Proposed Method Research Articles

Incremental Fuzzy Clustering-Based Neural Networks Driven With the Aid of Dynamic Input Space Partition and Quasi-Fuzzy Local Models.

Fuzzy clustering-based neural networks (FCNNs) based on information granulation techniques have been shown to be effective Takagi-Sugeno (TS)-type fuzzy models. However, the existing FCNNs could not cope well with sequential learning tasks. In this study, we introduce incremental FCNNs (IFCNNs), which could dynamically update themselves whenever new learning data (e.g., single datum or block data) are incorporated into the dataset. Specifically, we employ dynamic (incremental) fuzzy C-means (FCMs) clustering algorithms to reveal a structure in data and divide the entire input space into several subregions. In the aforementioned partition, the dynamic FCM adaptively adjusts the position of its prototypes by using sequential data. Due to the time-sharing arrival of training data, compared with batch learning models, incremental learning methods may lose classification (prediction) accuracy. In order to tackle this challenge, we utilize quasi-fuzzy local models (QFLMs) based on modified Schmidt neural networks to replace the popular linear functions in TS-type fuzzy models to refine and enhance the ability to represent the behavior of fuzzy subspaces. Meanwhile, the recursive least square error (LSE) estimation is utilized to update the weights of QFLMs from one-by-one or block-by-block (fixed or varying block size) learning data. In addition, the L2 regularization is considered to ameliorate the deterioration of generalization abilities caused by potential overfitting when carrying out weight estimation. The proposed method leads to the construction of FCNNs in a new way, which can effectively deal with incremental data as well as deliver sound generalization capability. Extensive machine-learning datasets and a real-world application are employed to show the validity and performance of the presented methods. From the experimental results, we show that the proposal can maintain sound classification accuracy when effectively processing sequential data.

Improved Common-Mode Leakage Current Measurement Method for Insulation Condition Monitoring in Distribution Grids

Common-mode (CM) leakage current is an essential insulation indicator for condition monitoring in distribution grids. However, online measurement of CM leakage current is a difficult task in the industry due to the disturbances of far larger load currents. To address this challenge, a novel CM leakage current measurement method based on multi-layer magnetic shield has been proposed in this paper. The analytical solution for the magnetic field around the shield is firstly derived and validated by finite element analysis. Comprehensive comparison with previous single-layer shield design is then conducted for a typical industrial case, proving that an improvement of about 46 dB can be achieved for CM leakage current measurement by the novel design. A sensor prototype has also been developed with optimized parameters and validated by experiments. The effectiveness of the proposed method is proved, with minor CM leakage current accurately extracted from load currents which are 3000 times larger.

Bayesian Estimation of Inverted Beta Mixture Models With Extended Stochastic Variational Inference for Positive Vector Classification.

The finite inverted beta mixture model (IBMM) has been proven to be efficient in modeling positive vectors. Under the traditional variational inference framework, the critical challenge in Bayesian estimation of the IBMM is that the computational cost of performing inference with large datasets is prohibitively expensive, which often limits the use of Bayesian approaches to small datasets. An efficient alternative provided by the recently proposed stochastic variational inference (SVI) framework allows for efficient inference on large datasets. Nevertheless, when using the SVI framework to address the non-Gaussian statistical models, the evidence lower bound (ELBO) cannot be explicitly calculated due to the intractable moment computation. Therefore, the algorithm under the SVI framework cannot directly use stochastic optimization to optimize the ELBO, and an analytically tractable solution cannot be derived. To address this problem, we propose an extended version of the SVI framework with more flexibility, namely, the extended SVI (ESVI) framework. This framework can be used in many non-Gaussian statistical models. First, some approximation strategies are applied to further lower the ELBO to avoid intractable moment calculations. Then, stochastic optimization with noisy natural gradients is used to optimize the lower bound. The excellent performance and effectiveness of the proposed method are verified in real data evaluation.

Disentanglement by Cyclic Reconstruction.

