Similar Feature Distributions Research Articles

Fault Diagnosis of Rolling Bearings in Primary Mine Fans under Sample Imbalance Conditions.

Rolling bearings are crucial parts of primary mine fans. In order to guarantee the safety of coal mine production, primary mine fans commonly work during regular operation and are immediately shut down for repair in case of failure. This causes the sample imbalance phenomenon in fault diagnosis (FD), i.e., there are many more normal state samples than faulty ones, seriously affecting the precision of FD. Therefore, the current study presents an FD approach for the rolling bearings of primary mine fans under sample imbalance conditions via symmetrized dot pattern (SDP) images, denoising diffusion probabilistic models (DDPMs), the image generation method, and a convolutional neural network (CNN). First, the 1D bearing vibration signal was transformed into an SDP image with significant characteristics, and the DDPM was employed to create a generated image with similar feature distributions to the real fault image of the minority class. Then, the generated images were supplemented into the imbalanced dataset for data augmentation to balance the minority class samples with the majority ones. Finally, a CNN was utilized as a fault diagnosis model to identify and detect the rolling bearings' operating conditions. In order to assess the efficiency of the presented method, experiments were performed using the regular rolling bearing dataset and primary mine fan rolling bearing data under actual operating situations. The experimental results indicate that the presented method can more efficiently fit the real image samples' feature distribution and generate image samples with higher similarity than other commonly used methods. Moreover, the diagnostic precision of the FD model can be effectively enhanced by gradually expanding and enhancing the unbalanced dataset.

Distilling Region-Wise and Channel-Wise Deep Structural Facial Relationships for FAU (DSR-FAU) Intensity Estimation

Facial emotions are expressed through a combination of facial muscle movements, namely, the Facial Action Units (FAUs). FAU intensity estimation aims to estimate the intensity of a set of structurally dependent FAUs. Contrary to the existing works that focus on improving FAU intensity estimation performance, this study investigates how knowledge distillation (KD) incorporated into a training model can improve FAU intensity estimation efficiency while achieving the comparable level of performance. Given the intrinsic structural characteristics of FAU, it is desirable to distill deep structural relationships, namely, DSR-FAU, using heatmap regression. Our methodology is as follows: First, a feature map-level distillation loss is applied to ensure that the student network and the teacher network share similar feature distributions. Second, the region-wise and channel-wise relationship distillation loss functions are introduced to penalize the difference in structural relationships. Specifically, the region-wise relationship can be represented by the structural correlations across the facial features, whereas the channel-wise relationship is represented by the implicit FAU co-occurrence dependencies. Third, we compare the model performance of DSR-FAU with the state-of-the-art models, based on two benchmarking datasets. It is shown that our model achieves comparable performance, with a lower number of model parameters and lower computation complexities.

Urban scene segmentation model based on multi-scale shuffle features.

The monitoring of urban land categories is crucial for effective land resource management and urban planning. To address challenges such as uneven parcel distribution, difficulty in feature extraction and loss of image information in urban remote sensing images, this study proposes a multi-scale feature shuffle urban scene segmentation model. The model utilizes a deep convolutional encoder-decoder network with BlurPool instead of MaxPool to compensate for missing translation invariance. GSSConv and SE module are introduced to enhance information interaction and filter redundant information, minimizing category misclassification caused by similar feature distributions. To address unclear boundary information during feature extraction, the model applies multi-scale attention to aggregate context information for better integration of boundary and global information. Experiments conducted on the BDCI2017 public dataset show that the proposed model outperforms several established segmentation networks in OA, mIoU, mRecall, P and Dice with scores of 83.1%, 71.0%, 82.7%, 82.7% and 82.5%, respectively. By effectively improving the completeness and accuracy of urban scene segmentation, this study provides a better understanding of urban development and offers suggestions for future planning.

Cross-Database Micro-Expression Recognition Based on a Dual-Stream Convolutional Neural Network

Cross-database micro-expression recognition (CDMER) under semi supervised conditions is a difficult task, where the target (testing) and source (training) samples come from different micro-expression (ME) databases, resulting in the inconsistency of the feature distributions between each other, and hence affecting the performance of many existing MER methods. To address this problem, we propose a dual-stream convolutional neural network (DSCNN) for dealing with CDMER tasks. In the DSCNN, two stream branches are designed to study temporal and facial region cues in ME samples with the goal of recognizing MEs. In addition, in the training process, the domain discrepancy loss is used to enforce the target and source samples to have similar feature distributions in some layers of the DSCNN. Extensive CDMER experiments are conducted to evaluate the DSCNN. The results show that our proposed DSCNN model achieves a higher recognition accuracy when compared with some representative CDMER methods.

