Search

The imbalanced data classification problem widely exists in many real-world applications. Data resampling is a promising technique to deal with imbalanced data through either oversampling or undersampling. However, the traditional data resampling approaches simply take into account the local neighbor information to generate new instances in linear ways, leading to the generation of incorrect and unnecessary instances. In this study, we propose a new data resampling technique, namely, Gaussian Distribution based Oversampling (GDO). In GDO, anchor instances are selected from the minority class instances in a probabilistic way by taking into account the density and distance information carried by the minority instances. Then new minority instances are generated following a Gaussian distribution model. The proposed method is validated by comparing with seven imbalanced learning approaches on 40 data sets from the KEEL repository and 10 large data sets from the UCI repository. Experimental results show that our method outperforms the other compared methods in terms of AUC, G-mean and memory usage with an increase in running time. We also apply GDO to two real imbalanced data classification problems: Internet video traffic identification and metastasis detection of esophageal cancer. The experimental results once again validate the effectiveness of our approach.

In this study, the powders of the Ba0.75Sr0.25TiO3 (BST) nanoparticles were directly synthesized by milling of Ba(OH)2·8H2O, Sr(OH)2·8H2O, and Ti(BuO)4 in ethanol at room temperature. They have homogenous grains of ~ 15 nm and high sintering activity. The dense ceramics with the density > 90% can be obtained at a sintering temperature of ≤ 950 °C by adding 3 wt% sintering aids of Bi2O3 and Li2CO3. Several Bi-related intermediate compounds act as perovskite-structured templates to sintering the ceramics at a different temperature. They enhance the mass transfer and promote the sintering densification. These compounds such as Ba2BiO4 and SrBiO4 appear at 800 °C, LiBa4Bi3O11 and Sr1.2Bi0.8O3 appear over 830 °C, and Bi8.11Ba0.89O13.05 appears at 950 °C. The cation Bi in them can have mixture valences of 3+ and 5+. It makes the ceramics as semiconducting state with the dark gray color and decreases the ceramic resistivities. With the sintering temperature increase, especially at 950 °C, the cation Bi tends back to single valence of +3 in the ceramics. The most of alkaline earth cations in Bi-related compounds will release and resorb into the lattice of BST and drive the sintering densification. The BST ceramics can have a peak dielectric constant > 6500 (at 53 °C) with loss 1012 Ω cm when sintered at a temperature of ≥ 900 °C with 3 wt% sintering aids. They have a potential application for multiple layer ceramic capacitors (MLCC) with silver inner electrodes.

Introduction Decreasing sleep quality is a challenge for older adults over age 65. Researchers have found significant associations between romantic relationships and sleep quality. However, most existing studies are cross-sectional; few have focused on the aging population who experience the most sleep problems; few have considered mechanisms linking close relationships and sleep quality; few have considered the specific effects of social interactions on sleep quality; and few have considered bidirectional relationships. To address these gaps in the literature, this study investigated the bidirectional pathways linking daily partner interactions to sleep quality in older adult couples. This study also tested daily negative affect, positive affect, and perceived partner responsiveness as three possible mediators explaining recursive associations between daily partner interactions and sleep quality. Methods This study recruited 238 older adult couples to complete baseline questionnaires including assessments of attachment orientation and demographics. Couple members also completed daily diaries for 7 consecutive days, reporting on their daily relationship, emotional, and sleep experiences. Results Results supported the predicted bidirectional relationship between daily partner interactions and sleep quality via three hypothesized pathways (i.e., negative affect, positive affect, and perceived partner responsiveness). Specifically, older adults who reported higher than average negative partner interactions during the day reported increases in negative affect, and decreases in positive affect and perceived partner responsiveness on the same day, which in turn were associated with worse sleep quality that night. The opposite was true for positive partner interactions. Bidirectionally, older adults who reported higher than average sleep quality on one night reported decreases in negative affect and increases in positive affect and perceived partner responsiveness the next day, which in turn were associated with fewer negative partner interactions and more positive partner interactions that day. Results also revealed moderating effects of attachment orientation and gender. Conclusion In conclusion, this study contributes to a sparse literature assessing the bidirectional relationship between daily partner interactions and sleep quality in the aging population. This research adds to our knowledge of the mechanisms underlying the recursive relationship and informs us of potential avenues for relationship-based interventions aimed at improving sleep quality. Support (if any)

