Search

Engaging the public in protecting water resources is a critical yet challenging task. A wealth of social science studies has identified psychological predictors for individual pro-environmental behaviors. These predictors can guide communication in public engagement and inform the allocation of engagement efforts. However, a thorny challenge is to select influential factors among many candidates. This paper addresses this challenge by using social science research to guide the development of strategies to motivate the public to protect the North American Great Lakes. We considered a variable selection technique, the LASSO regression, in the post-hoc analysis of the International Joint Commission's 2018 Binational Great Lakes Binational Poll data. The poll surveyed 4250 Canadian and U.S. residents of the Great Lakes basin. We fit LASSO logistic models to predict respondents' intentions to take three public actions to protect the Great Lakes, including contacting public officials, attending public meetings, and engaging in online forums and groups. The models included 41 predictors encompassing demographic characteristics as well as respondents' awareness, beliefs, and values that are pertinent to Great Lakes policy development and management. Results revealed eight variables that consistently predicted the three public actions, including indigenous status, political ideology, impacts of the specific policy issues of nuclear wastes, policy awareness and interests, and the Great Lakes values for personal benefits and wildlife. Based on these findings, we recommend strategies to motivate the public to take public actions to protect the Great Lakes.

With maximum atom-utilization efficiency, single atom catalysts (SACs) are surging as a new research frontier in catalysis science. However, fabricating SACs and maintaining their thermodynamic stability remain challenging and thus uneconomical, largely due to the lack of fundamental understanding of the formation and stabilization mechanisms of single atoms (SAs). Through systematic density functional theory (DFT) calculations and machine learning algorithms, we present a rational design guidance for the feasibility of transition metal SA formation on a defective carbon surface and the oxygen reduction reaction (ORR) activity of the resulting SAs. We show that the dispersion of a metal nanoparticle (NP) into a stable array of SAs on a defective carbon surface is governed by the decomposition energy barrier of the NP and the diffusion barrier of SAs on a carbon surface. An intrinsic descriptor that correlates the catalytic activity of a SAC with the topological, bonding, and electronic structures of the SAC and its bonding carbon defect site is revealed. The few-shot machine learning algorithm further enables a 130 000-fold reduction of the time needed to calculate the ORR activity of SACs from DFT, and thus allows us to predict the ORR activity of SACs of all transition metals within an error of 8.33%. The results from this study offer a mechanistic and quantitative guidance for rational selection of transition metal and optimal synthesis conditions to fabricate SACs with desirable electrocatalyst activity in emerging energy applications.

BACKGROUND Odontogenic infection is one of the common infectious diseases in oral and maxillofacial head and neck regions. Clinically, if early odontogenic infections such as acute periapical periodontitis, alveolar abscess, and pericoronitis of wisdom teeth are not treated timely, effectively and correctly, the infected tissue may spread up to the skull and brain, down to the thoracic cavity, abdominal cavity and other areas through the natural potential fascial space in the oral and maxillofacial head and neck. Severe multi-space infections are formed and can eventually lead to life-threatening complications (LTCs), such as intracranial infection, pleural effusion, empyema, sepsis and even death. CASE SUMMARY We report a rare case of death in a 41-year-old man with severe odontogenic multi-space infections in the oral and maxillofacial head and neck regions. One week before admission, due to pain in the right lower posterior teeth, the patient placed a cigarette butt dipped in the pesticide "Miehailin" into the "dental cavity" to relieve the pain. Within a week, the infection gradually spread bilaterally to the floor of the mouth, submandibular space, neck, chest, waist, back, temporal and other areas. The patient had difficulty breathing, swallowing and eating, and was transferred to our hospital as an emergency admission. Following admission, oral and maxillofacial surgeons immediately organized consultations with doctors in otolaryngology, thoracic surgery, general surgery, hematology, anesthesia and the intensive care unit to assist with treatment. The patient was treated with the highest level of antibiotics (vancomycin) and extensive abscess incision and drainage in the oral, maxillofacial, head and neck, chest and back regions. Unfortunately, the patient died of septic shock and multiple organ failure on the third day after admission. CONCLUSION Odontogenic infection can cause serious multi-space infections in the oral and maxillofacial head and neck regions, which can result in multiple LTCs. The management and treatment of LTCs such as multi-space infections should be multidisciplinary led by oral and maxillofacial surgeons.