Domain Adaptation for Remote Sensing Image Semantic Segmentation: An Integrated Approach of Contrastive Learning and Adversarial Learning

Although semantic segmentation models based on deep neural networks (DNNs) have achieved excellent results, generalizing well from one remote sensing dataset (source domain) to another dataset with different acquisition conditions (target domain) remains a major challenge. Many domain adaptation (DA) approaches have been proposed to address this problem. DA aims to help DNNs learn a generalizable representation space in which source and target domains have similar feature distributions, but most of the existing DA approaches have difficulty in aligning the high-dimensional image representations of two domains directly. In this study, we proposed a model integrating contrastive learning and adversarial learning in a unified framework for aligning two domains in both representation space and spatial layout. Specifically, the model consists of a semantic segmentation network for feature extraction and two branches for DA. The first branch is used for adaptation in representation space directly by a proposed pixelwise contrastive loss, while the second branch is used for adaptation in predicted results to help two domains have similar spatial layouts through a novel but simple entropy-based similarity discriminator. Additionally, a training strategy called category similarity matching sampling was proposed to provide source and target image pairs with similar category composition for each training iteration, which can help the two branches work better. Extensive experiments indicated that the two branches can benefit each other to gain a superior performance and DA pretraining by our methods can achieve impressive results with only a small number of target labeled samples.

Cross-Domain Scene Classification Based on a Spatial Generalized Neural Architecture Search for High Spatial Resolution Remote Sensing Images

By labelling high spatial resolution (HSR) images with specific semantic classes according to geographical properties, scene classification has been proven to be an effective method for HSR remote sensing image semantic interpretation. Deep learning is widely applied in HSR remote sensing scene classification. Most of the scene classification methods based on deep learning assume that the training datasets and the test datasets come from the same datasets or obey similar feature distributions. However, in practical application scenarios, it is difficult to guarantee this assumption. For new datasets, it is time-consuming and labor-intensive to repeat data annotation and network design. The neural architecture search (NAS) can automate the process of redesigning the baseline network. However, traditional NAS lacks the generalization ability to different settings and tasks. In this paper, a novel neural network search architecture framework—the spatial generalization neural architecture search (SGNAS) framework—is proposed. This model applies the NAS of spatial generalization to cross-domain scene classification of HSR images to bridge the domain gap. The proposed SGNAS can automatically search the architecture suitable for HSR image scene classification and possesses network design principles similar to the manually designed networks. To obtain a simple and low-dimensional search space, the traditional NAS search space was optimized and the human-the-loop method was used. To extend the optimized search space to different tasks, the search space was generalized. The experimental results demonstrate that the network searched by the SGNAS framework with good generalization ability displays its effectiveness for cross-domain scene classification of HSR images, both in accuracy and time efficiency.

Statistical harmonization can improve the development of a multicenter CT-based radiomic model predictive of nonresponse to induction chemotherapy in laryngeal cancers.

To develop a radiomic model predicting nonresponse to induction chemotherapy in laryngeal cancers, from multicenter pretherapeutic contrast-enhanced computed tomography (CE-CT) and evaluate the benefit of feature harmonization in such a context. Patients (n = 104) eligible for laryngeal preservation chemotherapy were included in five centers. Primary tumor was manually delineated on the CE-CT images. The following radiomic features were extracted with an in-house software (MIRAS v1.1, LaTIM UMR 1101): intensity, shape, and textural features derived from Gray-Level Co-occurrence Matrix (GLCM), Neighborhood Gray Tone Difference Matrix (NGTDM), Gray-Level Run Length Matrix (GLRLM), and Gray-Level Size Zone Matrix (GLSZM). Harmonization was performed using ComBat after unsupervised hierarchical clustering, used to determine labels automatically, given the high heterogeneity of imaging characteristics across and within centers. Patients with similar feature distributions were grouped with unsupervised clustering into an optimal number of clusters (2) determined with "silhouette scoring." Statistical harmonization was then carried out with ComBat on these 2 identified clusters. The cohort was split into training/validation (n = 66) and testing (n = 32) sets. Area under the receiver operating characteristics curves (AUC) were used to evaluate the ability of radiomic features (before and after harmonization) to predict nonresponse to chemotherapy, and specificity (Sp) and sensitivity (Se) were used to quantify their performance in the testing set. Without harmonization, none of the features identified as predictive in the training set remained significant in the testing set. After ComBat, one textural feature identified in the training set keeps a predictive trend in the testing set-Zone Percentage, derived from the GLSZM, was predictive of nonresponse in the training set (AUC = 0.62, Se = 70%, Sp = 64%, P = 0.04) and obtained a satisfactory performance in the testing set (Se = 80%, Sp = 67%, P = 0.03), although significance was limited by the size of the testing set. These results are consistent with previously published findings in head and neck cancers. Radiomic features from CE-CT could help in the selection of patients for induction chemotherapy in laryngeal cancers, with relatively good sensitivity and specificity in predicting lack of response. Statistical harmonization with ComBat and unsupervised clustering seems to improve the predictive value of features extracted in such a heterogeneous multicenter setting.

