学科服务 更多>>

关键词： Agricultural pest detection   Fusion features   Anchor-free   RCNN   Region proposals  

摘要： The frequent outbreaks of agricultural pests have resulted in the reduction of crop production and seriously restricted agricultural production. And many kinds of agricultural pests bring challenges to the accurate identification of agricultural pests for agricultural workers. Currently, the traditional methods of agricultural pest detection cannot satisfy the needs of agricultural production because of low efficiency and accuracy. In this paper, we put forward an anchor-free region convolutional neural network (AF-RCNN) for precision recognition and classification of 24-classes pests. First, a feature fusion module is designed to extract effective feature information of agricultural pests, especially small pests. Then, we propose an anchor-free region proposal network (AFRPN) that is used for getting high-quality object proposals as possible pest positions based on the fusion feature maps. Finally, our anchor-free region convolutional neural network (AF-RCNN) is employed to detect 24-classes pest via an end-to-end way by merging our AFRPN with Fast R-CNN into a single network. We evaluate the performance of our method on the pest dataset including 20k images and 24 classes. Experimental results demonstrate that our method is able to obtain 56.4% mAP and 85.1% mRecall on 24-classes pest dataset, 7.5% and 15.3% higher than Faster R-CNN, and 39.4% and 56.5% higher than YOLO detector. The running time could achieve 0.07 s per image, meeting the real-time detection. The proposed method is effective and applicable for accurate and real-time intelligent pest detection.

关键词： TI-RADS   Shear wave elastography   Contrast-enhanced ultrasound   Thyroid nodules  

摘要： Purpose To evaluate the diagnostic value of shear wave elastography (SWE) combined with contrast-enhanced ultrasonography (CEUS) in diagnosing thyroid imaging reporting and data system (TI-RADS) category 4a and 4b nodules. Methods TI-RADS, SWE, and CEUS features of 71 thyroid nodules (23 benign, 48 malignant) confirmed by postoperative pathological results were retrospectively analyzed. The diagnostic efficiency of each single method and that of a combination of three methods were compared. Results The sensitivity and specificity in diagnosing thyroid nodules were 70.83% and 65.22% for TI-RADS, 68.75% and 91.30% for SWE, 77.08% and 78.26% for CEUS, and 91.67% and 95.65% for TI-RADS + SWE + CEUS, respectively. The area under the curve for TI-RADS, SWE, CEUS, and TI-RADS + SWE + CEUS in diagnosing thyroid nodules were 0.680, 0.839, 0.799, and 0.937, respectively. A significant difference was observed between a combination of the three methods and any of them alone (p < 0.05). Conclusion Combining SWE and CEUS improves the differential diagnosis of TI-RADS category 4a and 4b nodules.

关键词： Categorical attributes   Coupling learning   Embedding learning   Heterogeneous fusion   Metric learning  

摘要： Categorical attributes are ubiquitous in real-world collected data. However, such attributes lack a well-defined distance metric and cannot be directly manipulated per algebraic operations, so many data mining algorithms are unable to work directly on them. Learning an appropriate metric or an effective numerical embedding is very vital yet challenging, for categorical attributes with multi-view heterogeneous data characteristics. This paper proposes a novel multi-view heterogeneous fusion model (MVHF), which first captures basic coupling information for each view and then fuses these heterogeneous information from different views by multi-kernel metric learning, to measure the intrinsic distances between this type of categorical attributes; based on these measured distances, further, we use the manifold learning method to learn a high-quality numerical embedding for each categorical value. Experiments on 33 mixed data sets demonstrate that MVHF-enabled classification significantly enhances the performance, compared with state-of-the-art distance metrics or embedding competitors. [ABSTRACT FROM AUTHOR]