Deep neural networks have demonstrated their ability to automatically extract meaningful features from data. However, in supervised learning, information specific to the dataset used for training, but irrelevant to the task at hand, may remain encoded in the extracted representations. This remaining information introduces a domain-specific bias, weakening the generalization performance. In this work, we propose splitting the information into a task-related representation and its complementary context representation. We propose an original method, combining adversarial feature predictors and cyclic reconstruction, to disentangle these two representations in the single-domain supervised case. We then adapt this method to the unsupervised domain adaptation (UDA) problem, consisting of training a model capable of performing on both a source and a target domain. In particular, our method promotes disentanglement in the target domain, despite the absence of training labels. This enables the isolation of task-specific information from both domains and a projection into a common representation. The task-specific representation allows the efficient transfer of knowledge acquired from the source domain to the target domain. In the single-domain case, we demonstrate the quality of our representations on information retrieval tasks and the generalization benefits induced by sharpened task-specific representations. We then validate the proposed method on several classical domain adaptation (DA) benchmarks and illustrate the benefits of disentanglement for DA.

Interpretable Latent Space for Meteorological Out-of-Distribution Detection via Weak Supervision

Deep neural networks (DNNs) are effective tools for learning-enabled cyber-physical systems (CPSs) that handle high-dimensional image data. However, DNNs may make incorrect decisions when presented with inputs outside the distribution of their training data. These inputs can compromise the safety of CPSs. So, it becomes crucial to detect inputs as out-of-distribution (OOD) and interpret the reasons for their classification as OOD. In this study, we propose an interpretable learning method to detect OOD caused by meteorological features like darkness, lightness, and rain. To achieve this, we employ a variational autoencoder (VAE) to map high-dimensional image data to a lower-dimensional latent space. We then focus on a specific latent dimension and encourage it to classify different intensities of a particular meteorological feature in a monotonically increasing manner. This is accomplished by incorporating two additional terms into the VAE’s loss function: a classification loss and a positional loss. During training, we optimize the utilization of label information for classification. Remarkably, our results demonstrate that using only 25% of the training data labels is sufficient to train a single pre-selected latent dimension to classify different intensities of a specific meteorological feature. We evaluate the proposed method on two distinct datasets, CARLA and Duckietown, employing two different rain-generation methods. We show that our approach outperforms existing approaches by at least 15 in the F1 score and precision when trained and tested on CARLA dataset.

Perceptual Quality-Oriented Rate Allocation via Distillation from End-to-End Image Compression

Mainstream image/video coding standards, exemplified by the state-of-the-art H.266/VVC, AVS3, and AV1, follow the block-based hybrid coding framework. Due to the block-based framework, encoders designed for these standards are easily optimized for peak signal-to-noise ratio (PSNR) but have difficulties optimizing for the metrics more aligned to perceptual quality, e.g., multi-scale structural similarity (MS-SSIM), since these metrics cannot be accurately evaluated at the small block level. We address this problem by leveraging inspiration from the end-to-end image compression built on deep networks, which is easily optimized through network training for any metric as long as the metric is differentiable. We compared the trained models using the same network structure but different metrics and observed that the models allocate rates in different ratios. We then propose a distillation method to obtain the rate allocation rule from end-to-end image compression models with different metrics and to utilize such a rule in the block-based encoders. We implement the proposed method on the VVC reference software—VTM and the AVS3 reference software—HPM, focusing on intraframe coding. Experimental results show that the proposed method on top of VTM achieves more than 10% BD-rate reduction than the anchor when evaluated with MS-SSIM or LPIPS, which leads to concrete perceptual quality improvement.

