Search

In the previous image segmentation methods based on affinity graph learning, it is difficult to obtain clear contour boundaries in the process of image preprocessing, over-segmentation leads to the separation of local regions and most traditional affinity graph learning methods cluster and segment with a single feature of natural images, ignoring the associative characteristics of multiple types of features and cannot effectively utilize the useful information of multiple features of natural images. Aiming at such problems, this paper proposes a tensor-based multi-feature affinity graph learning method for natural image segmentation (TRMFAL). First, the adaptive morphological rebuilding watershed transform is applied to original natural image, the obtained superpixel image contains a lot of boundary contour information, and the fusion of local regions is relatively good; secondly, extract multi-class features from the superpixel blocks in the superpixel image, effectively utilize the features of different characteristics, and integrate them into a multi-feature data matrix according to the corresponding rules; Then, a tensor-based multi-feature affinity graph learning algorithm is proposed, in which tensor is introduced in the algorithm to effectively obtain the higher-order information of image data, and use the projection matrix to embed the original data in the low-dimensional space to decrease the dimension, while minimizing the residual error of each view feature and assign appropriate weights according to the importance of the feature information; finally, use spectral clustering performs clustering and segmentation on affinity graphs to obtain final clustering results and segmented images. In addition, an optimization iterative algorithm based on the alternating multiplier direction method is designed, which effectively solves the problem of solving the TRMFAL model. Sufficient experimental comparisons are carried out on multiple public datasets, and the results prove that the proposed method achieves the optimal clustering performance and segmentation effect. [ABSTRACT FROM AUTHOR]

Graph-based semi-supervised learning (GSSL) has received more and more attention due to its efficiency and accuracy. Label propagation is a critical step in GSSL that propagates label information to unlabeled data through the structure of graph. However, the traditional label propagation algorithms treat all unlabeled samples as equivalent and blindly propagate label information to all neighbors without considering their reliabilities. In this case, some unreliable samples may mislead the process of label propagation, thus greatly reducing the accuracy of classification. In order to solve this problem, this paper proposes a novel label propagation algorithm called node influence-based label propagation (NILP). Based on the structure of graph, the NILP algorithm measures the influences of nodes by calculating their degrees and local densities. In the process of label propagation, the label information is preferentially transmitted to the influential neighbors to control the propagation sequence and prevent wrong propagation. Moreover, our algorithm improves the transition matrix by integrating label information and feature information. The experimental results on both synthetic and real-world benchmark datasets show that the proposed method is superior to some existing label propagation algorithms. Especially when the number of labeled samples is very small, the advantage of NILP algorithm is more obvious. [ABSTRACT FROM AUTHOR]

Selecting relevant feature subsets is vital in machine learning, and multiclass feature selection is harder to perform since most classifications are binary. The feature selection problem aims at reducing the feature set dimension while maintaining the performance model accuracy. Datasets can be classified using various methods. Nevertheless, metaheuristic algorithms attract substantial attention to solving different problems in optimization. For this reason, this paper presents a systematic survey of literature for solving multiclass feature selection problems utilizing metaheuristic algorithms that can assist classifiers selects optima or near optima features faster and more accurately. Metaheuristic algorithms have also been presented in four primary behavior-based categories, i.e., evolutionary-based, swarm-intelligence-based, physics-based, and human-based, even though some literature works presented more categorization. Further, lists of metaheuristic algorithms were introduced in the categories mentioned. In finding the solution to issues related to multiclass feature selection, only articles on metaheuristic algorithms used for multiclass feature selection problems from the year 2000 to 2022 were reviewed about their different categories and detailed descriptions. We considered some application areas for some of the metaheuristic algorithms applied for multiclass feature selection with their variations. Popular multiclass classifiers for feature selection were also examined. Moreover, we also presented the challenges of metaheuristic algorithms for feature selection, and we identified gaps for further research studies. [ABSTRACT FROM AUTHOR]

