Class Boundaries Research Articles

The evaluation model of light pollution degree based on Topsis method

Light pollution is an environmental problem that arises along with the economic development of modern society. To explore the influencing factors of light pollution degree, firstly, the evaluation indicator system was constructed. The system modified by factor analysis, then entropy weight method and multiple linear regression are used to calculate the weight values of factors respectively, and the comprehensive score index of 271 prefecture-level cities in China is obtained by substituting into TOPSIS. And K-Means++ clustering is performed on the scores to derive the grade classification boundary values, thus dividing the degree of light pollution into five criteria. Then the 271 prefecture-level cities in China are divided and visualized by the two rating systems, and the advantages and disadvantages are analyzed by comparing with satellite images. And the DN values were calculated with the composite score series derived from the two models separately to calculate the spearman coefficients, and the results showed a better fit of the multivariate linear TOPSIS. This article applies the superiority rating system of the previous work, selects four representative areas from each of the four scenarios with different levels of light pollution, calculates their comprehensive score index, and adopts the light pollution level classification and cause analysis. The results show that light pollution is very low in conservation land locations, low in rural communities, medium in suburban communities, and high in urban communities.

Soft-margin Ellipsoid generative adversarial networks

Generative adversarial networks (GANs) are of great significance for synthetizing realistic images. However, GANs are potentially unstable during the training process, posing some challenges for their development. By defining an integral probability metric (IPM) on the hypersphere, Sphere GAN enforces the discriminator to satisfy Lipschitz continuity and stabilizes the training process. Developed from Sphere GAN, a soft-margin Ellipsoid GAN is proposed for improving the quality of generated samples and the stability of training process. In the presented method, the geometric moment difference defined on the hypersphere is generalized to the hyperellipsoid. The hyperellipsoid is realized to relax the upper bound of the IPM by extending measurable functions space, thus the quality of generated samples can be improved. Furthermore, a nonlinear separating hyperellipsoid is designed to prevent the discriminator from gradient vanishing and exploding on the classification boundary. The proposed soft-margin Ellipsoid GAN is proved theoretically to have a global optimal solution, i.e., the probability density of generated samples approaches to that of real samples infinitely when both the discriminator and the generator are optimal. The CIFAR10 and LSUN-bedrooms datasets are selected to evaluate the performance of the proposed methods. The quantitative results show that the proposed approach decreases the Fréchet inception distance (FID) on the two datasets by 8.2% and 16.0%, respectively. The qualitative results show that the proposed soft-margin mechanism improves the stability of the training process.

Multi-source domain adaptation with spatio-temporal feature extractor for EEG emotion recognition

Electroencephalogram (EEG) emotion recognition suffers from cross-domain identification difficulties mainly caused by cross-subject and cross-session differences. Inspired by domain adaptation, the study constructs a novel multi-source domain adaptation with a Spatio-temporal feature extractor (MSDA-SFE) for EEG emotion recognition, which could reduce the signal differences between subjects and between sessions to make the target domain features align with the source domain features easily. In common feature learning, a Spatiotemporal feature extraction module is constructed to acquire domain-invariant characteristics of the source and target domains in which one subject is selected randomly as the target domain, and the remaining subjects are source domains. Whereafter, the obtained target domain features are paired with other source domain features respectively to generate N-1 pairs of concatenated features. And then these features are translated into N-1 domain-specific EEG features through N-1 parallel branches, where the target domain features are better aligned in the latent space, and the classification boundary of target samples is obtained by minimizing the disparity between branches. Finally, domain-specific features are fed to respective classifiers to get N-1 predictions, and their mean is put as the final result. Extensive experiments are implemented on SEED, SEED-IV, and DEAP benchmark databases. And, results indicate that the MSDA-SFE network outperforms the other comparison methods. Besides, the additional experiment also illustrates that the method has a powerful generalization capacity.