摘要： Objectives. To evaluate the impact of duration and service category on HIV health outcomes among low-income adults living with HIV and enrolled in a housing program in 2014 to 2017. Methods. We estimated relative risk of engagement in care, viral suppression, and CD4 improvement for 561 consumers at first and second year after enrollment to matched controls through the New York City HIV surveillance registry, by enrollment length (enrolled for more than 1 year or not) and service category (housing placement assistance [HPA], supportive permanent housing [SPH], and rental assistance [REN]). Results. The SPH and REN consumers were enrolled longer and received more services, compared with HPA consumers. Long-term SPH and REN consumers had better engagement in care, viral suppression, and CD4 count than controls at both first and second year after enrollment, but the effect did not grow bigger from year 1 to 2. HPA consumers did not have better outcomes than controls regardless of enrollment length. Conclusions. Longer enrollment with timely housing placement and a higher number and more types of services are associated with better HIV health outcomes for lowincome persons living with HIV with unmet housing needs.

关键词： Vector coloring   Lovasz nu number   Categorical graph product   Semidefinite programming   Hedetniemi's conjecture  

摘要： A vector t-coloring of a graph is an assignment of real vectors p(1), ... , p(n) to its vertices such that p(i)(T) p(i) = t - 1, for all i = 1, ... , n and p(i)(T) p(j) <= -1, whenever i and j are adjacent. The vector chromatic number of G is the smallest number t >= 1 for which a vector t-coloring of G exists. For a graph H and a vector t-coloring p(1), ... , p(n) of G, the map taking (i, l) is an element of V(G) x V(H) to p(i) is a vector t-coloring of the categorical product G x H. It follows that the vector chromatic number of G x H is at most the minimum of the vector chromatic numbers of the factors. We prove that equality always holds, constituting a vector coloring analog of the famous Hedetniemi Conjecture from graph coloring. Furthermore, we prove necessary and sufficient conditions under which all optimal vector colorings of G x H are induced by optimal vector colorings of the factors. Our proofs rely on various semidefinite programming formulations of the vector chromatic number and a theory of optimal vector colorings we develop along the way, which is of independent interest.

关键词： Visual analytics   multidimensional feature space   epidemiology   quantitative and categorical variables  

摘要： With the advances in science and technology, a rapid growth of multidimensional (multivariate) datasets is observed in different fields. Projection and visualization of such data to a lower dimensional space without losing the data structure is a challenging task. We propose an interactive visual analytics tool that is applied for the combined analysis of multidimensional numerical and categorical data. The tool helps the analyst not only to find the clusters of similar objects but also to identify the important features specific to these clusters. The efficacy of the various functionalities of the tool is examined analyzing epidemiological data to understand the pathogenesis of obstructive sleep apnea. Our approach helps the user to visually analyze the data and get a better understanding of the data. The tool would be a valuable resource for analysts working on numerical and categorical data.

关键词： Bug severity   Stack traces   Machine learning mining software repositories   Software maintenance  

摘要： Context: The severity of a bug is often used as an indicator of how a bug negatively affects system functionality. It is used by developers to prioritize bugs which need to be fixed. The problem is that, for various reasons, bug submitters often enter the incorrect severity level, delaying the bug resolution process. Techniques that can automatically predict the severity of a bug can significantly reduce the bug triaging overhead. In our previous work, we showed that the accuracy of description-based severity prediction techniques could be significantly improved by using stack traces as a source of information. Objective: In this study, we expand our previous work by exploring the effect of using categorical features, in addition to stack traces, to predict the severity of bugs. These categorical features include faulty product, faulty component, and operating system. We experimented with other features and observed that they do not improve the severity prediction accuracy. A Software system is composed of many products;each has a set of components. Components interact with each to provide the functionality of the product. The operating system field refers to the operating system on which the software was running on during the crash. Method: The proposed approach uses a linear combination of stack trace and categorical features similarity to predict the severity. We adopted a cost sensitive K Nearest Neighbor approach to overcome the unbalance label distribution problem and improve the classifier accuracy. Results: Our experiments on bug reports of Eclipse submitted between 2001 and 2015 and Gnome submitted between 1999 and 2015 show that the accuracy of our severity prediction approach can be improved from 5% to 20% by considering categorical features, in addition to stack traces. Conclusion: The accuracy of predicting the severity of bugs is higher when combining stack traces and three categorical features, product, component, and operating system.