Facial expression recognition (FER) in the wild is very challenging due to occlusion, posture, illumination, and other uncontrolled factors. Learning discriminant features for FER using Convolutional Neural Networks is a momentous task for the significant class imbalance, wrong labels, inter-class similarities, and intra-class variations. The traditional method utilizes the Cross entropy loss function to optimize the convolutional network to obtain discriminative features for classification. However, this loss function cannot effectively solve the above problems in practice and cannot contribute to obtaining highly discriminant facial features for further analysis. Center loss improves the learning efficiency by reducing the intra-class distance of similar expressions, while the improvement of inter-class similarity, class imbalance, and generalization is insufficient. In this paper, we propose a lightweight Effective Attention Feature Reconstruction loss (EAFR loss), which can further optimize the feature space and enhance the discriminability of expression. The loss model is composed of the Focal Smoothing loss (FS loss) and the Aggregation-Separation loss (AS loss). Firstly, the FS loss can improve the poor recognition performance caused by imbalanced classes and prevent paranoid knowledge learning behaviors. Meanwhile, AS loss further accurately condenses the intra-class expression features and expands the inter-class distance, which is achieved by using progressive stage max-pooling channel and position attention mechanism and lightweight asymmetric autoencoder model for feature reconstruction. Finally, the EAFR loss joins the above two loss functions to more comprehensively solve the above typical problems for FER in the wild. We validate the proposed loss function on three most commonly used large-scale wild expression datasets (RAF-DB, FERPlus, and AffectNet), and the results show that our model achieves superior performance to several state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The existing view-based 3D object classification and recognition methods ignore the inherent hierarchical correlation and distinguishability of views, making it difficult to further improve the classification accuracy. In order to solve this problem, this paper proposes an end-to-end multi-view dual attention network framework for high-precision recognition of 3D objects. On one hand, we obtain three feature layers of query, key, and value through the convolution layer. The spatial attention matrix is generated by the key-value pairs of query and key, and each feature in the value of the original feature space branch is assigned different importance, which clearly captures the prominent detail features in the view, generates the view space shape descriptor, and focuses on the detail part of the view with the feature of category discrimination. On the other hand, a channel attention vector is obtained by compressing the channel information in different views, and the attention weight of each view feature is scaled to find the correlation between the target views and focus on the view with important features in all views. Integrating the two feature descriptors together to generate global shape descriptors of the 3D model, which has a stronger response to the distinguishing features of the object model and can be used for high-precision 3D object recognition. The proposed method achieves an overall accuracy of 96.6% and an average accuracy of 95.5% on the open-source ModelNet40 dataset, compiled by Princeton University when using Resnet50 as the basic CNN model. Compared with the existing deep learning methods, the experimental results demonstrate that the proposed method achieves state-of-the-art performance in the 3D object classification accuracy. [ABSTRACT FROM AUTHOR]

Speech diagnosis of Parkinson's disease (PD) as a non-invasive and simple diagnosis method is particularly worth exploring. However, the number of samples of speech-based PD is relatively small, and there exist discrepancies in the distribution between subjects. In order to solve the two problems, a novel unsupervised two-step sparse transfer learning is proposed in this paper to tackle with PD speech diagnosis. In the first step, convolution sparse coding with the coordinate selection of samples and features is designed to learn speech structure from the source domain to replenish sample information of the target domain. In the second step, joint local structure distribution alignment is designed to maintain the neighbor relationship between the respective samples of the training set and test set, and reduce the distribution difference between the two domains at the same time. Two representative public PD speech datasets and one real-world PD speech dataset were exploited to verify the proposed method on PD speech diagnosis. Experimental results demonstrate that each step of the proposed method has a positive effect on the PD speech classification results, and it also delivers superior performance over the existing relative methods. [ABSTRACT FROM AUTHOR]

In recent years, researchers have proposed many methods to solve the problem of obstacle detection. However, computer vision-based vehicle detection and recognition technology is still not mature enough. This research combines machine learning technology to construct a traffic object recognition system and applies innovative technology to the computer vision recognition system to construct an automatic identification system suitable for current traffic demand and improve the stability of the traffic system. Moreover, this study uses a combination of a monocular camera and a binocular camera to sense the traffic environment and obtain vehicle position and velocity information. In addition, this study is based on the binocular stereo camera to find the obstacle space and obtain the obstacle relative to the position and speed of the vehicle and combine the obstacle space information to optimize the obstacle frame of the target vehicle. Through experimental research and analysis, it can be seen that the algorithm proposed in this study has certain recognition effect and can be applied to traffic object recognition. [ABSTRACT FROM AUTHOR]

This paper considers a scheduling problem with general job-dependent learning curves and controllable processing times on a single machine. The objective is to determine the optimal compressions of the processing times and the optimal sequence of jobs so as to minimize some total cost functions, which consist of regular and non-regular functions and the processing time compressions. It shows that the problem can be solved by an assignment problem and thus can be solved in polynomial time. Some extensions of the problem are also given. [ABSTRACT FROM AUTHOR]