Based on the complementary potential energy variational principle, in this work, we proposed a stress-driven homogenization procedure to compute overall effective material properties for elastic composites with locally heterogenous micro-structures. We have developed a novel incremental variational formulation for homogenization problems of both infinitesimal and finite deformations where the macro-stress-based complementary potential energy is obtained for hyperelastic materials for a global minimization problem with respect to fine-scale displacement fluctuation field. The point of departure of our approach is a general complementary variational principle formulation that can determine material responses of elastic composites with heterogeneous micro structures. We have implemented the proposed stress-driven computational homogenization procedure with the finite element method. By comparing the numerical results with the analytical method and the strain-driven homogenization method, we find that the stress-driven homogenization offers the lower bound estimate of materials properties for elastic composites.

In the last decade, the increasing video traffic, especially illegal videos brought big challenges for Internet management. Generally, abnormal videos, such as illegal videos only account for a small percentage which makes the detection of such videos to be a typical imbalanced classification problem. In this study, we propose a new imbalanced learning method, namely, the imbalanced data gravitation classification model based the gradient descent (IDGC-GD), to handle imbalanced problems. In IDGC-GD model, we use the gradient descent algorithm to optimize feature weights of the imbalanced data gravitation classification (IDGC) model. Then, we try to build an accurate video traffic identification solution using IDGC-GD. We conduct a set of comparing experiments between IDGC-GD and seven imbalanced learning algorithms using 21 open data sets and four video traffic data sets collected from the real application. Experimental results show that our method is promising for solving imbalanced problems, including Internet video traffic identification.

Recently, China advanced a policy of establishing a protected area system with national parks as the main body, and so the development of protected areas has entered a transitional period. However, to promote the coordinated development of economic construction and ecological protection of protected areas, their management needs to be more comprehensive and refined. Therefore, a more quantitative and refined spatial planning and management method for protected areas is urgently needed. This study took Sanjiangyuan National Park as the research object and considered the three biodiversity elements of species, ecosystems and landscape. The maximum entropy (MaxEnt) model, Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model, and Fragstats software were used to determine the areas with highly suitable species habitat, areas of high ecosystem service value and areas of high landscape diversity. Based on these areas, Marxan software was used to calculate the irreplaceability value, identify the boundary of Sanjiangyuan National Park and clarify the gaps of the existing Sanjiangyuan National Park in the space. We suggest that at least 18,569 km

Layout representation, which models visual elements and their inter-relations in a canvas, plays a crucial role in graphic design intelligence. With a large variety of layout designs and the unique characteristic of layouts that visual elements are defined as a list of categorical (e.g., type) and numerical (e.g., position and size) properties, it is challenging to learn general and compact representations with limited data. Inspired by the recent success of self-supervised pre-training techniques in various natural language processing tasks, in this paper, we propose CanvasEmb (Canvas Embedding), which pre-trains deep representations from unlabeled graphic designs by jointly conditioning on all the context elements in a canvas, with a multi-dimensional feature encoder and a multi-task learning objective. The pre-trained CanvasEmb model can be fine-tuned with just one additional output layer and with a small size of training data to create models for a wide range of downstream tasks. We verify our approach with presentation slides data. We construct a large-scale dataset with more than one million slides and propose two layout understanding tasks with human-labeled sets, namely element role labeling and image captioning. Evaluation results on these two tasks show that our model with fine-tuning achieves state-of-the-art performance. Furthermore, we conduct a deep analysis aiming to understand the modeling mechanism of CanvasEmb and demonstrate its great potential with two extended applications: layout auto completion and layout retrieval.

In the context of today’s rapid economic development and increasingly strict ecological protection, interactions between humans and wildlife will intensify. Therefore, we need to identify the spatial distribution of such interactions and to take measures to mitigate possible negative interactions. Using the Three-Rivers Source Region as a case study, we combined species habitat suitability, species feeding habits, species conservation levels and human impact intensity to construct a “human-wildlife spatial interaction index” (HWSII) to evaluate the intensity of human-wildlife spatial interactions. The overall level of HWSII was relatively low in the Three-River Source Region. Higher potential human-wildlife interactions occurred in areas with roads and railways. Areas of relatively high interactions between carnivores and livestock were mainly located in the southern part of the study area, and those with relatively high interactions between herbivores and livestock appeared in the western and eastern parts. Managers in these areas should consider mitigation measures. Although developed for a specific region of China, the ‘human-wildlife spatial interaction index’ could have application in other areas of the world for setting a baseline reference to identify areas with the potential for high levels of human-wildlife interactions and pinpoint places where potential negative interactions can be mitigated.