关键词： Online review   Motivating factors   Product sales   Product category   Meta-analysis   Moderating effect  

摘要： Online reviews, which significantly influence product sales, have been a central research topic in the field of marketing. Meanwhile, some motivating factors related to online retailers have been linked to product sales. While several articles have examined the impact between online reviews and motivating factors on product sales, many of the conclusions drawn are contradictory. From 28 studies focusing on online reviews and sales, this study performs a meta-analysis to analyze the true impacts of six review-related factors (i.e., the number of reviews, star ratings, standard deviation of ratings, helpfulness, review length and sentiment), and two motivating factors (i.e., price discounts and special shipping) on product sales. Meanwhile, this paper also studies how one product-related factor (i.e., product age) and one reviewer-related factor (i.e., reviewer's reputation) influence the relationship between online reviews and product sales. In addition, to study the moderating effect of product category, we divide the selected literature into two subgroups which are search and experience products. The results indicate that only review length and special shipping have no significant impact on product sales, while product category has a valid and specific moderating effect on the relationship between these determinants and sales. The presented conclusions will have important implications for academic research and for future industry practice.

关键词： Clinical endpoint prediction   Electronic health records   Categorical variables embedding  

摘要： Patients with advanced cancer are burdened physically and psychologically, so there is an urgent need to pay more attention to their health-related quality of life (HRQOL). With an expected clinical endpoint prediction, over-treatment can be effectively eliminated by the means of palliative care at the right time. This paper de-velops a deep learning based approach for cancer clinical endpoint prediction based on patient's electronic health records (EHR). Due to the pervasive existence of categorical information in EHR, it brings unavoidably obstacles to the effective numerical learning algorithms. To address this issue, we propose a novel cross-field categorical attributes embedding (CCAE) model to learn a vectorized representation for cancer patients in attribute-level by orders, in which the strong semantic coupling among categorical variables are well exploited. By transforming the order-dependency modeling into a sequence learning task in an ingenious way, recurrent neural network is adopted to capture the semantic relevance among multi-order representations. Experimental results from the SEER-Medicare EHR dataset have illustrated that the proposed model can achieve competitive prediction performance compared with other baselines.

关键词： Sodium   Sodium-modi fied foods   Potassium   Packaged food   Packaged Nutrient information  

摘要： Background Lowering excess sodium in packaged foods is part of a public health strategy to reduce cardiovascular disease risk. Sales of foods with labeled sodium claims increased during the past decade. Yet, it is unclear whether sugars or fats were added during the reformulation of foods that might counter the bene fits of sodium reduction. Objective It was hypothesized that the nutrient content of packaged foods with lower sodium label claims (ie, sodium -modi fied) would differ from their regular (ie, un- modi fied) counterparts. Design This cross-sectional study compared label data of 153 sodium -modi fied foods and 141 regular, matched counterparts within four food categories: soups, processed meats, vegetables, and savory snacks. Foods were identi fied by searching manufacturer websites of the top -10 brands in each category. Sodium, calories, total carbohydrate, sugar, protein, total fat, saturated fat, and potassium (when reported) were compared by labeled serving and per 100 g food. Results The average amount in milligrams of sodium per serving in regular foods ranged from 162 mg for savory snacks to 782 mg for soups. Compared with regular foods, the matched lower sodium foods had signi ficantly less sodium per serving (-95 to -387 mg) and per 100 g (-184 to -462 mg) ( P 0.01 for all comparisons), except for soups per 100 g ( P = 0.166), and were similar to their regular counterparts in calories, total carbohydrate, sugar, protein, total fat, and saturated fat ( P 0.05 for all compari- sons). Of the soups that reported potassium on the label, potassium was 244 mg/serving ( P=0.004) and 139 mg/100 g ( P=0.0 02) higher among matched lower sodium soups. Conclusions The similarity in macronutrient contents on the labels for sodium - modi fied foods and their regular counterparts suggests that reformulation did not include the addition of signi ficant amounts of sugars, fats, or other macronutrients among major food brands in the selected categories. Potassium c