Cloud computing provides a powerful yet low-cost environment for distributed deep learning workloads. However, training complex deep learning models often requires accessing large amounts of data, which can easily exceed the capacity of local disks. Prior research often overlooks this training data problem by implicitly assuming that data is available locally or via low latency network-based data storage. Such implicit assumptions often do not hold in a cloud-based training environment, where deep learning practitioners create and tear down dedicated GPU clusters on demand, or do not have the luxury of local storage, such as in serverless workloads. In this work, we investigate the performance of distributed training that leverages training data residing entirely inside cloud storage buckets. These buckets promise low storage costs, but come with inherent bandwidth limitations that make them seem unsuitable for an efficient training solution. To account for these bandwidth limitations, we propose the use of two classical techniques, namely caching and pre-fetching, to mitigate the training performance degradation. We implement a prototype, DELI, based on the popular deep learning framework PyTorch by building on its data loading abstractions. We then evaluate the training performance of two deep learning workloads using Google Cloud's NVIDIA K80 GPU servers and show that we can reduce the time that the training loop is waiting for data by 85.6%-93.5% compared to loading directly from a storage bucket - thus achieving comparable performance to loading data directly from disk - while only storing a fraction of the data locally at a time. In addition, DELI has the potential of lowering the cost of running a training workload, especially on models with long per-epoch training times., To appear in IC2E 2021

Ever since the commercial offerings of the Cloud started appearing in 2006, the landscape of cloud computing has been undergoing remarkable changes with the emergence of many different types of service offerings, developer productivity enhancement tools, and new application classes as well as the manifestation of cloud functionality closer to the user at the edge. The notion of utility computing, however, has remained constant throughout its evolution, which means that cloud users always seek to save costs of leasing cloud resources while maximizing their use. On the other hand, cloud providers try to maximize their profits while assuring service-level objectives of the cloud-hosted applications and keeping operational costs low. All these outcomes require systematic and sound cloud engineering principles. The aim of this paper is to highlight the importance of cloud engineering, survey the landscape of best practices in cloud engineering and its evolution, discuss many of the existing cloud engineering advances, and identify both the inherent technical challenges and research opportunities for the future of cloud computing in general and cloud engineering in particular., Comment: author copy/preprint of a paper published in the IEEE International Conference on Cloud Engineering (IC2E 2021)

This study identifies and proposes techniques to alleviate two key bottlenecks to executing deep neural networks in trusted execution environments (TEEs): page thrashing during the execution of convolutional layers and the decryption of large weight matrices in fully-connected layers. For the former, we propose a novel partitioning scheme, y-plane partitioning, designed to (i) provide consistent execution time when the layer output is large compared to the TEE secure memory; and (ii) significantly reduce the memory footprint of convolutional layers. For the latter, we leverage quantization and compression. In our evaluation, the proposed optimizations incurred latency overheads ranging from 1.09X to 2X baseline for a wide range of TEE sizes; in contrast, an unmodified implementation incurred latencies of up to 26X when running inside of the TEE., To Appear in the 9th IEEE International Conference on Cloud Engineering (IC2E 21)

Multiprincipal element alloys (MPEAs) have gained surging interest due to their exceptional properties unprecedented in traditional alloys. However, identifying an MPEA with desired properties from a huge compositional space via a cost-effective design remains a grand challenge. To address this challenge, the authors present a highly efficient design strategy of MPEAs through a coherent integration of molecular dynamics (MD) simulation, machine learning (ML) algorithms, and genetic algorithm (GA). The ML model can be effectively trained from 54 MD simulations to predict the stiffness and critical resolved shear stress (CRSS) of CoNiCrFeMn alloys with a relative error of 2.77% and 2.17%, respectively, with a 12 600-fold reduction of computation time. Furthermore, by combining the highly efficient ML model and a multi-objective GA, one can predict 100 optimal compositions of CoNiCrFeMn alloys with simultaneous high stiffness and CRSS, as verified by 100 000 ML-accelerated predictions. The highly efficient and precise design strategy can be readily adapted to identify MPEAs of other principal elements and thus substantially accelerate the discovery of other high-performance MPEA materials.

In order to improve the economy of island integrated energy system, an optimization model of electric-gas-thermal coupling integrated island energy system operation considering the demand response of building is proposed. The building thermal inertia model including HVAC load and integrated energy system model including gas turbine, diesel generator, wind turbine, electric boiler and gas storage tank are modeled. Considering the coupling property of time-sharing electricity price, gas price and multi energy, the demand side management is carried out through building thermal inertia, and the minimum operating cost of the system is taken as the goal. The yalmip toolbox and cplex solver are used to solve the model and obtain the power of each equipment and load in the scheduling period. The results show that the demand response of the buildings can play a role of peak load reduction and valley filling on the premise of ensuring the user's requirements for indoor temperature comfort, stabilize the fluctuation of wind power, and make the supply and demand more balanced. Through the coordinated interaction of the source, net and load, as well as the multiple energy sources on the supply side support and reserve each other, the efficiency of energy interactive utilization gets improved, and the system operation cost is reduced as well.