Extended rough sets model based on fuzzy granular ball and its attribute reduction

Attribute reduction is one of the core steps of data analysis. The attribute reduction method based on neighborhood rough sets (NRS) is widely used. However, the time complexity of this method is excessively high because its radius selection relies on grid search. Granular ball neighborhood rough sets (GBNRS) can generate different neighborhoods adaptively and has more generality and flexibility than the NRS method. Nevertheless, due to the purity of granular ball is required to be 1, the GBNRS algorithm will generate many granular balls with the sample number of 1 at the boundaries of classes. These granular balls will be regarded as outliers and deleted, resulting in the loss of class boundary information. Moreover, due to the strict reduction conditions of GBNRS, the reduction effect cannot be achieved on some datasets. To solve the above problems, the fuzzy granular ball is defined by relaxing the generation conditions of the granular ball, and an extended rough sets model is proposed on the basis of the fuzzy granular ball (FGBERS). In this model, we retain as much class boundary information as possible by preserving the fuzzy granular ball of class boundary. A new forward heuristic attribute reduction algorithm is designed on the basis of this model. To demonstrate the performance of FGBERS, we conducted sufficient experiments on 18 real datasets from different fields. The experimental results show that FGBERS not only compensates for the shortcomings of GBNRS, but also demonstrates higher classification accuracy, especially on the KNN classifier, the classification accuracy of the FGBERS algorithm is 5% higher on average, and the highest improvement is 20% on high-dimensional dataset ALLAML. In addition, compared with the comparison algorithms, FGBERS has excellent reduction performance in both large-scale datasets and high-dimensional datasets.

Dual-feature Fusion Attention Network for Small Object Segmentation

Accurate segmentation of medical images is an important step during radiotherapy planning and clinical diagnosis. However, manually marking organ or lesion boundaries is tedious, time-consuming, and prone to error due to subjective variability of radiologist. Automatic segmentation remains a challenging task owing to the variation (in shape and size) across subjects. Moreover, existing convolutional neural networks based methods perform poorly in small medical objects segmentation due to class imbalance and boundary ambiguity. In this paper, we propose a dual feature fusion attention network (DFF-Net) to improve the segmentation accuracy of small objects. It mainly includes two core modules: the dual-branch feature fusion module (DFFM) and the reverse attention context module (RACM). We first extract multi-resolution features by multi-scale feature extractor, then construct DFFM to aggregate the global and local contextual information to achieve information complementarity among features, which provides sufficient guidance for accurate small objects segmentation. Moreover, to alleviate the degradation of segmentation accuracy caused by blurred medical image boundaries, we propose RACM to enhance the edge texture of features. Experimental results on datasets NPC, ACDC, and Polyp demonstrate that our proposed method has fewer parameters, faster inference, and lower model complexity, and achieves better accuracy than more state-of-the-art methods.

Multiobjective Evolutionary Optimization for Prototype-Based Fuzzy Classifiers

Evolving intelligent systems (EISs), particularly, the zero-order ones have demonstrated strong performance on many real-world problems concerning data stream classification, while offering high model transparency and interpretability thanks to their prototype-based nature. Zero-order EISs typically learn prototypes by clustering streaming data online in a “one pass” manner for greater computation efficiency. However, such identified prototypes often lack optimality, resulting in less precise classification boundaries, thereby hindering the potential classification performance of the systems. To address this issue, a commonly adopted strategy is to minimise the training error of the models on historical training data or alternatively, to iteratively minimise the intra-cluster variance of the clusters obtained via online data partitioning. This recognises the fact that the ultimate classification performance of zero-order EISs is driven by the positions of prototypes in the data space. Yet, simply minimising the training error may potentially lead to overfitting, whilst minimising the intra-cluster variance does not necessarily ensure the optimised prototype-based models to attain improved classification outcomes. To achieve better classification performance whilst avoiding overfitting for zero-order EISs, this paper presents a novel multi-objective optimisation approach, enabling EISs to obtain optimal prototypes via involving these two disparate but complementary strategies simultaneously. Five decision-making schemes are introduced for selecting a suitable solution to deploy from the final non-dominated set of the resulting optimised models. Systematic experimental studies are carried out to demonstrate the effectiveness of the proposed optimisation approach in improving the classification performance of zero-order EISs.

Study the effect of MRF Model based NC classifier with different distance measures and parameters

The accuracy of satellite image classification and the computational complexity is reduced due to the image's noisy pixels. Therefore, spatial contextual information-based classifiers are required to handle the noisy pixels and obtain the neighborhood information. This paper represents Noise clustering (NC) based Markov Random Field (MRF) models (SP, DA (H1, H2, H3, and H4)) that handle the noisy pixels and provide the information. The Smoothing Prior (SP) and Discontinuity Adaptive (DA) models are useful for reducing noise by smoothing the images and showing the boundary of classes, respectively. This study has carried out a comparative study among MRF model-based NC classifiers SP and DA for different distance measures and parameters. MRF models based on NC classifiers were tested for classifying Eucalyptus, Water, Riverine sand, Grassland, Dense Forest, and Wheat classes using the Formosat-2 and Landsat-8 multispectral images of the Haridwar area. The DA (H1) model provides the best overall accuracy (85.09%) for m=1.3, λ=0.2, δ=104,γ=0.8, and Mean Absolute Difference.