Feature distribution-based label correlation in multi-label classification

In multi-label classification, multiple label variables in output space are equally important and can be predicted according to a common set of input variables. To improve the accuracy and efficiency of multi-label learner, measuring and utilizing label correlation is the core breakthrough. Extensive research on label correlation focuses on the co-occurrence or mutual exclusion frequency of label values in output space. In this paper, to handle the multi-label learning tasks, a novel method, named FL-MLC, is proposed by considering the influence of feature-label dependencies on inter-label correlations. In order to describe the intrinsic relationship between feature variable and label variable, the discriminant weight of any feature to label is first defined. Therefore, the concept of feature distribution for inputs on label is proposed to reflect the discriminant weights of features to the label. The corresponding calculation process is also designed based on multiple kernel learning and kernel alignment. Furthermore, the feature distributions on different labels are integrated into the feature distribution-based label correlation by using two different aggregation strategies. Obviously, arbitrary label variables with highly similar feature distributions have strong relevance. Thus, the feature distribution-based label correlation is applied to adjust the distance between the parameters for different labels in the predictive learner of FL-MLC method. Finally, the experimental results on twelve real-world datasets demonstrate that our methods achieves good effectiveness and versatility for multi-label classification.

Transfer Subspace Learning for Unsupervised Cross-Corpus Speech Emotion Recognition

In many practical applications, a speech emotion recognition model learned on a source (training) domain but applied to a novel target (testing) domain degenerates even significantly due to the mismatch between the two domains. Aiming at learning a better speech emotion recognition model for the target domain, the paper investigates this interesting problem, i.e., unsupervised cross-corpus speech emotion recognition (SER), in which the training and testing speech signals come from two different speech emotion corpora. Meanwhile, the training speech signals are labeled, while the label information of the testing speech signals is entirely unknown. To deal with this problem, we propose a simple yet effective method called transfer subspace learning (TRaSL). TRaSL aims at learning a projection matrix with which we can transform the source and target speech signals from the original feature space to the label space. The transformed source and target speech signals in the label space would share similar feature distributions. Consequently, the classifier learned on the labeled source speech signals can effectively predict the emotional states of the unlabeled target speech signals. To evaluate the performance of the proposed TRaSL method, we carry out extensive cross-corpus SER experiments on four speech emotion corpora including IEMOCAP, EmoDB, eNTERFACE, and AFEW 4.0. Compared with recent state-of-the-art cross-corpus SER methods, the proposed TRaSL can achieve more satisfactory overall results.

Standardization-refinement domain adaptation method for cross-subject EEG-based classification in imagined speech recognition

Recent advances in imagined speech recognition from EEG signals have shown their capability of enabling a new natural form of communication, which is posed to improve the lives of subjects with motor disabilities. However, differences among subjects may be an obstacle to the applicability of a previously trained classifier to new users, since a significant amount of labeled samples must be acquired for each new user, making this process tedious and time-consuming. In this sense, unsupervised domain adaptation (UDA) methods, especially those based on deep learning (D-UDA), arise as a potential solution to address this issue by reducing the differences among feature distributions of subjects. It has been shown that the divergence in the marginal and conditional distributions must be reduced to encourage similar feature distributions. However, current D-UDA methods may become sensitive under adaptation scenarios where a low discriminative feature space among classes is given, reducing the accuracy performance of the classifier. To address this issue, we introduce a D-UDA method, named Standardization-Refinement Domain Adaptation (SRDA), which combines Adaptive Batch Normalization (AdaBN) with a novel loss function based on the variation of information (VOI), in order to build an adaptive classifier on EEG data corresponding to imagined speech. Our proposal, applied over two imagined speech datasets, resulted in SRDA outperforming standard classifiers for BCI and existing D-UDA methods, achieving accuracy performances of 61.02±08.14% and 62.99±04.78%, assessed using leave-one-out cross-validation.

