (ii) recently developed algorithms, including Grey Wolf Optimizer (GWO), Gravitational Search Algorithm (GSA), and Salp Swarm Algorithm (SSA); and (iii) high performance optimizers, including CMA-ES, SHADE, and LSHADE-SPACMA. Using average rank of Friedman test, for all 58 mathematical functions EO is able to outperform PSO, GWO, GA, GSA, SSA, and CMA-ES by 60%, 69%, 94%, 96%, 77%, and 64%, respectively, while it is outperformed by SHADE and LSHADE-SPACMA by 24% and 27%, respectively. The Bonferroni–Dunnand Holm’s tests for all functions showed that EO is significantly a better algorithm than PSO, GWO, GA, GSA, SSA and CMA-ES while its performance is statistically similar to SHADE and LSHADE-SPACMA. The source code of EO is publicly availabe at https://github.com/afshinfaramarzi/Equilibrium-Optimizer , http://built-envi.com/portfolio/equilibrium-optimizer/ and http://www.alimirjalili.com/SourceCodes/EOcode.zip .
TL;DR: New generator and discriminator of Generative Adversarial Network (GAN) are designed in this paper to generate more discriminant fault samples using a scheme of global optimization to solve the problem of unbalanced fault samples.
Abstract: Deep learning can be applied to the field of fault diagnosis for its powerful feature representation capabilities. When a certain class fault samples available are very limited, it is inevitably to be unbalanced. The fault feature extracted from unbalanced data via deep learning is inaccurate, which can lead to high misclassification rate. To solve this problem, new generator and discriminator of Generative Adversarial Network (GAN) are designed in this paper to generate more discriminant fault samples using a scheme of global optimization. The generator is designed to generate those fault feature extracted from a few fault samples via Auto Encoder (AE) instead of fault data sample. The training of the generator is guided by fault feature and fault diagnosis error instead of the statistical coincidence of traditional GAN. The discriminator is designed to filter the unqualified generated samples in the sense that qualified samples are helpful for more accurate fault diagnosis. The experimental results of rolling bearings verify the effectiveness of the proposed algorithm.
TL;DR: Li et al. as mentioned in this paper proposed dyngraph2vec, which learns the temporal transitions in the network using a deep architecture composed of dense and recurrent layers to predict unseen links with higher precision.
Abstract: Learning graph representations is a fundamental task aimed at capturing various properties of graphs in vector space. The most recent methods learn such representations for static networks. However, real-world networks evolve over time and have varying dynamics. Capturing such evolution is key to predicting the properties of unseen networks. To understand how the network dynamics affect the prediction performance, we propose an embedding approach which learns the structure of evolution in dynamic graphs and can predict unseen links with higher precision. Our model, dyngraph2vec, learns the temporal transitions in the network using a deep architecture composed of dense and recurrent layers. We motivate the need for capturing dynamics for the prediction on a toy dataset created using stochastic block models. We then demonstrate the efficacy of dyngraph2vec over existing state-of-the-art methods on two real-world datasets. We observe that learning dynamics can improve the quality of embedding and yield better performance in link prediction.
TL;DR: This survey discusses the role of deep learning in intrusion detection, the impact of intrusion detection datasets, and the efficiency and effectiveness of the proposed approaches, and provides a novel fine-grained taxonomy that categorizes the current state-of-the-art deep learning-based IDSs with respect to different facets.
Abstract: The massive growth of data that are transmitted through a variety of devices and communication protocols have raised serious security concerns, which have increased the importance of developing advanced intrusion detection systems (IDSs). Deep learning is an advanced branch of machine learning, composed of multiple layers of neurons that represent the learning process. Deep learning can cope with large-scale data and has shown success in different fields. Therefore, researchers have paid more attention to investigating deep learning for intrusion detection. This survey comprehensively reviews and compares the key previous deep learning-focused cybersecurity surveys. Through an extensive review, this survey provides a novel fine-grained taxonomy that categorizes the current state-of-the-art deep learning-based IDSs with respect to different facets, including input data, detection, deployment, and evaluation strategies. Each facet is further classified according to different criteria. This survey also compares and discusses the related experimental solutions proposed as deep learning-based IDSs. By analysing the experimental studies, this survey discusses the role of deep learning in intrusion detection, the impact of intrusion detection datasets, and the efficiency and effectiveness of the proposed approaches. The findings demonstrate that further effort is required to improve the current state-of-the art. Finally, open research challenges are identified, and future research directions for deep learning-based IDSs are recommended.
TL;DR: The results show that PO outperforms all other algorithms, and consistency in performance on such a comprehensive suite of benchmark functions proves the versatility of the algorithm.