With the rapid development of autonomous driving, real-time object detection on 360° images becomes more and more important. In this paper, we propose a panoramic virtual dataset for training object detectors on 360° images. The most important feature of our dataset includes (1) an auto-generated city scene is created for rendering 360° dataset. (2) annotation work for this dataset is automatic. In addition, we propose a modified YOLOv3 model called Pano-YOLO for real-time panoramic object detection. Compared with YOLOv3, mAP of Pano-YOLO drops 0.39%. While speed is 32.47% faster. Experiments are performed to show that models trained on our virtual dataset can be applied in real world. And Pano-YOLO is capable of real-time object detection task on high-resolution 360° panoramic images and videos.

For many freshwater systems, mitigating agricultural runoff of nutrients is a key requirement for curbing eutrophication and reducing subsequent ecological threats. However, defining the best way to achieve reductions in agricultural runoff can be a contentious issue. A policy debate is currently unfolding in Ohio focused on whether the state government should introduce regulatory policies on agriculture to reduce nutrient loadings from watersheds in an attempt to also reduce harmful algal blooms in Lake Erie. To inform policy development, we used a survey instrument to gauge public acceptance of regulatory policies and examined the psychological determinants of Ohio residents’ support for a regulatory policy proposal that would introduce fines on excessive agricultural runoff. We designed a survey instrument with nine predictors of people’s willingness to support regulations: 1) effectiveness of voluntary programs, 2) risk perception, 3) water quality perception, 4) trust in farmers, 5) trust in state government, 6) belief about fertilizer runoff as a major cause of HABs, 7) belief that farmers alone should not bear the burden to restore water quality in Lake Erie, 8) belief that regulation is necessary to keep farmers accountable, and 9) belief that regulation harms economy and employment. We also measured variables that represented different levels of self-interests, awareness of reduction goals, political party affiliation, and demographic characteristics. We collected a sample of 1000 respondents, who were representative of Ohio residents by age, gender, race, and education level. Most predictors were significant and in the directions hypothesized, with exception of water quality perception and belief about regulation and jobs. One’s a priori belief that regulations are necessary to keep farmers accountable for their land management practices had the largest enhancing effect for accepting a regulatory policy of fines, while trust for farmers had the largest inhibiting effect. In comparison, water quality perception was not significant in predicting individual policy attitudes. This study informs the public engagement and communication efforts and suggest directions for future research on public policy support.

Popular big data frameworks, ranging from Hadoop MapReduce to Spark, rely on garbage-collected languages, such as Java and Scala. Big data applications are especially sensitive to the effectiveness of garbage collection (i.e., GC), because they usually process a large volume of data objects that lead to heavy GC overhead. Lacking in-depth understanding of GC performance has impeded performance improvement in big data applications. In this paper, we conduct the first comprehensive evaluation on three popular garbage collectors, i.e., Parallel, CMS, and G1, using four representative Spark applications. By thoroughly investigating the correlation between these big data applications' memory usage patterns and the collectors' GC patterns, we obtain many findings about GC inefficiencies. We further propose empirical guidelines for application developers, and insightful optimization strategies for designing big-data-friendly garbage collectors.

Omnidirectional lighting provides the foundation for achieving spatially-variant photorealistic 3D rendering, a desirable property for mobile augmented reality applications. However, in practice, estimating omnidirectional lighting can be challenging due to limitations such as partial panoramas of the rendering positions, and the inherent environment lighting and mobile user dynamics. A new opportunity arises recently with the advancements in mobile 3D vision, including built-in high-accuracy depth sensors and deep learning-powered algorithms, which provide the means to better sense and understand the physical surroundings. Centering the key idea of 3D vision, in this work, we design an edge-assisted framework called Xihe to provide mobile AR applications the ability to obtain accurate omnidirectional lighting estimation in real time. Specifically, we develop a novel sampling technique that efficiently compresses the raw point cloud input generated at the mobile device. This technique is derived based on our empirical analysis of a recent 3D indoor dataset and plays a key role in our 3D vision-based lighting estimator pipeline design. To achieve the real-time goal, we develop a tailored GPU pipeline for on-device point cloud processing and use an encoding technique that reduces network transmitted bytes. Finally, we present an adaptive triggering strategy that allows Xihe to skip unnecessary lighting estimations and a practical way to provide temporal coherent rendering integration with the mobile AR ecosystem. We evaluate both the lighting estimation accuracy and time of Xihe using a reference mobile application developed with Xihe's APIs. Our results show that Xihe takes as fast as 20.67ms per lighting estimation and achieves 9.4% better estimation accuracy than a state-of-the-art neural network.