There is emerging evidence for links between relationship factors and sleep quality. Existing research linking attachment orientation to sleep quality has yielded inconsistent effects, has focused on younger samples, and has not considered underlying mechanisms of action. This research addressed these gaps in two studies that investigated the links between attachment orientation and sleep quality in both younger/middle-aged (Study 1) and older (Study 2) adult couples using Actor–Partner Interdependence Models. We also tested mediating effects of relationship-specific security and negative affect. In both studies, participants completed surveys assessing their attachment orientation, sleep quality, and the proposed mediators. Both studies revealed that relationship-specific security and negative affect mediated the negative association between insecure attachment and one’s own sleep quality. This research enhances our understanding of how attachment orientation affects sleep quality, provides a foundation for future research on relationship influences on sleep, and suggests avenues for improving sleep quality.

Understanding dislocation multiplication, interactions, and hence crystal plasticity has been a long-standing challenge of major scientific significance. Combining atomistic enriched constitutive law with continuum finite element formulation, we developed a multiscale framework to capture geometrically-compatible dislocation patterns in Body Centered Cubic (BCC) single crystals. This dislocation pattern dynamics framework was successfully applied to study crystal plasticity in Face Centered Cubic (FCC) crystals before, coined as the Multiscale Crystal Defect Dynamics (MCDD) method. Inheriting from MCDD method, we now established a multiscale dislocation pattern dynamics model for BCC crytals with the following novelties: (1) Comparing with molecular dynamics approach, MCDD can predict crystal plasticity in BCC crystals in a more efficient way and in larger scale; (2) MCDD can capture the geometrically-compatible dislocation pattern distribution and evolution in BCC crystals, and (3) MCDD can capture the size effect of crystal plasticity in single BCC crystals at nanoscale, or to simulate size-dependent plasticity in BCC single crystals. In particular, we have successfully simulated crystal plasticity in a BCC crystal at the sub-micron scale by using a serial computing code in a desktop environment.

Since 2013, the Chinese government has issued a set of action plans catered toward the ecological civilization construction. To quantify and evaluate the ecological civilization development, this study has introduced the ecological civilization index of China (ECI), using different standardization methods (extreme standardization, single-benchmark progressive standardization, and double benchmark progressive standardization), to evaluate ecological civilization development on city-level (337 cities) of China. The issues of China's ecological civilization construction were analyzed. The evaluation results suggest that in the double benchmark progressive method, every evaluation score has its practical meaning and is therefore suitable for policy making. This study obtained an average ECI score of 61.16, for 337 cities. This is China's overall ECI score, which is classified as C-grade. Only 4.75% of cities' green environment (GE) score obtained A-grade, while 53.41% of cities' GE scored below the passing level, suggesting that the quality of the ecology and environment is neglected in the development of China eco-civilization. This study provides valuable knowledge for both the ecological and environment management and eco-civilization construction.

Imbalanced data classification, an important type of classification task, is challenging for standard learning algorithms. There are different strategies to handle the problem, as popular imbalanced learning technologies, data level imbalanced learning methods have elicited ample attention from researchers in recent years. However, most data level approaches linearly generate new instances by using local neighbor information rather than based on overall data distribution. Differing from these algorithms, in this study, we develop a new data level method, namely, generative learning (GL), to deal with imbalanced problems. In GL, we fit the distribution of the original data and generate new data on the basis of the distribution by adopting the Gaussian mixed model. Generated data, including synthetic minority and majority classes, are used to train learning models. The proposed method is validated through experiments performed on real-world data sets. Results show that our approach is competitive and comparable with other methods, such as SMOTE, SMOTE-ENN, SMOTE-TomekLinks, Borderline-SMOTE, and safe-level-SMOTE. Wilcoxon signed rank test is applied, and the testing results show again the significant superiority of our proposal.