Abstract: This paper proposes a novel global optimization algorithm called Political Optimizer (PO), inspired by the multi-phased process of politics. PO is the mathematical mapping of all the major phases of politics such as constituency allocation, party switching, election campaign, inter-party election, and parliamentary affairs. The proposed algorithm assigns each solution a dual role by logically dividing the population into political parties and constituencies, which facilitates each candidate to update its position with respect to the party leader and the constituency winner. Moreover, a novel position updating strategy called recent past-based position updating strategy (RPPUS) is introduced, which is the mathematical modeling of the learning behaviors of the politicians from the previous election. The proposed algorithm is benchmarked with 50 unimodal, multimodal, and fixed dimensional functions against 15 state of the art algorithms. We show through experiments that PO has an excellent convergence speed with good exploration capability in early iterations. Root cause of such behavior of PO is incorporation of RPPUS and logical division of the population to assign dual role to each candidate solution. Using Wilcoxon rank-sum test, PO demonstrates statistically significant performance over the other algorithms. The results show that PO outperforms all other algorithms, and consistency in performance on such a comprehensive suite of benchmark functions proves the versatility of the algorithm. Furthermore, experiments demonstrate that PO is invariant to function shifting and performs consistently in very high dimensional search spaces. Finally, the applicability on real-world applications is demonstrated by efficiently solving four engineering optimization problems.
TL;DR: Through verifying the benchmark functions, the advanced binary GWO is superior to the original BGWO in the optimality, time consumption and convergence speed.
Abstract: Grey Wolf Optimizer (GWO) is a new swarm intelligence algorithm mimicking the behaviours of grey wolves. Its abilities include fast convergence, simplicity and easy realization. It has been proved its superior performance and widely used to optimize the continuous applications, such as, cluster analysis, engineering problem, training neural network and etc. However, there are still some binary problems to optimize in the real world. Since binary can only be taken from values of 0 or 1, the standard GWO is not suitable for the problems of discretization. Binary Grey Wolf Optimizer (BGWO) extends the application of the GWO algorithm and is applied to binary optimization issues. In the position updating equations of BGWO, the a parameter controls the values of A and D , and influences algorithmic exploration and exploitation. This paper analyses the range of values of A D under binary condition and proposes a new updating equation for the a parameter to balance the abilities of global search and local search. Transfer function is an important part of BGWO, which is essential for mapping the continuous value to binary one. This paper includes five transfer functions and focuses on improving their solution quality. Through verifying the benchmark functions, the advanced binary GWO is superior to the original BGWO in the optimality, time consumption and convergence speed. It successfully implements feature selection in the UCI datasets and acquires low classification errors with few features.
TL;DR: The thrust of this review is to outline emerging applications of DL and provide a reference to researchers seeking to use DL in their work for pattern recognition with unparalleled learning capacity and the ability to scale with data.
Abstract: Deep learning (DL) has solved a problem that a few years ago was thought to be intractable — the automatic recognition of patterns in spatial and temporal data with an accuracy superior to that of humans. It has solved problems beyond the realm of traditional, hand-crafted machine learning algorithms and captured the imagination of practitioners who are inundated with all types of data. As public awareness of the efficacy of DL increases so does the desire to make use of it. But even for highly trained professionals it can be daunting to approach the rapidly increasing body of knowledge in the field. Where does one start? How does one determine if a particular DL model is applicable to their problem? How does one train and deploy them? With these questions in mind, we present an overview of some of the key DL architectures. We also discuss some new automatic architecture optimization protocols that use multi-agent approaches. Further, since guaranteeing system uptime is critical to many applications, a section dwells on using DL for fault detection and mitigation. This is followed by an exploratory survey of several areas where DL emerged as a game-changer: fraud detection in financial applications, financial time-series forecasting, predictive and prescriptive analytics, medical image processing, power systems research and recommender systems. The thrust of this review is to outline emerging applications of DL and provide a reference to researchers seeking to use DL in their work for pattern recognition with unparalleled learning capacity and the ability to scale with data.
TL;DR: A normalized convolutional neural network is proposed for the diagnosis of different fault severities and orientations considering data imbalance and variable working conditions and results show that the proposed method has excellent diagnosis accuracy and admirable robustness, and also has sufficient stability on the data imbalance.