Patterns of Integration: A Network Perspective on Popular Religious Connections in China’s Lower Yangzi, 1150–1350

The spread of cults from their original homelands in the Song dynasty (960–1279) created crisscrossing ties between local communities and fostered social and cultural integration in Chinese society that transcended class and geographic boundaries. Scholars have produced numerous case studies on these translocal cults and their implications, but the pattern of connections across space created by these cults is yet to be explored. Using the data collected from local gazetteers that have survived from the Southern Song and Yuan dynasties, this article takes a bird’s-eye view of the spatial distribution of popular cults in China’s Lower Yangzi region between 1150 and 1350 and employs the method of network analysis to study the pattern of connections formed through these religious ties. It reveals seven statistically significant subregional clusters of popular cults and three complementary mechanisms that tied these clusters together. It argues that integration across space was achieved not only through the spread of a cult and the attendant formation of a unified religious culture, but also through a multitude of less prominent cults which were each confined in their geographical scope of influence but collectively created a crisscrossing web of ties linking one subregional cluster to another. Host to a diversity of popular deities that were each associated with a different subregional cluster, the prefectural seats and the Southern Song capital Lin’an played a critical role in the social and cultural integration by providing a welcoming meeting ground for divergent communities of devotees.

Imitating the oracle: Towards calibrated model for class incremental learning

Class-incremental learning (CIL) aims to recognize classes that emerged in different phases. The joint-training (JT), which trains the model jointly with all classes, is often considered as the upper bound of CIL. In this paper, we thoroughly analyze the difference between CIL and JT in feature space and weight space. Motivated by the comparative analysis, we propose two types of calibration: feature calibration and weight calibration to imitate the oracle (ItO), i.e., JT. Specifically, on the one hand, feature calibration introduces deviation compensation to maintain the class decision boundary of old classes in feature space. On the other hand, weight calibration leverages forgetting-aware weight perturbation to increase transferability and reduce forgetting in parameter space. With those two calibration strategies, the model is forced to imitate the properties of joint-training at each incremental learning stage, thus yielding better CIL performance. Our ItO is a plug-and-play method and can be implemented into existing methods easily. Extensive experiments on several benchmark datasets demonstrate that ItO can significantly and consistently improve the performance of existing state-of-the-art methods. Our code is publicly available at https://github.com/Impression2805/ItO4CIL.

Gender, race and class at work: enlisting African health labour into the Gold Coast Medical Service, 1860–1957

From the mid-nineteenth century, the people of the Gold Coast formed an essential component of the missionary and early colonial medical services (CMS). The labour of the people was mainly...

Bourdieusian Boundary-Making, Social Networks, and Capital Conversion: Inequality among International Degree Holders in Hong Kong

Sociological research richly documents the many ways through which education becomes a form of convertible capital, but focuses less on the cultural schemas that graduates possess and use to respond to disruptions of capital conversion processes. Using the case of international degree holders in Hong Kong, this article draws upon Bourdieu’s theory of practice to interrogate the cultural schemas that valorize international degrees when their conversion pathways to economic capital are subjectively perceived to weaken. This article unearths the role of social networks in embedding cultural schemas and their effects on relations within the field: when faced with diminishing economic returns, international degree holders hold fast to their schemas vis-à-vis fellow international graduates and reconceptualize their degrees as symbolic capital to cope with the loss by enacting symbolic violence against domestic degree holders. Class boundaries are ultimately entrenched when international degree graduates valorize their cultural capital gains and legitimate their economic capital losses. Doing so compromises their class interests by forcing themselves into an interstitial position between different fields: though they occupy dominating homologous positions in the cultural field, they choose to overlook their dominated homologous positions in the economic field.