Learning partial differential equations (LPDEs) from training data for particular tasks has been successfully applied to many image processing problems. In this paper, we propose a more effective LPDEs model for vector-valued image tasks. The PDEs are also formulated as a linear combination of fundamental differential invariants, but have several distinctions. First, we simplify the current LPDEs system by omitting a PDE which works as an indicate function in current ones. Second, instead of using L2-norm, we use the L1-norm to regularize the coefficients with respect to the fundamental differential invariants. Third, as the objective function is not smooth, we resort to the alternating direction method to optimize it. We illustrate the properties of our LPDEs system by several examples in denoising and demosaicking of RGB color images. The experiments demonstrate the advantage of the proposed method over other PDE-based methods. [ABSTRACT FROM AUTHOR]

In this paper, a novel distributed elastic random weights network (DERWN) is proposed to achieve robust surface reconstruction from highly noisy point clouds sampled from real surface. The designed elastic regularization with l 1 and l 2 penalty items makes the network more resilient to noise and effectively capture the intrinsic shape of surface. Sparsity constraints of output weight vectors and threshold-based nodes removal are conducive to determining appropriate number of hidden nodes of network and optimizing the distribution of hidden nodes. The distributed optimization manner in DERWN on the basis of alternating direction method of multipliers solves the problem that traditional RWN learning algorithm suffers from the limitation of memory with large-scale data. The proposed DERWN achieves a solution to global problem by solving local subproblems coordinately. Experimental results show that the proposed DERWN algorithm can robustly reconstruct the unknown surface in case of highly noisy data with satisfying accuracy and smoothness. [ABSTRACT FROM AUTHOR]

Structured output prediction aims to learn a predictor to predict a structured output from a input data vector . The structured outputs include vector, tree, sequence, etc. We usually assume that we have a training set of input-output pairs to train the predictor. However, in many real-world applications, it is difficult to obtain the output for a input, and thus for many training input data points, the structured outputs are missing. In this paper, we discuss how to learn from a training set composed of some input-output pairs and some input data points without outputs. This problem is called semi-supervised structured output prediction. We propose a novel method for this problem by constructing a nearest neighbor graph from the input space to present the manifold structure and use it to regularize the structured output space directly. We define a slack structured output for each training data point and propose to predict it by learning a structured output predictor. The learning of both slack structured outputs and the predictor are unified within one single minimization problem. In this problem, we propose to minimize the structured loss between the slack structured outputs of neighboring data points and the prediction error measured by the structured loss. The problem is optimized by an iterative algorithm. Experiment results over three benchmark data sets show its advantage. [ABSTRACT FROM AUTHOR]

In this paper, we study the problem of tag completion. Given an image and a set of tags, only a few of the tags are known to be associated with this image or not, and the problem is to predict whether the other tags are associated with the image. To solve this problem, we propose to learn a tag scoring vector for each image and use it to predict the associated tags of the image. To learn the tag scoring vector, we use the method of local linear learning. A local linear function is used in the neighborhood of each image to predict the tag scoring vectors of its neighboring images. We construct a unified objective function for the learning of both tag scoring vectors and local linear function parameters. In this objective, we impose the learned tag scoring vectors to be consistent with the known associations to the tags of each image and also minimize the prediction error of each local linear function, while reducing the complexity of each local function. The objective function is optimized by an alternate optimization strategy and gradient descent methods in an iterative algorithm. We compare the proposed algorithm against different state-of-the-art tag completion methods, and the results show its advantages. [ABSTRACT FROM AUTHOR]

In this paper, a new approach for solving system of fully fuzzy nonlinear equations based on fuzzy neural network is presented. This method can also lead to improve numerical methods. In this work, an architecture of fuzzy neural networks is also proposed to find a fuzzy root of a system of fuzzy nonlinear equations (if exists) by introducing a learning algorithm. We propose a learning algorithm from the cost function for adjusting of fuzzy weights. Finally, we illustrate our approach by numerical examples. [ABSTRACT FROM AUTHOR]