Abstract: Intelligent fault detection and diagnosis, as an important approach, play a crucial role in ensuring the stable, reliable and safe operation of rolling bearings, which is one of the most important components in the rotating machinery. In real industries, it is common to face that the issues of severe data imbalance and distribution difference since the number of fault data is small and the equipments frequently change the working conditions according to the production. To accurately and automatically identify the conditions of rolling bearings, a normalized convolutional neural network is proposed for the diagnosis of different fault severities and orientations considering data imbalance and variable working conditions. First, the batch normalization is adopted as a novel application to eliminate feature distribution difference, which is the prerequisite for ensuring generalization ability under different working conditions. Then, a special model structure is established and the overall performances of the proposed model are optimized by iterative update, which combines the exponential moving average technology. Finally, the proposed model is applied to the fault diagnosis under different data imbalance cases and working conditions. The effectiveness of the proposed method is verified based on two popular experiment dataset, and the diagnosis performance is widely evaluated in different scenarios. Comparisons with other commonly used methods and related works on the same dataset demonstrate the superiority of the proposed method. The results show that the proposed method has excellent diagnosis accuracy and admirable robustness, and also has sufficient stability on the data imbalance.
TL;DR: A novel approach named deep transfer multi-wavelet auto-encoder is presented for gearbox intelligent fault diagnosis with few training samples and transfer diagnosis cases for different fault severities and compound faults of gearbox confirm the feasibility of the proposed approach.
Abstract: Lack of typical fault samples remains a huge challenge for intelligent fault diagnosis of gearbox. In this paper, a novel approach named deep transfer multi-wavelet auto-encoder is presented for gearbox intelligent fault diagnosis with few training samples. Firstly, new-type deep multi-wavelet auto-encoder is designed for learning important features of the collected vibration signals of gearbox. Secondly, high-quality auxiliary samples are selected based on similarity measure to well pre-train a source model sharing similar characteristics with the target domain. Thirdly, parameter knowledge acquired from the source model is transferred to target model using very few target training samples. Transfer diagnosis cases for different fault severities and compound faults of gearbox confirm the feasibility of the proposed approach even if the working conditions have significant changes.
TL;DR: A new decision fusion strategy is designed to flexibly fuse each individual target CNN to obtain the comprehensive result of the proposed ensemble transfer convolutional neural networks driven by multi-channel signals.
Abstract: Automatic and reliable fault diagnosis of rotating machinery cross working conditions is of practical importance. For this purpose, ensemble transfer convolutional neural networks (CNNs) driven by multi-channel signals are proposed in this paper. Firstly, a series of source CNNs modified with stochastic pooling and Leaky rectified linear unit (LReLU) are pre-trained using multi-channel signals. Secondly, the learned parameter knowledge of each individual source CNN is transferred to initialize the corresponding target CNN which is then fine-tuned by a few target training samples. Finally, a new decision fusion strategy is designed to flexibly fuse each individual target CNN to obtain the comprehensive result. The proposed method is used to analyze multi-channel signals measured from rotating machinery. The comparison result shows the superiorities of the proposed method over the existing deep transfer learning methods.
TL;DR: A new type of group decision making problems in which experts will provide his/her interval fuzzy preference relations over alternatives under social network environment is considered and a new model to help experts reach consensus is proposed.
Abstract: With the rapid development of information, communication and techniques, social network group decision making problems which allow information exchange and communication among experts are more and more common in recent years. How to use social relationships generated by social networks to promote consensus among experts has been becoming a hot topic in the field of group decision making. In this paper, we consider a new type of group decision making problems in which experts will provide his/her interval fuzzy preference relations over alternatives under social network environment and propose a new model to help experts reach consensus. In the proposed model, we first define the individual consensus measure and the group consensus measure, and then use a network partition algorithm to detect sub-networks of experts, based on which the leadership of experts can be identified. Afterwards, by considering the leadership and the bounded confidence levels of experts, a new feedback mechanism which can provide acceptable advice to experts who need to modify their opinions is devised and a consensus reaching algorithm is further developed. To demonstrate the performance of the proposed consensus model and algorithm, a hypothetical application and some simulation analysis are provided eventually.
TL;DR: A novel method based on enhanced deep gated recurrent unit and complex wavelet packet energy moment entropy and a modified training algorithm based on learning rate decay strategy is developed to enhance the prognosis capability of the constructed deep model.
Abstract: Early fault prognosis of bearing is a very meaningful yet challenging task to improve the security of rotating machinery. For this purpose, a novel method based on enhanced deep gated recurrent unit and complex wavelet packet energy moment entropy is proposed in this paper. First, complex wavelet packet energy moment entropy is defined as a new monitoring index to characterize bearing performance degradation. Second, deep gated recurrent unit network is constructed to capture the nonlinear mapping relationship hidden in the defined monitoring index. Finally, a modified training algorithm based on learning rate decay strategy is developed to enhance the prognosis capability of the constructed deep model. The proposed method is applied to analyze the simulated and experimental signals of bearing. The results demonstrate that the proposed method is more superior in sensibility and accuracy to the existing methods.