Randomized learning-based classification of sound quality using spectrogram image and time-series data: A practical perspective

Audio classification is an important research topic in the deep learning field, and classification accuracy has developed exponentially. However, as the model performance increases, it becomes difficult to apply it to actual industrial data owing to limitations such as the requirement of a large amount of data or increased training time. In this study, we developed a method for a small dataset with ambiguous classification boundaries, similar to most industrial data. We used randomized learning methods to reduce the training time and the risk of overfitting that required low computing time and few hyperparameters. Vehicle interior noise data were used as industrial data. The data were classified into luxury, powerful, and sporty, according to the human emotional reactions. Data was recorded from approximately 300 vehicles in the same environment and preprocessed into two types: spectrogram images and time-series data. Through deep fuzzy c-means clustering, it was confirmed that the classification boundary of the vehicle interior noise data was ambiguous. We selected the random vector functional link (RVFL), deep RVFL, and random kernel transformation (ROCKET) as randomized learning methods and compared the results with convolutional neural network and long short term memory. RVFL and deep RVFL showed similar classification accuracy to deep learning but significantly reduced training time by approximately 99%. In the case of ROCKET, the training time was reduced by approximately 91% and the classification performance was improved by approximately 3%. In addition, all randomized learning methods have few hyperparameters, which significantly reduces the classification model design time.

On the Minimal Adversarial Perturbation for Deep Neural Networks With Provable Estimation Error.

Although Deep Neural Networks (DNNs) have shown incredible performance in perceptive and control tasks, several trustworthy issues are still open. One of the most discussed topics is the existence of adversarial perturbations, which has opened an interesting research line on provable techniques capable of quantifying the robustness of a given input. In this regard, the euclidean distance of the input from the classification boundary denotes a well-proved robustness assessment as the minimal affordable adversarial perturbation. Unfortunately, computing such a distance is highly complex due the non-convex nature of DNNs. Despite several methods have been proposed to address this issue, to the best of our knowledge, no provable results have been presented to estimate and bound the error committed. This paper addresses this issue by proposing two lightweight strategies to find the minimal adversarial perturbation. Differently from the state-of-the-art, the proposed approach allows formulating an error estimation theory of the approximate distance with respect to the theoretical one. Finally, a substantial set of experiments is reported to evaluate the performance of the algorithms and support the theoretical findings. The obtained results show that the proposed strategies approximate the theoretical distance for samples close to the classification boundary, leading to provable robustness guarantees against any adversarial attacks.

Data-Based On-Board Diagnostics for Diesel Engine NOx-Reduction Aftertreatment Systems

Abstract The NOx conversion efficiency of a combined selective catalytic reduction and ammonia slip catalyst (SCR-ASC) in a diesel aftertreatment (AT) system degrades with time. A novel model-informed data-driven on-board diagnostic (OBD) binary classification strategy is proposed in this paper to distinguish an end of useful life (EUL) SCR-ASC catalyst from degreened (DG) ones. An optimized, supervised machine learning model was used for the classification with a calibrated single-cell three-state continuous stirred tank reactor (CSTR) observer used for state estimation. Percentage of samples classified as EUL (%EUL), with respect to classification boundary of 50%, was used as an objective criterion of classification. The method resulted in 87.5% classification accuracy when tested on eight day-files from four trucks (2 day-files per truck; 1 DG and 1 EUL) operating in real-world on-road conditions. Each day-file had ∼86,000 samples of data. Mileage of the same truck was used as ground truth for classification. However, mileage across different trucks cannot be used for classification since the operating conditions would vary across trucks.

Classification optimization of traditional Chinese medicine materials based on additive physical properties of powder

According to the method of predicting the physical properties of oily powder based on the additive physical properties of Chinese medicinal powder, Dioscoreae Rhizoma and calcined Ostreae Concha with high sieve rate and good fluidity were mixed and crushed with Persicae Semen, Platycladi Semen, Raphani Semen, Ziziphi Spinosae Semen, and other typical oily materials with high fatty oil content in proportion to obtain 23 mixed powders. Fifteen physical properties such as bulk density, water absorption, and maximum torque force were measured, and the physical properties of typical oily powders were predicted. When the mixing and grinding ratio was in the range of 5∶1-1∶1, the r value in the correlation equation between the weighted average score of the mixed powder and the powder proportion ranged from 0.801 to 0.986, and the linearity was good, indicating that the method of predicting the physical properties of oily powder based on the additive physical properties of traditional Chinese medicine(TCM)powder was feasible. The results of cluster analysis showed that the classification boundaries of the five kinds of TCM materials were clear, and the similarity of the physical fingerprints of powdery and oily materials decreased from 80.6% to 37.2%, which solved the problem of fuzzy classification boundaries of powdery and oily materials due to the lack of representativeness of oily material model drugs. The classification of TCM materials was optimized, laying a foundation for optimizing the prediction model of the prescription of personalized water-paste pills.