It is now widely accepted that memristive devices are promising candidates for the emulation of the behavior of biological synapses in neuromorphic systems. This is mainly because of the fact that like the strength of synapse, memristance of the memristive device can be tuned actively for example by the application of voltage or current. In addition, it is also possible to fabricate high density of memristive devices through the nano-crossbar structures. In this paper, we will show that there are some problems associated with memristive devices, which are playing the role of biological synapses. For example, we show that the variation rate of the memristance depends completely on the initial state of the device, and therefore, it can change significantly with time during the learning phase. This phenomenon can degrade the performance of learning methods like spike timing-dependent plasticity and cause the corresponding neuromorphic systems to become unstable. We also illustrate that using two serially connected memristive devices with different polarities as a synapse can somewhat fix the aforementioned problem. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel data representation-classification model learning algorithm. The model is a convolutional neural network (CNN), and we learn its parameters to achieve A-optimality. The input multi-instance data are represented by a CNN model, and then classified by a linear classification model. The A-optimality of a classification model is measured by the trace of the covariance matrix of the model parameter vector. To achieve the A-optimality of the CNN model, we minimize the classification errors and a regularization term to present the classification model parameter as a function of the CNN filter parameters, and minimize its trace of the covariance matrix. We show that the minimization problem can be solved easily by transferring it to another coupled minimization problem. In the experiments over benchmark data sets of molecular, image, and seismic waveform, we show the advantages of the proposed A-optimal CNN model. [ABSTRACT FROM AUTHOR]

In this paper, a new Chinese character recognition (CCR) approach is proposed based on the fuzzy clustering analysis theory. Chinese characters (CCs) have various similar radicals and stroke components, which make it difficult to recognize features in the CCR process. At the same time, the recognition accuracy and the efficiency are lower when the objects to be recognized are complex. In order to solve these problems, a fuzzy clustering analysis method is introduced to enhance the computing efficiency. At first, the CCs including learning samples and testing samples are transformed into binarization templates in the form of matrixes. Then, the minimum distance algorithm is applied to calculate 'distances' between the testing sample templates and the learning sample templates. At last, the character recognition can be achieved by searching the minimum distance from the results. The experiment results of the CCR process can prove the effectiveness and accuracy of the new method. [ABSTRACT FROM AUTHOR]

Traditional projection twin support vector machines (SVMs) ignore the differences between the categories when establishing the objective functions, which would lessen their generalization performance. To solve the issue, an improved projection twin SVM (abbreviated as IPTSVM) is proposed in this paper, which aims to find two projected directions via a single quadratic programming problem. In their respective subspace, the projected sample points belonging to each category are far from those of the other class. Meanwhile, to enhance the performance, the recursive arithmetic seeks for more than one projection directions for each class. Besides, an effective clipping dual coordinate descent model is adopted to solve the dual problem to accelerate the training process. The linear IPTSVM model could be changed into the nonlinear model by using the kernel metric. Furthermore, the multi-label version of IPTSVM model is developed to deal with the multi-label learning problems. Experiments on a set of public datasets show that the IPTSVM model has significant advantages over the other models in terms of generalization performance. [ABSTRACT FROM AUTHOR]

Transfer learning is a problem defined over two domains. These two domains share the same feature space and class label space, but have significantly different distributions. One domain has sufficient labels, named as source domain, and the other domain has few labels, named as target domain. The problem is to learn an effective classifier for the target domain. In this paper, we propose a novel transfer learning method for this problem by learning a partially shared classifier for the target domain, and weighting the source domain data points. We learn some shared subspaces for both the data points of the two domains, and a shared classifier in the shared subspaces. We hope that in the shared subspaces, the distributions of two domain can match each other well, and to match the distributions, we weight the source domain data points with different weighting factors. Moreover, we adapt the shared classifier to each domain by learning different adaptation functions. To learn the subspace transformation matrices, the classifier parameters, and the adaptation parameters, we build an objective function with weighted classification errors, parameter regularization, local reconstruction regularization, and distribution matching. This objective function is minimized by an iterative algorithm. Experiments show its effectiveness over benchmark data sets, including travel destination review data set, face expression data set, spam email data set. [ABSTRACT FROM AUTHOR]