TL;DR: Empirical results demonstrate that the selective naive Bayes shows superior classification accuracy, yet at the same time maintains the simplicity and efficiency.
Abstract: Naive Bayes is one of the most popular data mining algorithms. Its efficiency comes from the assumption of attribute independence, although this might be violated in many real-world data sets. Many efforts have been done to mitigate the assumption, among which attribute selection is an important approach. However, conventional efforts to perform attribute selection in naive Bayes suffer from heavy computational overhead. This paper proposes an efficient selective naive Bayes algorithm, which adopts only some of the attributes to construct selective naive Bayes models. These models are built in such a way that each one is a trivial extension of another. The most predictive selective naive Bayes model can be selected by the measures of incremental leave-one-out cross validation. As a result, attributes can be selected by efficient model selection. Empirical results demonstrate that the selective naive Bayes shows superior classification accuracy, yet at the same time maintains the simplicity and efficiency.
TL;DR: This survey is going to take a glance at the evolution of both semantic and instance segmentation work based on CNN, and specified comparative architectural details of some state-of-the-art models.
Abstract: From the autonomous car driving to medical diagnosis, the requirement of the task of image segmentation is everywhere. Segmentation of an image is one of the indispensable tasks in computer vision. This task is comparatively complicated than other vision tasks as it needs low-level spatial information. Basically, image segmentation can be of two types: semantic segmentation and instance segmentation. The combined version of these two basic tasks is known as panoptic segmentation. In the recent era, the success of deep convolutional neural networks (CNN) has influenced the field of segmentation greatly and gave us various successful models to date. In this survey, we are going to take a glance at the evolution of both semantic and instance segmentation work based on CNN. We have also specified comparative architectural details of some state-of-the-art models and discuss their training details to present a lucid understanding of hyper-parameter tuning of those models. We have also drawn a comparison among the performance of those models on different datasets. Lastly, we have given a glimpse of some state-of-the-art panoptic segmentation models.
TL;DR: A simple but fast DPeak, namely FastDPeak, 1 is proposed, which runs in about O ( n l o g ( n ) ) expected time in the intrinsic dimensionality and replaces density with kNN-density, which is computed by fast kNN algorithm such as cover tree, yielding huge improvement for density computations.
Abstract: Density Peak (DPeak) clustering algorithm is not applicable for large scale data, due to two quantities, i.e, ρ and δ , are both obtained by brute force algorithm with complexity O ( n 2 ) . Thus, a simple but fast DPeak, namely FastDPeak, 1 is proposed, which runs in about O ( n l o g ( n ) ) expected time in the intrinsic dimensionality. It replaces density with kNN-density, which is computed by fast kNN algorithm such as cover tree, yielding huge improvement for density computations. Based on kNN-density, local density peaks and non-local density peaks are identified, and a fast algorithm, which uses two different strategies to compute δ for them, is also proposed with complexity O ( n ) . Experimental results show that FastDPeak is effective and outperforms other variants of DPeak.
TL;DR: Experiments on a real-world dataset demonstrate that the proposed model outperformed the other state-of-the-art methods, and this provides evidence pointing to the success of employing deep learning and multi-criteria in recommendation systems.
Abstract: Recommender systems have been in existence everywhere with most of them using single ratings in prediction. However, multi-criteria predictions have been proved to be more accurate. Recommender systems have many techniques; collaborative filtering is one of the most commonly used. Deep learning has achieved impressive results in many domains such as text, voice, and computer vision. Lately, deep learning for recommender systems began to gain massive interest, and many recommendation models based on deep learning have been proposed. However, as far as we know, there is not yet any study which gathers multi-criteria recommendation and collaborative filtering with deep learning. In this work, we propose a novel multi-criteria collaborative filtering model based on deep learning. Our model contains two parts: in the first part, the model obtains the users and items’ features and uses them as an input to the criteria ratings deep neural network, which predicts the criteria ratings. Those criteria ratings constitute the input to the second part, which is the overall rating deep neural network and is used to predict the overall rating. Experiments on a real-world dataset demonstrate that our proposed model outperformed the other state-of-the-art methods, and this provides evidence pointing to the success of employing deep learning and multi-criteria in recommendation systems.
TL;DR: Experimental results have shown that the proposed detection strategy outperforms recent techniques as it introduces the maximum accuracy rate.