Semi-supervised Support Vector Machine

The article considers a new approach to constructing a support vector machine with semi-supervised learning for solving a classification problem. It is assumed that the distributions of the classes may overlap. The cost function has been modified by adding elements of a penalty to it for labels not in their class. The penalty is represented as a linear function of the distance between the label and the class boundary. To overcome the problem of multicriteria, a global optimization method known as continuation is proposed. For a combination of predictions, it is suggested to use the voting method of models with different kernels. The Optuna framework was chosen as the tool for configuring hyperparameters. The following were considered as training samples: type_dataset, banana, banana_inverse, c_circles, two_moons_classic, two_moons_tight, two_moons_wide.

Learning defense transformations for counterattacking adversarial examples

Deep neural networks (DNNs) are vulnerable to adversarial examples with small perturbations. Adversarial defense thus has been an important means which improves the robustness of DNNs by defending against adversarial examples. Existing defense methods focus on some specific types of adversarial examples and may fail to defend well in real-world applications. In practice, we may face many types of attacks where the exact type of adversarial examples in real-world applications can be even unknown. In this paper, motivated by that adversarial examples are more likely to appear near the classification boundary and are vulnerable to some transformations, we study adversarial examples from a new perspective that whether we can defend against adversarial examples by pulling them back to the original clean distribution. We empirically verify the existence of defense affine transformations that restore adversarial examples. Relying on this, we learn defense transformations to counterattack the adversarial examples by parameterizing the affine transformations and exploiting the boundary information of DNNs. Extensive experiments on both toy and real-world data sets demonstrate the effectiveness and generalization of our defense method. The code is avaliable at https://github.com/SCUTjinchengli/DefenseTransformer.

Shaped patch based nonparametric discriminant analysis for hyperspectral image classification through the CNN model

ABSTRACT Convolutional neural network (CNN) has shown superior classification results for hyperspectral image classification. Due to high dimensionality of the hyperspectral image, usually a feature reduction method is applied to the hyperspectral data before feeding it to the CNN model. Principal component analysis is the most widely used approach to this end. However, it has two main disadvantages. It uses the mean square error measure and ignores the class separability. Moreover, it is a pixel-based method, which does not utilize the spatial information. To deal with these weaknesses, the patch-based nonparametric discriminant analysis method is proposed for hyperspectral image feature reduction in this work. However, due to the complex and heterogeneous nature of the hyperspectral images, the use of fixed square patches is not so appropriate. To deal with this difficulty, a shaping approach is introduced for the generation of adaptive patches with consistent shapes in each local region. The input patches of the CNN model are also weighted according to the spectral similarity and class boundaries information. Moreover, a simple residual CNN model is suggested for the classification of the extracted shaped input patches. The efficiency of the proposed framework is assessed in different cases using various evaluation measures on six real hyperspectral images.

Violence is Islam, Violence is Not Islam: Meaning-Making Among Muslim Men in Norway

Abstract This article employs cultural repertoire theory to investigate how 84 Muslim men in Norway make meaning of adopting or rejecting political violence. Previous studies have addressed political violence among Muslims, but little attention has been paid to how its adoption and rejection involve self-ascription and ascriptions by others. The participants made meaning by drawing on stories about their past, exclusion and belonging, in addition to religious worldviews and political knowledge, including boundaries of class, crime, violence, race, religion and gender. Muslims are informed by mainstream ascriptions of them as extreme others and inherently radicalized in their meaning-making. This finding has important implications for how Muslim radicalization should be understood and countered.

Learning the Parameters of ELECTRE-Based Primal-Dual Sorting Methods that Use Either Characteristic or Limiting Profiles

Two multicriteria-sorting methods that generalize the relational paradigm have been recently presented in the literature. One uses objects representative of classes, the other uses objects in the limiting boundaries of classes; both can use either a reflexive or an asymmetric preference relation. However, defining the parameters of relation-based methods is not straightforward. The present work operationalizes those methods with a methodology that takes examples provided by the decision-maker and, using an accuracy measure that specifically fits the characteristics of the methods, exploits an evolutionary algorithm to determine the parameters that best reproduce such examples. The assessment of the proposal showed that (i) it can achieve considerably high levels of out-of-sample effectiveness with only a few decision examples; (ii) the inference process is more effective learning the parameters of the method based on representative objects; (iii) it tends to be more effective with a reflexive relation; (iv) the effectiveness decreases while increasing the number of classes, which is not always the case when increasing the number of criteria. Theoretical properties of the proposed methodology will be investigated in future works.