In this paper, we invest the domain transfer learning problem with multi-instance data. We assume we already have a well-trained multi-instance dictionary and its corresponding classifier from the source domain, which can be used to represent and classify the bags. But it cannot be directly used to the target domain. Thus we propose to adapt them to the target domain by adding an adaptive term to the source domain classifier. The adaptive function is a linear function based on a domain transfer multi-instance dictionary. Given a target domain bag, we first map it to a bag-level feature space using the domain transfer dictionary and then apply a linear adaptive function to its bag-level feature vector. To learn the domain transfer dictionary and the adaptive function parameter, we simultaneously minimize the average classification error of the target domain classifier over the target domain training set, and the complexities of both the adaptive function parameter and the domain transfer dictionary. The minimization problem is solved by an iterative algorithm which updates the dictionary and the function parameter alternately. Experiments over several benchmark data sets show the advantage of the proposed method over existing state-of-the-art domain transfer multi-instance learning methods. [ABSTRACT FROM AUTHOR]

Multiple kernel learning (MKL) approach has been proposed for kernel methods and has shown high performance for solving some real-world applications. It consists on learning the optimal kernel from one layer of multiple predefined kernels. Unfortunately, this approach is not rich enough to solve relatively complex problems. With the emergence and the success of the deep learning concept, multilayer of multiple kernel learning (MLMKL) methods were inspired by the idea of deep architecture. They are introduced in order to improve the conventional MKL methods. Such architectures tend to learn deep kernel machines by exploring the combinations of multiple kernels in a multilayer structure. However, existing MLMKL methods often have trouble with the optimization of the network for two or more layers. Additionally, they do not always outperform the simplest method of combining multiple kernels (i.e., MKL). In order to improve the effectiveness of MKL approaches, we introduce, in this paper, a novel backpropagation MLMKL framework. Specifically, we propose to optimize the network over an adaptive backpropagation algorithm. We use the gradient ascent method instead of dual objective function, or the estimation of the leave-one-out error. We test our proposed method through a large set of experiments on a variety of benchmark data sets. We have successfully optimized the system over many layers. Empirical results over an extensive set of experiments show that our algorithm achieves high performance compared to the traditional MKL approach and existing MLMKL methods. [ABSTRACT FROM AUTHOR]

In this paper, we treat optimization problems as a kind of reinforcement learning problems regarding an optimization procedure for searching an optimal solution as a reinforcement learning procedure for finding the best policy to maximize the expected rewards. This viewpoint motivated us to propose a Q-learning-based swarm optimization (QSO) algorithm. The proposed QSO algorithm is a population-based optimization algorithm which integrates the essential properties of Q-learning and particle swarm optimization. The optimization procedure of the QSO algorithm proceeds as each individual imitates the behavior of the global best one in the swarm. The best individual is chosen based on its accumulated performance instead of its momentary performance at each evaluation. Two data sets including a set of benchmark functions and a real-world problem-the economic dispatch (ED) problem for power systems-were used to test the performance of the proposed QSO algorithm. The simulation results on the benchmark functions show that the proposed QSO algorithm is comparable to or even outperforms several existing optimization algorithms. As for the ED problem, the proposed QSO algorithm has found solutions better than all previously found solutions. [ABSTRACT FROM AUTHOR]

In order to overcome the disadvantage of the traditional algorithm for SLFN (single-hidden layer feedforward neural network), an improved algorithm for SLFN, called extreme learning machine (ELM), is proposed by Huang et al. However, ELM is sensitive to the neuron number in hidden layer and its selection is a difficult-to-solve problem. In this paper, a self-adaptive mechanism is introduced into the ELM. Herein, a new variant of ELM, called self-adaptive extreme learning machine (SaELM), is proposed. SaELM is a self-adaptive learning algorithm that can always select the best neuron number in hidden layer to form the neural networks. There is no need to adjust any parameters in the training process. In order to prove the performance of the SaELM, it is used to solve the Italian wine and iris classification problems. Through the comparisons between SaELM and the traditional back propagation, basic ELM and general regression neural network, the results have proven that SaELM has a faster learning speed and better generalization performance when solving the classification problem. [ABSTRACT FROM AUTHOR]

This paper proposes a probably approximately correct (PAC) algorithm that directly utilizes online data efficiently to solve the optimal control problem of continuous deterministic systems without system parameters for the first time. The dependence on some specific approximation structures is crucial to limit the wide application of online reinforcement learning (RL) algorithms. We utilize the online data directly with the kd-tree technique to remove this limitation. Moreover, we design the algorithm in the PAC principle. Complete theoretical proofs are presented, and three examples are simulated to verify its good performance. It draws the conclusion that the proposed RL algorithm specifies the maximum running time to reach a near-optimal control policy with only online data. [ABSTRACT FROM AUTHOR]