SMoGW-based deep CNN: Plant disease detection and classification using SMoGW-deep CNN classifier

Diagnosing plant disease is a major role to reduce adequate losses in yield production, which further leads to economic losses. The various disease control measures are accessible without a proper diagnosis of the disease which results in a waste of expenses and time. The diagnosis of disease using images leads to unsatisfactory results in the prevalent methods due to the image clarity. It is mainly caused by the worst performance of the existing pre-trained image classifiers. This issue can be controlled by the SMoGW-deep convolutional neural network (deep CNN) classifier for the accurate and precise classification of plant leaf disease. The developed method transforms the poor-quality captured images into high quality by the preprocessing technique. The preprocessed input images contain pixels on their dimension and also the value of the threshold is analyzed by the Otsu method by which the particular disease-affected region is extracted based on the image pixels. The region of interest is separated from the other parts of the input leaf image using the K-means segmentation technique. The stored features in the feature vector are fed forward to the deep CNN classifier for training and are optimized by the SMoGW optimization approach. The experiments are done and achieved an accuracy of 94.5% sensitivity of 94.525%, specificity of 94.6%, precision of 95% with 90% of training data and under K-fold training with 95% of accuracy, 95% of sensitivity, 94.1% of specificity, and 92.1% of precession is achieved for the SMoGW-optimized classifier approach that is superior to the prevalent techniques for disease classification and detection. The potential, as well as the capability of the proposed method, is experimentally demonstrated for plant leaf disease classification and identification.

Multi-attribute decision-making analysis based on the bipolar N-soft PROMETHEE method

The multi-attribute decision-making (MADM) methods can deeply mine hidden information in data and make a more reliable decision with actual needs and human cognition. For this reason, this paper proposes the bipolar N-soft PROMETHEE (preference ranking organization method for enrichment of evaluation) method. The method fully embodies the advantages of the PROMETHEE method, which can limit the unconditional compensation between attribute values and effectively reflect the priority between attribute values. Further, by introducing an attribute threshold to filter research objects, the proposed method not only dramatically reduces the amount of computation but also considers the impact of the size of the attribute value itself on decision-making. Secondly, the paper proposes the concepts of attribute praise, attribute popularity, total praise, and total popularity for the first time, fully mining information from bipolar N-soft sets, which can effectively handle situations where attribute values have different orders of magnitude. In addition, this paper presents the decision-making process of the new method, closely integrating theoretical models with real life. Finally, this paper analyses and compares the proposed method with the existing ones, further verifying the effectiveness and flexibility of the proposed method.

A machine-based corpus optimization method for extracting domain-oriented technical words: an example of COVID-19 corpus data

Although natural language processing (NLP) refers to a process involving the development of algorithms or computational models that empower machines to understand, interpret, and generate human language, machines are still unable to fully grasp the meanings behind words. Specifically, they cannot assist humans in categorizing words with general or technical purposes without predefined standards or baselines. Empirically, prior researches have relied on inefficient manual tasks to exclude these words when extracting technical words (i.e., terminology or terms used within a specific field or domain of expertise) for obtaining domain information from the target corpus. Therefore, to enhance the efficiency of extracting domain-oriented technical words in corpus analysis, this paper proposes a machine-based corpus optimization method that compiles an advanced general-purpose word list (AGWL) to serve as the exclusion baseline for the machine to extract domain-oriented technical words. To validate the proposed method, this paper utilizes 52 COVID-19 research articles as the target corpus and an empirical example. After compared to traditional methods, the proposed method offers significant contributions: (1) it can automatically eliminate the most common function words in corpus data; (2) through a machine-driven process, it removes general-purpose words with high frequency and dispersion rates –57% of word types belonging to general-purpose words, constituting 90% of the total words in the target corpus. This results in 43% of word types representing domain-oriented technical words that makes up 10% of the total words in the target corpus are able to be extracted. This allows future researchers to focus exclusively on the remaining 43% of word types in the optimized word list (OWL), enhancing the efficiency of corpus analysis for extracting domain knowledge. (3) The proposed method establishes a set of standard operation procedure (SOP) that can be duplicated and generally applied to optimize any corpus data.

Enhancing the transferability of adversarial attacks with diversified input strategies

Abstract Adversarial examples are often used to test and evaluate the security and robustness of image classification models. Though adversarial attacks under white-box setting can achieve a high attack success rate, due to overfitting, the success rate of black-box attacks is relatively low. To this end, this paper proposes diversified input strategies to improve the transferability of adversarial examples. In this method, various transformation methods are applied to randomly transform the original image multiple times, thereby generating a batch of transformed images. Then, in the process of back-propagation, the loss function gradient of the transformed images is calculated, and a weighted average of the obtained gradient values is performed to generate adversarial perturbation, which is iteratively added to the original image to generate adversarial examples. Meanwhile, by increasing the variety of data augmentation transformation types and the number of input images, the proposed method effectively alleviates overfitting and improves the transferability of adversarial examples. Extensive experiments on the ImageNet dataset indicate that the proposed method can perform black-box attacks better than benchmark methods, with an average of 97.2% success rate attacking multiple models simultaneously.