A dummy-based user privacy protection approach for text information retrieval

Text retrieval enables people to efficiently obtain the desired data from massive text data, so has become one of the most popular services in information retrieval community. However, while providing great convenience for users, text retrieval results in a serious issue on user privacy. In this paper, we propose a dummy-based approach for text retrieval privacy protection. Its basic idea is to use well-designed dummy queries to cover up user queries and thus protect user privacy. First, we present a client-based system framework for the protection of user privacy, which requires no change to the existing algorithm of text retrieval, and no compromise to the accuracy of text retrieval. Second, we define a user privacy model to formulate the requirements that ideal dummy queries should meet, i.e., (1) having highly similar feature distributions with user queries, and (2) effectively reducing the significance of user query topics. Third, by means of the knowledge derived from Wikipedia, we present an implementation algorithm to construct a group of ideal dummy queries that can well meet the privacy model. Finally, we demonstrate the effectiveness of our approach by theoretical analysis and experimental evaluation. The results show that by constructing dummy queries that have similar feature distributions but unrelated topics with user queries, the privacy behind users’ textual queries can be effectively protected, under the precondition of not compromising the accuracy and usability of text retrieval.

Model Reuse in Machine Learning for Author Name Disambiguation: An Exploration of Transfer Learning

Machine learning for author name disambiguation is usually conducted on the training and test subsets of labeled data created for a specific task. As a result, disambiguation models learned on heterogeneous labeled data are often inapplicable for other purposes that either do not use the same labeled data or do not make use of any labeled data at all. This article explores the idea of transfer learning in a new context, author name disambiguation. We focus on cases where a disambiguation task lacking labeled training data uses models trained on labeled data generated for other tasks. For this purpose, two labeled source datasets are used for training of disambiguation models to be applied to three test target datasets that are deficient of labeled training data. Our results show that transfer learning can produce disambiguation performances similar to those achievable by traditional machine learning in which training and test datasets come from the same labeled data source. The good performance through transfer learning are possible when training source datasets have similar feature distributions as test target datasets. This study suggests that through transfer learning, rich disambiguation models in previous studies can be retained and reused across ambiguous bibliographic data from different fields and data sources, motivating further research on how to correct feature distribution differences between source and target datasets to expand the application of transfer learning in author name disambiguation beyond the model sharing explored in this research.

Domain Regeneration for Cross-Database Micro-Expression Recognition.

Recently, micro-expression recognition has attracted lots of researchers' attention due to its potential value in many practical applications, e.g., lie detection. In this paper, we investigate an interesting and challenging problem in micro-expression recognition, i.e., cross-database micro-expression recognition, in which the training and testing samples come from different micro-expression databases. Under this problem setting, the consistent feature distribution between the training and testing samples originally existing in conventional micro-expression recognition would be seriously broken and hence the performance of most current well-performing micro-expression recognition methods may sharply drop. In order to overcome it, we propose a simple yet effective framework called Domain Regeneration (DR) in this paper. DR framework aims at learning a domain regenerator to regenerate the micro-expression samples from source and target databases respectively such that they can abide by the same or similar feature distributions. Thus, we are able to use the classifier learned based on the labeled source micro-expression samples to predict the label information of the unlabeled target micro-expression samples. To evaluate the proposed DR framework, we conduct extensive cross-database micro-expression recognition experiments designed based on SMIC and CASME II databases. Experimental results show that compared with recent state-of-the-art cross-database emotion recognition methods, the proposed DR framework has more promising performance.

Contextual-based top-down saliency feature weighting for target detection

Target detection using attention models has recently become a major research topic in active vision. One of the major problems in this area of research is how to appropriately weight low-level features to get high quality top-down saliency maps that highlight target objects. Learning of such weights has previously been done using example images having similar feature distributions without considering contextual information. In this paper, we propose a model that we refer to as the top-down contextual weighting (TDCoW) that incorporates high-level knowledge of the gist context of images to apply appropriate weights to the features. The proposed model is tested on four challenging datasets, two for cricket balls, one for bikes and one for person detection. The obtained results show the effectiveness of contextual information for modelling the TD saliency by producing better feature weights than those produced without contextual information.