In this work, a variational Bayesian learning-based computation algorithm is developed to “inversely” identify the deformation field of a crashed car and hence their residual strain fields based on its final damaged structural configuration (wreckage), which is important in three-dimensional traffic collision reconstruction and its forensic analysis. Different from our previous Generalized Bayesian Regularization Network (GBRN) algorithm (Xie et al. [2002] Computational Mechanics 69, 1191–1212), the present method is based on Variational Bayesian Learning theory coupled with a Feed-forward Neural Network architecture, and it provides a higher computation efficiency. This is because it requires less number of iterations and produces more accurate registration results, since the locality of nodes are greatly preserved during registration process. In this work, we have demonstrated that the developed machine learning algorithm has a unique capability to practically identify the deformation field of a real crashed car and to recover its initial pre-crash state based on residual damaged geometric configuration, and it shows great potential in forensic analysis of car crash and vehicle crashworthiness evaluation.

Burial of organic carbon (OC) in rift lakes on plateau is an important part of the global cycle. It is affected by natural and anthropogenic factors. In this study, we selected the sediment records of 7 rift lakes on the Yunnan-Guizhou Plateau to study spatial-temporal variation, sources and driving factors of organic carbon burial since 1850. The analysis of the temporal and spatial trend of carbon burial shows that the TOC concentration, TOC flux, C: N and mass accumulate rate have increased significantly since 1850. Co-occurrence network analysis indicated that a strong correlation between the TOC concentration and silty. TOC concentration were identified as core genera due to their high concentration. Carbon isotope tracing results show that before 1950, endogenous OC input played a dominant role, and after 1950, the proportion of exogenous OC increased. Canonical correlation analysis indicated that after 1950, agriculture intensification and population increase become one of the factors affecting the carbon burial of lakes in this area. The result of this study indicate that anthropogenic factors have become the main factors promoting carbon burial in rift lakes on the plateau.

This paper proposes the problem of Deep Question Generation (DQG), which aims to generate complex questions that require reasoning over multiple pieces of information of the input passage. In order to capture the global structure of the document and facilitate reasoning, we propose a novel framework which first constructs a semantic-level graph for the input document and then encodes the semantic graph by introducing an attention-based GGNN (Att-GGNN). Afterwards, we fuse the document-level and graph-level representations to perform joint training of content selection and question decoding. On the HotpotQA deep-question centric dataset, our model greatly improves performance over questions requiring reasoning over multiple facts, leading to state-of-the-art performance. The code is publicly available at https://github.com/WING-NUS/SG-Deep-Question-Generation., ACL 2020 (long paper)

Recent question generation (QG) approaches often utilize the sequence-to-sequence framework (Seq2Seq) to optimize the log-likelihood of ground-truth questions using teacher forcing. However, this training objective is inconsistent with actual question quality, which is often reflected by certain global properties such as whether the question can be answered by the document. As such, we directly optimize for QG-specific objectives via reinforcement learning to improve question quality. We design three different rewards that target to improve the fluency, relevance, and answerability of generated questions. We conduct both automatic and human evaluations in addition to a thorough analysis to explore the effect of each QG-specific reward. We find that optimizing question-specific rewards generally leads to better performance in automatic evaluation metrics. However, only the rewards that correlate well with human judgement (e.g., relevance) lead to real improvement in question quality. Optimizing for the others, especially answerability, introduces incorrect bias to the model, resulting in poor question quality. Our code is publicly available at https://github.com/YuxiXie/RL-for-Question-Generation., Comment: COLING 2020 (long paper)

Elevated evidences show that microRNAs (miRNAs) play vital roles in tumor progression regulation. However, the functional role of let-7b in hepatocellular carcinoma (HCC) is still largely unknown. In this study, we try to investigate the biological activity of let-7b in human HCC cells and try to find the potential regulatory signaling pathway. Our results indicate that let- 7b was remarkably down-regulated in human HCC tissues by qRT-PCR. In addition, let-7b overexpression decreased the expression of β-catenin and c-Myc, while upregulated E-cadherin expression in HCC cells which was verified by quantitative real-time PCR (qRT-PCR) and western blotting. Furthermore, Wnt/β-catenin was involved in let-7b biological activity which was revealed by luciferase assay. Moreover, Wnt/β-catenin signaling inhibitor blocks HCC cell proliferation which is as the same pattern as let-7b overexpression inhibits in HCC cells proliferation. In conclusion, down-regulated let-7b promotes HCC cell proliferation through Wnt/β-catenin signaling in HCC cells. These results suggested that appropriate manipulation of let-7b might be a new treatment of human HCC in the future. Keywords: Let-7b, Wnt/β-catenin signaling, Cell proliferation