Reliability analysis of intelligent manufacturing systems based on improved FMEA combined with machine learning

Along with the booming of intelligent manufacturing, the reliability management of intelligent manufacturing systems appears to be becoming more significant. Failure mode and effects analysis (FMEA) is a prospective reliability management instrument extensively utilized to manage failure modes of systems, products, processes, and services in various industries. However, the conventional FMEA method has been criticized for its inherent limitations. Machine learning can handle large amounts of data and has merits in reliability analysis and prediction, which can help in failure mode classification and risk management under limited resources. Therefore, this paper devises a method for complex systems based on an improved FMEA model combined with machine learning and applies it to the reliability management of intelligent manufacturing systems. First, the structured network of failure modes is constructed based on the knowledge graph for intelligent manufacturing systems. Then, the grey relation analysis (GRA) is applied to determine the risk prioritization of failure modes. Hereafter, the k-means algorithm in unsupervised machine learning is employed to cluster failure modes into priority classes. Finally, a case study and further comparative analysis are implemented. The results demonstrate that failure modes in system security, production quality, and information integration are high-risk and require more resources for prevention. In addition, recommendations for risk prevention and monitoring of intelligent manufacturing systems were given based on the clustering results. In comparison to the conventional FMEA method, the proposed method can more precisely capture the coupling relationship between the failure modes compared with. This research provides significant support for the reliability and risk management of complex systems such as intelligent manufacturing systems.

An improved IPA approach driven by big data and its application to customer satisfaction research of energy-saving appliance

Against the backdrop of increasing global demand for reducing greenhouse gas emissions, promoting the use of energy-saving and environmentally friendly products has become a crucial aspect of low-carbon economic development. Customer satisfaction plays a vital role in the promotion of these products. To address the challenges of dealing with big data in the conventional customer satisfaction analysis tool, Importance Performance Analysis (IPA), a machine learning-based method is proposed to improve IPA. Firstly, the Latent Dirichlet Allocation (LDA) model is used to capture users’ opinions on different product topics. Then, the Support Vector Machine (SVM) and Random Forest (RF) algorithms are employed respectively to assess the satisfaction and importance of product attributes, enabling an objective measurement of customer satisfaction and adapting to the current trend of big data. The proposed method is applied to the analysis of water heater satisfaction on the JD platform, obtaining satisfaction levels for 10 topics. The research findings demonstrate that the improved IPA method based on SVM-RF effectively explores customer satisfaction and can provide some improvement strategies for platform managers and manufacturers.

Constructing a MOEA approach for product form Kansei design based on text mining and BPNN

Kansei Engineering (KE) is a product design method that aims to develop products to meet users’ emotional preferences. However, traditional KE faces the problem that the acquisition of Kansei factors does not represent the real consumers demands based on manual and reports, and using traditional methods to calculate relationship between Kansei factors and specific design elements, which can lead to the omission of key information. To address these problems, this study adopts text mining and backward propagation neural networks (BPNN) to propose a product form design method from a multi-objective optimization perspective. Firstly, Term Frequency-Inverse Document Frequency (TF-IDF) and WordNet are used to extract key user Kansei requirements from online review texts to obtain more accurate Kansei knowledge. Secondly, the BPNN is used to establish the non-linear relationship between product Kansei factors and specific design elements, and a preference mapping prediction model is constructed. Finally, BPNN is transformed into an iterative prediction value of non-dominated sorting genetic algorithm-II (NSGA-II), and the model is solved through multi-objective evolutionary algorithm (MOEA) to obtain the Pareto optimal solution set that satisfies the user’s multiple emotional needs, and the fuzzy Delphi method is used to obtain the best product form design scheme that meets the user’s multiple emotional images. Using the example of electric bicycle form design could show that this proposed method can effectively complete multi-objective product solutions innovation design.