This paper proposes a subspace learning method, named as sparse tensor canonical correlation analysis (ST-CCA), for color face recognition. A sample image is formalized as high-order tensors to preserve the inherent structure of the color face images. We utilize sparse canonical correlation analysis (SCCA) to choose gene. For each pair of tensors, SCCA generates the sparse loadings alternately, which is helpful for choosing significant variables to reduce dimensions and eliminate the redundancies of tensors. We use the elastic net as constraint condition to attack the collinearity problem by decorrelating and selecting the sufficient variables irrespective of the limited dimensions. Furthermore, ST-CCA gains stable recognition rates because the alternating least square solution converges. ST-CCA is convex with different initials of the projection matrices. Experimental results on AR face database and LFW face database show the superior performance of our method over the state-of-the-art ones. [ABSTRACT FROM AUTHOR]

The harmony search (HS) method is a popular meta-heuristic optimization algorithm, which has been extensively employed to handle various engineering problems. However, it sometimes fails to offer a satisfactory convergence performance under certain circumstances. In this paper, we propose and study a hybrid HS approach, HS-PBIL, by merging the HS together with the population-based incremental learning (PBIL). Numerical simulations demonstrate that our HS-PBIL is well capable of outperforming the regular HS method in dealing with nonlinear function optimization and a practical wind generator optimization problem. [ABSTRACT FROM AUTHOR]

Electrical Impedance Tomography (EIT) is a non-invasive technique used to obtain the electrical internal conductivity distribution from the interior of bodies. This is a promising method from the manufacturing viewpoint, since it could be used to estimate different physical inner body properties during the production of goods. Nevertheless, this technique requires dealing with an inverse problem that makes its usage in real-time processes challenging. Recently, Machine Learning techniques have been proposed to solve the inverse problem accurately. However, the majority of prior research is focused on qualitative results, and they typically lack a systematic methodology to determine the optimal hyperparameters appropriately. This work presents a systematic comparison of six popular Machine Learning algorithms: Artificial Neural Network, Random Forest, K-Nearest Neighbors, Elastic Net, Ada Boost, and Gradient Boosting. Particularly, the last two algorithms were based on decision tree learners. Furthermore, we studied the relationship between model performance and different EIT configurations. Specifically, we analyzed whether the measurement pattern and the number of used electrodes could increase the model performance. Experiments revealed that tree-based models present high performance, even better than Neural Networks, the most widely-used Machine Learning model to deal with EIT. Experiments also showed a model performance improvement when the EIT configuration was optimized. Most favorable metrics were attained using the tree-based Gradient Boosting model with a combination of both adjacent and mono measurement patterns as well as with 32 electrodes deployed during the tomographic process. With this particular setting, we achieved an accuracy of 99.14% detecting internal artifacts and a Root Mean Square Error of 4.75 predicting internal conductivity distributions. [ABSTRACT FROM AUTHOR]

Big data analytics and machine learning applications are often used to detect and classify anomalous behaviour in telecom network measurement data. The accuracy of findings during the analysis phase greatly depends on the quality of the training dataset. If the training dataset contains data from network elements (NEs) with high number of failures and high failure rates, such behaviour will be assumed as normal. As a result, the analysis phase will fail to detect NEs with such behaviour. Effective post-processing techniques are needed to analyse the anomalies, to determine the different kinds of anomalies, as well as their relevance in real-world scenarios. Manual post-processing of anomalies detected in an Anomaly Detection experiment is a cumbersome task, and ways to automate this process are not much researched upon. There exists no universally accepted method for effective classification of anomalous behaviour. High failure ratios have traditionally been considered as signs of faults in NEs. Operators use well-known key performance indicators (KPIs) such as drop call ratio and handover failure ratio to identify misbehaving NEs. The main problem with these KPIs based on failure ratios is their unstable nature. This paper proposes a method of measuring the significance of failures. The usage of this method is proposed in two stages of anomaly detection: training set filtering (pre-processing stage) and classification of anomalies (post-processing stage) using an automated process. [ABSTRACT FROM AUTHOR]