Abstract: COVID-19 infection is growing in a rapid rate. Due to unavailability of specific drugs, early detection of (COVID-19) patients is essential for disease cure and control. There is a vital need to detect the disease at early stage and instantly quarantine the infected people. Many research have been going on, however, none of them introduces satisfactory results yet. In spite of its simplicity, K-Nearest Neighbor (KNN) classifier has proven high flexibility in complex classification problems. However, it can be easily trapped. In this paper, a new COVID-19 diagnose strategy is introduced, which is called COVID-19 Patients Detection Strategy (CPDS). The novelty of CPDS is concentrated in two contributions. The first is a new hybrid feature selection Methodology (HFSM), which elects the most informative features from those extracted from chest Computed Tomography (CT) images for COVID-19 patients and non COVID-19 peoples. HFSM is a hybrid methodology as it combines evidence from both wrapper and filter feature selection methods. It consists of two stages, namely; Fast Selection Stage (FS2) and Accurate Selection Stage (AS2). FS2relies on filter, while AS2 uses Genetic Algorithm (GA) as a wrapper method. As a hybrid methodology, HFSM elects the significant features for the next detection phase. The second contribution is an enhanced K-Nearest Neighbor (EKNN) classifier, which avoids the trapping problem of the traditional KNN by adding solid heuristics in choosing the neighbors of the tested item. EKNN depends on measuring the degree of both closeness and strength of each neighbor of the tested item, then elects only the qualified neighbors for classification. Accordingly, EKNN can accurately detect infected patients with the minimum time penalty based on those significant features selected by HFSM technique. Extensive experiments have been done considering the proposed detection strategy as well as recent competitive techniques on the chest CT images. Experimental results have shown that the proposed detection strategy outperforms recent techniques as it introduces the maximum accuracy rate.
TL;DR: A novel multi-label relevance–redundancy feature selection method based on Ant colony optimization (ACO) for the first time, called MLACO, which is compared against five well-known and state-of-the-art feature selection methods using ML-KNN classifier.
Abstract: Nowadays, with emerge the multi-label datasets, the multi-label learning processes attracted interest and increasingly applied to different fields. In such learning processes, unlike single-label learning, instances have more than one class label simultaneously. Also, multi-label learning suffers from the curse of dimensionality, and thus, the feature selection becomes a difficult task. In this paper, we propose a novel multi-label relevance–redundancy feature selection method based on Ant colony optimization (ACO) for the first time, called MLACO. By introducing two unsupervised and supervised heuristic functions, MLACO tries to search in the features space to find the most promising features with the lowest redundancy (unsupervised) and highest relevancy with class labels (supervised) through several iterations. For speeding up the convergence of the algorithm, the normalized cosine similarity between features and class labels have been used as the initial pheromone of each ant. The proposed method does not take into account any learning algorithm, and it can be classified as a filter-based method. We compare the performance of the MLACO against five well-known and state-of-the-art feature selection methods using ML-KNN classifier. The experimental results on several frequently used datasets show the superiority of the MLACO in different multi-label evaluation measures criteria and runtime.
TL;DR: The LR-SMOTE algorithm is proposed to make the newly generated samples close to the sample center, avoid generating outlier samples or changing the distribution of data sets, and shows better performance than the SMOTE algorithm in terms of G-means value, F-measure value and AUC.
Abstract: Machine learning classification algorithms are currently widely used. One of the main problems faced by classification algorithms is the problem of unbalanced data sets. Classification algorithms are not sensitive to unbalanced data sets, therefore, it is difficult to classify unbalanced data sets. There is also a problem of unbalanced data categories in the field of loose particle detection of sealed electronic components. The signals generated by internal components are always more than the signals generated by loose particles, which easily leads to misjudgment in classification. To classify unbalanced data sets more accurately, in this paper, based on the traditional oversampling SMOTE algorithm, the LR-SMOTE algorithm is proposed to make the newly generated samples close to the sample center, avoid generating outlier samples or changing the distribution of data sets. Experiments were carried out on four sets of UCI public data sets and six sets of self-built data sets. Unmodified data sets balanced by LR-SMOTE and SMOTE algorithms used random forest algorithm and support vector machine algorithm respectively. The experimental results show that the LR-SMOTE has better performance than the SMOTE algorithm in terms of G-means value, F-measure value and AUC.
TL;DR: A novel hybrid deep learning model is presented, which combines the VMD (variational mode decomposition) method and the LSTM (long short-term memory) network to construct a forecasting model, which has superior performance for non-ferrous metals price forecasting.