GRU combined model based on multi-objective optimization for short-term residential load forecasting

Reliable and accurate short-term forecasting of residential load plays an important role in DSM. However, the high uncertainty inherent in single-user loads makes them difficult to forecast accurately. Various traditional methods have been used to address the problem of residential load forecasting. A single load forecast model in the traditional method does not allow for comprehensive learning of data characteristics for residential loads, and utilizing RNNs faces the problem of long-term memory with vanishing or exploding gradients in backpropagation. Therefore, a gated GRU combined model based on multi-objective optimization is proposed to improve the short-term residential load forecasting accuracy in this paper. In order to demonstrate the effectiveness, GRUCC-MOP is first experimentally tested with the unimproved model to verify the model performance and forecasting effectiveness. Secondly the method is evaluated experimentally with other excellent forecasting methods: models such as DBN, LSTM, GRU, EMD-DBN and EMD-MODBN. By comparing simulation experiments, the proposed GRU combined model can get better results in terms of MAPE on January, April, July, and November load data, so this proposed method has better performance than other research methods in short-term residential load forecasting.

Method for wind power forecasting based on support vector machines optimized and weighted composite gray relational analysis

This study proposes a weighted composite approach for grey relational analysis (GRA) that utilizes a numerical weather prediction (NWP) and support vector machine (SVM). The approach is optimized using an improved grey wolf optimization (IGWO) algorithm. Initially, the dimension of NWP data is decreased by t-distributed stochastic neighbor embedding (t-SNE), then the weight of sample coefficients is calculated by entropy-weight method (EWM), and the weighted grey relational of data points is calculated for different weather numerical time series data. At the same time, a new weighted composite grey relational degree is formed by combining the weighted cosine similarity of NWP values of the historical day and to be measured day. The SVM’s regression power prediction model is constructed by the time series data. To improve the accuracy of the system’s predictions, the grey relational time series data is chosen as the input variable for the SVM, and the influence parameters of the ideal SVM are discovered using the IGWO technique. According to the simulated prediction and analysis based on NWP, it can be observed that the proposed method in this study significantly improves the prediction accuracy of the data. Specifically, evaluation metrics such as root mean squared error (RMSE), regression correlation coefficient (r2), mean absolute error (MAE) and mean absolute percent error (MAPE) all show corresponding enhancements, while the computational burden remains relatively low.

Optimized multiscale deep bidirectional gated recurrent neural network fostered practical teaching of university music course

Music education has a rich historical background. Nevertheless, the introduction of modern teaching methods is relatively delayed. In recent years, there has been a remarkable acceleration in the advancement of music education. A promising tool that has emerged to revolutionize education as a whole is Virtual Reality (VR) technology, which offers immersive and interactive experiences across various disciplines. At the university level, integrating VR technology into music education opens up exciting opportunities to enhance practical teaching methods and provide students with enriched musical experiences. Virtual Reality together with Internet of Things (IoT) demonstrates its capabilities in various tasks, but its widespread availability in online learning remainders a pressing challenge that needs to be addressed. In pre-processing, it removes noise data using Dynamic Context-Sensitive Filtering (DCSF). VR technology creates an unparalleled learning environment, it transporting students to virtual concert halls, recording studios, or historical music venues. Hence the Multiscale deep bidirectional gated recurrent neural Network (MDBGNN) improves the practical teaching of music course concept, like Music theory, harmony, and rhythm can be visualized and experienced in VR. Finally, Dung Beetle Optimization Algorithm (DBOA) is employed to optimize the weight parameters of MDBGNN. The proposed MDBGNN-DBO-UMC-VRT is implemented in Python. The proposed method is analysed with the help of performance metrics, like precision, accuracy, F1-score, Recall (Sensitivity), Specificity, Error rate, Computation time and RoC. The proposed MDBGNN-DBO-UMC-VRT method attains 13.11%, 18.12% and 18.73% high specificity, 11.13%, 11.04% and 19.51% lower computation Time, 15.29%, 15.365% and 14.551% higher ROC and 13.65%, 15.98%, and 17.15% higher Accuracy compared with existing methods, such as Enhancing Vocal Music Teaching through the Fusion of Artificial Intelligence Algorithms and VR Technology (CNN-UMC-VRT), Exploring the Efficacy of VR Technology in Augmenting Music Art Teaching (BPNN-UMC-VRT) and Implementing an Interactive Music-Assisted Teaching System Using VR Technology (DNN-UMC-VRT) respectively.