Discriminative sparse neighbor coding

Sparse coding has received extensive attention in the literature of image classification. Traditional sparse coding strategies tend to approximate local features in terms of a linear combination of basis vectors, without considering feature neighboring relationships. In this scenario, similar instances in the feature space may result in totally different sparse codes. To address this shortcoming, we investigate how to develop new sparse representations which preserve feature similarities. We commence by establishing two modules to improve the discriminative ability of sparse representation. The first module selects discriminative features for each class, and the second module eliminates non-informative visual words. We then explore the distribution of similar features over the dominant basis vectors for each class. We incorporate the feature distribution into the objective function, spanning a class-specific low dimensional subspace for effective sparse coding. Extensive experiments on various image classification tasks validate that the proposed approach consistently outperforms several state-of-the-art methods.

Middle Eocene Carnivoramorpha of the Uinta Basin, Utah

The middle Eocene of North American was an important time in carnivoramorphan evolution, because it is when the order showed increases in diversity taxonomically and ecomorphologically. We describe new fossils of “miacoids” from the Uinta Formation of northeastern Utah, and assess their phylogenetic and paleobiological significance. A new species of Miocyon (Matthew, 1909), is described, Miocyon vallisrubrae , which, in several aspects of lower molar morphology, is intermediate between the smaller Miocyon major (Matthew, 1909) and the larger Miocyon scotti (Wortman and Matthew, 1899). In addition, new well-preserved dental and gnathic specimens of M. scotti are described, along with fragmentary cranial and postcranial remains. The taxonomy of Miocyon is reviewed in light of the new material, which highlights the divergent specializations of the genus. These include an m1 with a relatively broad talonid and an m2 that is large, broad and relatively flat with little difference in height between the trigonid and talonid. These features suggest a shift from a shearing-dominated diet to a greater reliance on grinding. An ecological analogue for the radiation of this lineage among living carnivorans may be the sympatric African species of jackals. The distribution of similar features in Miocyon as well as other “miacoids” suggests an ecomorphologically similar phylogenetic group containing Miocyon, Uintacyon Leidy, 1872, and Vassacyon Matthew, 1909. New specimens from the Uinta Formation are also described of Tapocyon Stock, 1934, Procynodictis Wortman and Matthew, 1899, a new unnamed miacid, and an unnamed small viverravid. These highlight the adaptive diversity of carnivoramorphans during the Uintan North American Land Mammal Age.

Stratigraphy, Sedimentary Structures, and Textures of the Late Neoproterozoic Doushantuo Cap Carbonate in South China

Abstract The 3- to 5-m-thick Doushantuo cap carbonate in south China overlies the glaciogenic Nantuo Formation (ca. 635 Ma) and consists of laterally persistent, thinly laminated and normally graded dolomite and limestone indicative of relatively deep-water deposition, most likely below storm wave base. The basal portion of this carbonate contains a distinctive suite of closely associated tepee-like structures, stromatactis-like cavities, layer-parallel sheet cracks, and cemented breccias. The cores of tepees are composed of stacked cavities lined by cements and brecciated host dolomicrite. Onlap by laminated sediment indicates synsedimentary disruption of bedding that resulted in a positive seafloor expression. Cavities and sheet cracks contain internal sediments, and they are lined by originally aragonitic isopachous botryoidal cements with acicular radiating needles, now replaced by dolomite and silica. Pyrite and barite are common, and calcite is locally retained as a primary mineral. These features share morphological and petrographic attributes with modern and ancient methane seeps in which methane gas and fluids provide both a force for physical disruption from buoyancy and a source of alkalinity for significant cementation. The presence of δ13C values as low as −41‰ in well preserved limestone crusts and cements within and immediately above the tepee-like structures provides unequivocal evidence for methane influence, and the widespread distribution of identical sedimentary structures and paragenetic cement sequences across the entire basin at the same basal cap carbonate level is consistent with gas hydrate destabilization and the development of methane seeps as a result of postglacial warming of the ocean. Considering the broad distribution of similar features at the same stratigraphic level in other cap carbonates globally, we suggest that the late Neoproterozoic postglacial methane release may have influenced the oceanic oxygen level as well as contributed to postglacial warming via the greenhouse effects of methane.

Stagnancy of the Illinoian Glacial Lobe East of the Illinois and Mississippi Rivers

The hitherto unrecognized features of Leverett's ridged drift of the Kaskaskia Basin throughout its course from the Shelbyville moraine in south central Illinois nearly to Belleville in southwestern Illinois; its genetic crevasse ridges, subcrevasse channels on the land surface, and moulin kames; together with sparse but wide distribution of similar features over the greater part of the Illinoian drift area east of Illinois and Mississippi rivers attest to the stagnant condition of the Illinoian glacial lobe for this region after it had reached its terminus and made its marginal deposits.