Abstract: Non-ferrous metals are indispensable industrial materials and strategic supports of national economic development. The price forecasting of non-ferrous metals is critical for investors, policymakers, and researchers. Nevertheless, an accurate and robust non-ferrous metals price forecasting is a difficult yet challenging problem due to severe fluctuations and irregular cycles in the metal price evolution. Motivated by the ”Divide-and-Conquer” principle, we present a novel hybrid deep learning model, which combines the VMD (variational mode decomposition) method and the LSTM (long short-term memory) network to construct a forecasting model in this paper. Here, the VMD method is firstly employed to disassemble the original price series into several components. The LSTM network is used to forecast for each component. Lastly, the forecasting results of each component are aggregated to formulate an ultimate forecasting output for the original price series. To investigate the forecasting performance of the proposed model, extensive experiments have been executed using the LME (London Metal Exchange) daily future prices of Zinc, Copper and Aluminum, and other six state-of-the-art methods are included for comparison. The experiment results demonstrate that the proposed model has superior performance for non-ferrous metals price forecasting.
TL;DR: In this article, a multi-view spectral clustering model is proposed to fuse different views into one graph, and the fusion graph approximates the original graph of each individual view but maintains an explicit cluster structure.
Abstract: A panoply of multi-view clustering algorithms has been developed to deal with prevalent multi-view data. Among them, spectral clustering-based methods have drawn much attention and demonstrated promising results recently. Despite progress, there are still two fundamental questions that stay unanswered to date. First, how to fuse different views into one graph. More often than not, the similarities between samples may be manifested differently by different views. Many existing algorithms either simply take the average of multiple views or just learn a common graph. These simple approaches fail to consider the flexible local manifold structures of all views. Hence, the rich heterogeneous information is not fully exploited. Second, how to learn the explicit cluster structure. Most existing methods do not pay attention to the quality of the graphs and perform graph learning and spectral clustering separately. Those unreliable graphs might lead to suboptimal clustering results. To fill these gaps, in this paper, we propose a novel multi-view spectral clustering model which performs graph fusion and spectral clustering simultaneously. The fusion graph approximates the original graph of each individual view but maintains an explicit cluster structure. Experiments on four widely used data sets confirm the superiority of the proposed method.
TL;DR: A novel hybrid algorithm called HSGWO-MSOS is proposed by combining simplified grey wolf optimizer (SGWO) and modified symbiotic organisms search (MSOS) and its performance is superior to the GWO, SOS and SA algorithm.
Abstract: Unmanned aerial vehicle (UAV) path planning problem is an important component of UAV mission planning system, which needs to obtain optimal route in the complicated field. To solve this problem, a novel hybrid algorithm called HSGWO-MSOS is proposed by combining simplified grey wolf optimizer (SGWO) and modified symbiotic organisms search (MSOS). In the proposed algorithm, the exploration and exploitation abilities are combined efficiently. The phase of the GWO algorithm is simplified to accelerate the convergence rate and retain the exploration ability of the population. The commensalism phase of the SOS algorithm is modified and synthesized with the GWO to improve the exploitation ability. In addition, the convergence analysis of the proposed HSGWO-MSOS algorithm is presented based on the method of linear difference equation. The cubic B-spline curve is used to smooth the generated flight route and make the planning path be suitable for the UAV. The simulation experimental results show that the HSGWO-MSOS algorithm can acquire a feasible and effective route successfully, and its performance is superior to the GWO, SOS and SA algorithm.
TL;DR: The proposed Sinusoidal-BDA outperforms the comparable feature selection algorithms and the proposed updating mechanism has a high impact on the algorithm performance when tackling Feature Selection (FS) problems.
Abstract: Dragonfly Algorithm (DA) is a recent swarm-based optimization method that imitates the hunting and migration mechanisms of idealized dragonflies. Recently, a binary DA (BDA) has been proposed. During the algorithm iterative process, the BDA updates its five main coefficients using random values. This updating mechanism can be improved to utilize the survival-of-the-fittest principle by adopting different functions such as linear, quadratic, and sinusoidal. In this paper, a novel BDA is proposed. The algorithm uses different strategies to update the values of its five main coefficients to tackle Feature Selection (FS) problems. Three versions of BDA have been proposed and compared against the original DA. The proposed algorithms are Linear-BDA, Quadratic-BDA, and Sinusoidal-BDA. The algorithms are evaluated using 18 well-known datasets. Thereafter, they are compared in terms of classification accuracy, the number of selected features, and fitness value. The results show that Sinusoidal-BDA outperforms other proposed methods in almost all datasets. Furthermore, Sinusoidal-BDA exceeds three swarm-based methods in all the datasets in terms of classification accuracy and it excels in most datasets when compared in terms of the fitness function value. In a nutshell, the proposed Sinusoidal-BDA outperforms the comparable feature selection algorithms and the proposed updating mechanism has a high impact on the algorithm performance when tackling FS problems.