Powerlifting total score prediction based on an improved random forest regression algorithm

This paper discusses the use of an improved random forest regression algorithm (RFRA) to predict the total score of powerlifters. The paper collected the age, weight, and total score of multiple powerlifters, and then used an improved RFRA to build a predictive model. The parameters of this model are optimized by a differential squirrel search algorithm. The experimental results show that our proposed method can effectively predict the total score of powerlifters with an error of less than 10%, which can provide a reference for experts and athletes before training or competition.

Intelligent control for solar-electric vehicles: integrating gannet optimization algorithm and quantum neural network for power tracking and high gain conversion

This paper presents a hybrid approach for optimizing the maximum power point tracking of photovoltaic (PV) systems in electric vehicles. The hybrid technique involves the simultaneous utilization of the Gannet Optimization Algorithm (GOA) and Quantum Neural Network (QNN), collectively referred to as the GOA-QNN technique. The primary aim is to enhance the efficiency and maximize the power output of PV systems. The proposed hybrid methodology boosts the performance of the photovoltaic system by managing the power interface. A high step-up DC/DC converter is employed to adjust the photovoltaic source power and load, ensuring optimal power transfer under various operating conditions. The proposed method optimally determines the duty cycle of the converter. Subsequently, the model is implemented in the MATLAB/Simulink platform, and its execution is evaluated using established procedures. The results clearly demonstrate the superiority of the proposed method over existing approaches in terms of power quality, settling time, and controller stability. The proposed technique achieves an impressive efficiency level of 95%, exceeding the efficiency of other existing techniques.

Multi-label text classification of cardiovascular drug attributes based on BERT and BiGRU

Extracting and digitizing drug attributes from medical literature is the first step to build a knowledge computing system for precision disease treatment. In order to build a cardiovascular drug knowledge base, this paper proposes a multi-label text classification method for cardiovascular drug attributes from the Chinese drug guideline. The drug attributes are characterized by a BERT pre-trained model, and a dual-feature extraction structure is proposed based on the BiGRU neural network to capture high-level semantic information. Label categorization of cardiovascular drug attributes, such as indications and mode of administration, is accomplished. The F1 score of 0.8431 was obtained using 5-fold cross-validation. Comparing KNN and Naïve bayes, and conducting CNN and BiGRU control experiments on the basis of Word2Vec characterization of medication guidelines, the proposed multi-label text classification method is effective and the F1 value is significantly improved. Proved by analysis of ablation and crossover experiments, the proposed method can achieve a high accuracy rate averaged at 0.8339.

A text extraction framework of financial report in traditional format with OpenCV

Due to intensified off-balance sheet disclosure by regulatory authorities, financial reports now contain a substantial amount of information beyond the financial statements. Consequently, the length of footnotes in financial reports exceeds that of the financial statements. This poses a novel challenge for regulators and users of financial reports in efficiently managing this information. Financial reports, with their clear structure, encompass abundant structured information applicable to information extraction, automatic summarization, and information retrieval. Extracting headings and paragraph content from financial reports enables the acquisition of the annual report text’s framework. This paper focuses on extracting the structural framework of annual report texts and introduces an OpenCV-based method for text framework extraction using computer vision. The proposed method employs morphological image dilation to distinguish headings from the main body of the text. Moreover, this paper combines the proposed method with a traditional, rule-based extraction method that exploits the characteristic features of numbers and symbols at the beginning of headings. This combination results in an optimized framework extraction method, producing a more concise text framework.