The control of surface defects is of critical importance in manufacturing quality control systems. Today, automatic defects detection using imaging and deep learning algorithms has produced more successful results than manual inspections. Thanks to these automatic applications, manufacturing systems will increase the production quality, and thus financial losses will be prevented. However, since the appearance and dimensions of the defects on the surface are very variable, automatic surface defect detection is a complex problem. In this study, multi-dimensional feature extraction-based deep encoder–decoder network (MFE-DEDNet) network developed to solve such problems. An effective encoder–decoder model with lower parameters compared to the state-of-the-art methods is developed using the depthwise separable convolutions (DSC) layers in the proposed model. In addition, the 3D spectral and spatial features extract (3DFE) module of the proposed model is developed to extract deep spectral and spatial features, as well as deep semantic features. During the combination of these features, the multi-input attention gate (MIAG) module is used so that important details are not lost. As a result, the proposed MFE-DEDNet model based on these structures enabled the extraction of powerful and effective features for defect detection in surface datasets containing few images. In experimental studies, MVTec and MT datasets were used to evaluate the performance of the MFE-DEDNet. The experimental results achieved 80.01% and 56% mean intersection-over-union (mIoU) scores for the MT and MVTec datasets, respectively. In these results, it was observed that the proposed model produced higher success compared to other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Motor fault diagnosis is critical to predictive maintenance of electrical motor condition monitoring, whereas the conventional motor fault diagnosis method cannot effectively diagnose conditions caused by motors' complex structure, non-stationary signals and mechanical big data. To solve the mentioned problems and enhance the fault diagnostic accuracy and generalization performance under different actual motor conditions, this study proposes an efficient, noise-resistant, end-to-end deep learning algorithm based on a novel capsule network with gate-structure dilated convolutions (GDCCN) for motor fault diagnosis; such algorithm is subtly incorporated with the input gate structure of long short-term memory network (LSTM), the dilated convolutions, as well as the capsule network. In the GDCCN model, the raw vibration signals are directly fed into the input gate structure of LSTM, which are employed to effectively remove noise and harvest more valuable information from the input sample. The dilated convolution is exploited in the output denoising feature maps to exponentially expand the receptive field of convolution kernel, so more redundant information can be acquired to reduce the effect of randomness. The capsule network is introduced to generate a set of vector neurons to represent an entity existing in the feature maps; as a result, more specific feature representations can be extracted, and the feature can be comprehended, thereby enhancing the diagnostic accuracy. As revealed from the experimental results, the GDCCN-based intelligent motor fault diagnosis method outperforms the other classical DL algorithms in diagnosis accuracy, noise resistance, generalization and transfer-learning performance of different workloads. [ABSTRACT FROM AUTHOR]

Load disaggregation technology is a key technology to realize real-time nonintrusive load monitoring (NILM), and deep learning method has shown great promise for NILM. However, current load disaggregation models based on deep learning are prone to the problems of gradient disappearance and model degradation, and it is difficult to extract effective features from load time series. In order to solve these problems, a new dilated residual attention deep network is proposed for load disaggregation. The proposed model adopts residual learning to extract high-level load features, reduces the difficulty of network optimization and solves the problem of network gradient disappearance. Dilated convolution is introduced to increase the receptive field of convolution kernels, which solves the problem that long-load time-series data are difficult to be learned. Most important of all, the proposed bottom-up and top-down attention mechanism can effectively extract the features of the abrupt points in mains power, improve the accuracy of judging the on/off state of electrical appliances and at the same time improve the learning ability of electrical appliances with low usage. Experiments on WikiEnergy dataset and UK-DALE dataset show that the proposed method achieves more accurate load disaggregation tasks than existing studies, which is of great significance for realizing practical NILM. [ABSTRACT FROM AUTHOR]

We propose an adaptive learning algorithm for cascades of boosted ensembles that is designed to handle the problem of concept drift in nonstationary environments. The goal was to create a real-time adaptive algorithm for dynamic environments that exhibit varying degrees of drift in high-volume streaming data. This we achieved using a hybrid of detect-and-retrain and constant-update approaches. The uniqueness of our method is found in two aspects of our framework. The first is the manner in which individual weak classifiers within each cascade layer of an ensemble are clustered during training and assigned a competence value. Secondly, the idea of learning optimal cascade-layer thresholds during runtime, which enables rapid adaptation to dynamic environments. The proposed adaptive learning method was applied to a binary-class problem with rare-event detection characteristics. For this, we chose the domain of face detection and demonstrate experimentally the ability of our algorithm to achieve an effective trade-off between accuracy and speed of adaptations in dense data streams with unknown rates of change. [ABSTRACT FROM AUTHOR]