TL;DR: An end-to-end Regional-Asymmetric Convolutional Neural Network (RACNN) for emotion recognition, which consists of temporal, regional and asymmetric feature extractors, which can capture the discriminative information between left and right hemispheres of the brain.
Abstract: Emotion recognition based on electroencephalography (EEG) is of great important in the field of Human–Computer Interaction (HCI), which has received extensive attention in recent years. Most traditional methods focus on extracting features in time domain and frequency domain. The spatial information from adjacent channels and symmetric channels is often ignored. To better learn spatial representation, in this paper, we propose an end-to-end Regional-Asymmetric Convolutional Neural Network (RACNN) for emotion recognition, which consists of temporal, regional and asymmetric feature extractors. Specifically, continuous 1D convolution layers are employed in temporal feature extractor to learn time–frequency representations. Then, regional feature extractor consists of two 2D convolution layers to capture regional information among physically adjacent channels. Meanwhile, we propose an Asymmetric Differential Layer (ADL) in asymmetric feature extractor by taking the asymmetry property of emotion responses into account, which can capture the discriminative information between left and right hemispheres of the brain. To evaluate our model, we conduct extensive experiments on two publicly available datasets, i.e., DEAP and DREAMER. The proposed model can obtain recognition accuracies over 95% for valence and arousal classification tasks on both datasets, significantly outperforming the state-of-the-art methods.
TL;DR: This paper proposes improving robustness, accuracy and reliability of the detection of small objects handled similarly using binarization techniques, and creates a database considering six objects: pistol, knife, smartphone, bill, purse and card.
Abstract: The capability of distinguishing between small objects when manipulated with hand is essential in many fields, especially in video surveillance. To date, the recognition of such objects in images using Convolutional Neural Networks (CNNs) remains a challenge. In this paper, we propose improving robustness, accuracy and reliability of the detection of small objects handled similarly using binarization techniques. We propose improving their detection in videos using a two level methodology based on deep learning, called Object Detection with Binary Classifiers. The first level selects the candidate regions from the input frame and the second level applies a binarization technique based on a CNN-classifier with One-Versus-All or One-Versus-One. In particular, we focus on the video surveillance problem of detecting weapons and objects that can be confused with a handgun or a knife when manipulated with hand. We create a database considering six objects: pistol, knife, smartphone, bill, purse and card. The experimental study shows that the proposed methodology reduces the number of false positives with respect to the baseline multi-class detection model.
TL;DR: By constructing dummy queries that have similar feature distributions but unrelated topics with user queries, the privacy behind users’ textual queries can be effectively protected, under the precondition of not compromising the accuracy and usability of text retrieval.
Abstract: Text retrieval enables people to efficiently obtain the desired data from massive text data, so has become one of the most popular services in information retrieval community. However, while providing great convenience for users, text retrieval results in a serious issue on user privacy. In this paper, we propose a dummy-based approach for text retrieval privacy protection. Its basic idea is to use well-designed dummy queries to cover up user queries and thus protect user privacy. First, we present a client-based system framework for the protection of user privacy, which requires no change to the existing algorithm of text retrieval, and no compromise to the accuracy of text retrieval. Second, we define a user privacy model to formulate the requirements that ideal dummy queries should meet, i.e., (1) having highly similar feature distributions with user queries, and (2) effectively reducing the significance of user query topics. Third, by means of the knowledge derived from Wikipedia, we present an implementation algorithm to construct a group of ideal dummy queries that can well meet the privacy model. Finally, we demonstrate the effectiveness of our approach by theoretical analysis and experimental evaluation. The results show that by constructing dummy queries that have similar feature distributions but unrelated topics with user queries, the privacy behind users’ textual queries can be effectively protected, under the precondition of not compromising the accuracy and usability of text retrieval.
TL;DR: Comparative evaluation of algorithmic performance with the state-of-the-art schemes manifest that the seizure detection scheme proposed herein outperforms competing algorithms in terms of accuracy, sensitivity, specificity, and Cohen’s Kappa coefficient.
Abstract: Background: Epileptic seizure detection is traditionally performed by visual observation of Electroencephalogram (EEG) signals. Owing to its onerous and time-consuming nature, seizure detection based on visual inspection hinders epilepsy diagnosis, monitoring, and large-scale data analysis in epilepsy research. So, there is a dire need of an automatic seizure detection scheme. Method: An automated scheme for epileptic seizure identification is developed in this study. Here we utilize a signal processing technique, namely-complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) for epileptic seizure identification. First, we decompose segments of EEG signals into intrinsic mode functions by CEEMDAN. The mode functions are then modeled by normal inverse Gaussian (NIG) pdf parameters. In this work, NIG modeling is employed in conjunction with CEEMDAN for epileptic seizure detection for the first time. The efficacy of the NIG parameters in the CEEMDAN domain is demonstrated by intuitive, graphical, and statistical analyses. Adaptive Boosting, an eminent ensemble learning based classification model, is implemented to perform classification. Results: Experimental outcomes suggest that the algorithmic performance of the proposed scheme is promising in all the cases of clinical significance. Comparative evaluation of algorithmic performance with the state-of-the-art schemes manifest that the seizure detection scheme proposed herein outperforms competing algorithms in terms of accuracy, sensitivity, specificity, and Cohen’s Kappa coefficient. Conclusions: Upon its implementation in clinical practice, the proposed seizure detection scheme will eliminate the onus of medical professionals and expedite epilepsy research and diagnosis.
TL;DR: A learning path recommendation model is designed for satisfying different learning needs based on the multidimensional knowledge graph framework, which can generate and recommend customized learning paths according to the e-learner’s target learning object.
Abstract: E-learners face a large amount of fragmented learning content during e-learning. How to extract and organize this learning content is the key to achieving the established learning target, especially for non-experts. Reasonably arranging the order of the learning objects to generate a well-defined learning path can help the e-learner complete the learning target efficiently and systematically. Currently, knowledge-graph-based learning path recommendation algorithms are attracting the attention of researchers in this field. However, these methods only connect learning objects using single relationships, which cannot generate diverse learning paths to satisfy different learning needs in practice. To overcome this challenge, this paper proposes a learning path recommendation model based on a multidimensional knowledge graph framework. The main contributions of this paper are as follows. Firstly, we have designed a multidimensional knowledge graph framework that separately stores learning objects organized in several classes. Then, we have proposed six main semantic relationships between learning objects in the knowledge graph. Secondly, a learning path recommendation model is designed for satisfying different learning needs based on the multidimensional knowledge graph framework, which can generate and recommend customized learning paths according to the e-learner’s target learning object. The experiment results indicate that the proposed model can generate and recommend qualified personalized learning paths to improve the learning experiences of e-learners.
TL;DR: This model firstly introduces bidirectional attention mechanism with position encoding to model aspect-specific representations between each aspect and its context words, then employs GCN over the attention mechanism to capture the sentiment dependencies between different aspects in one sentence.
Abstract: Aspect-level sentiment classification aims to distinguish the sentiment polarities over one or more aspect terms in a sentence. Existing approaches mostly model different aspects in one sentence independently, which ignore the sentiment dependencies between different aspects. However, such dependency information between different aspects can bring additional valuable information for aspect-level sentiment classification. In this paper, we propose a novel aspect-level sentiment classification model based on graph convolutional networks (GCN) which can effectively capture the sentiment dependencies between multi-aspects in one sentence. Our model firstly introduces bidirectional attention mechanism with position encoding to model aspect-specific representations between each aspect and its context words, then employs GCN over the attention mechanism to capture the sentiment dependencies between different aspects in one sentence. The proposed approach is evaluated on the SemEval 2014 datasets. Experiments show that our model outperforms the state-of-the-art methods. We also conduct experiments to evaluate the effectiveness of GCN module, which indicates that the dependencies between different aspects are highly helpful in aspect-level sentiment classification 1 .
TL;DR: A new Syntax- and Knowledge-based Graph Convolutional Network (SK-GCN) model is proposed for aspect-level sentiment classification, which leverages the syntactic dependency tree and commonsense knowledge via GCN to enhance the representation of the sentence toward the given aspect.
Abstract: Aspect-level sentiment classification is a fundamental subtask of fine-grained sentiment analysis. The syntactic information and commonsense knowledge are important and useful for aspect-level sentiment classification, while only a limited number of studies have explored to incorporate them via flexible graph convolutional neural networks (GCN) for this task. In this paper, we propose a new Syntax- and Knowledge-based Graph Convolutional Network (SK-GCN) model for aspect-level sentiment classification, which leverages the syntactic dependency tree and commonsense knowledge via GCN. In particular, to enhance the representation of the sentence toward the given aspect, we develop two strategies to model the syntactic dependency tree and commonsense knowledge graph, namely SK-GCN 1 and SK-GCN 2 respectively. SK-GCN 1 models the dependency tree and knowledge graph via Syntax-based GCN (S-GCN) and Knowledge-based GCN (K-GCN) independently, and SK-GCN 2 models them jointly. We also apply pre-trained BERT to this task and obtain new state-of-the-art results. Extensive experiments on five benchmark datasets demonstrate that our approach can effectively improve the performance of aspect-level sentiment classification compared with the state-of-the-art methods.