Executing deep-learning inference on cloud servers enables the usage of high complexity models for mobile devices with limited resources. However, pre-execution time-the time it takes to prepare and transfer data to the cloud-is variable and can take orders of magnitude longer to complete than inference execution itself. This pre-execution time can be reduced by dynamically deciding the order of two essential steps, preprocessing and data transfer, to better take advantage of on-device resources and network conditions. In this work, we present PieSlicer, a system for making dynamic preprocessing decisions to improve cloud inference performance using linear regression models. PieSlicer then leverages these models to select the appropriate preprocessing location. We show that for image classification applications PieSlicer reduces median and 99th percentile pre-execution time by up to 50.2ms and 217.2ms respectively when compared to static preprocessing methods.

Stochastic Gradient Descent (SGD) has become the de facto way to train deep neural networks in distributed clusters. A critical factor in determining the training throughput and model accuracy is the choice of the parameter synchronization protocol. For example, while Bulk Synchronous Parallel (BSP) often achieves better converged accuracy, the corresponding training throughput can be negatively impacted by stragglers. In contrast, Asynchronous Parallel (ASP) can have higher throughput, but its convergence and accuracy can be impacted by stale gradients. To improve the performance of synchronization protocol, recent work often focuses on designing new protocols with a heavy reliance on hard-to-tune hyper-parameters. In this paper, we design a hybrid synchronization approach that exploits the benefits of both BSP and ASP, i.e., reducing training time while simultaneously maintaining the converged accuracy. Based on extensive empirical profiling, we devise a collection of adaptive policies that determine how and when to switch between synchronization protocols. Our policies include both offline ones that target recurring jobs and online ones for handling transient stragglers. We implement the proposed policies in a prototype system, called Sync-Switch, on top of TensorFlow, and evaluate the training performance with popular deep learning models and datasets. Our experiments show that Sync-Switch achieves up to 5.13X throughput speedup and similar converged accuracy when comparing to BSP. Further, we observe that Sync-Switch achieves 3.8% higher converged accuracy with just 1.23X the training time compared to training with ASP. Moreover, Sync-Switch can be used in settings when training with ASP leads to divergence errors. Sync-Switch achieves all of these benefits with very low overhead, e.g., the framework overhead can be as low as 1.7% of the total training time., Comment: 15 pages, 16 figures, 6 tables, ICDCS'21

It has been nearly eleven years since the reform of China’s medical and health system entered the final sprint stage in early 2008. During this period, great changes have taken place in China’s medical system, which has had a huge impact on the operation of public hospitals in China. This paper analyzes the key financial indicators of public hospitals in Chongqing from 2007 to 2017, including revenue and expenditure situation and structure, assets and liabilities situation and structure, finds out the problems in operation and puts forward suggestions. The key financial indicators of public hospitals in Chongqing were analyzed and evaluated by trend analysis, ratio analysis and DuPont analysis. The total revenue and expenditure of public hospitals in Chongqing increases with GDP year by year, the structure of revenue and expenditure has changed greatly, the debt level is reasonable, and the operating capacity is at a low level. The balance of drug revenue and expenditure in public hospitals is unbalanced, so it is necessary to strengthen the implementation of drug price adjustment plan; Public hospitals should strengthen the cost control, especially the cost control of drug expenditure, in order to meet the requirements of national policies and improve their own operating capacity.

Summary The oxygen evolution reaction (OER) is a critical reaction for energy-related applications, yet suffers from its slow kinetics and large overpotential. It is desirable to develop effective OER electrocatalysts, such as single-atom catalysts (SACs). Here, we demonstrate machine learning (ML)-accelerated prediction of OER overpotential of all transition metals. Based on density functional theory (DFT) calculations of 15 species of SACs, we design a topological information-based ML model to map the OER overpotentials with atomic properties of the corresponding SACs. The trained ML model not only yields remarkable prediction precision (relative error of 6.49%) but also enables a 130,000-fold reduction of prediction time in comparison with pure DFT calculation. Furthermore, an intrinsic descriptor that correlates the overpotential of an SAC with its atomic properties is revealed. The approach and results from this study can be readily applicable to screen other SACs and significantly accelerate the design of high-performance catalysts for many other reactions., Graphical abstract, Highlights • We present a topology-based machine learning (ML) approach to predict OER activity • The prediction by the ML model is of high precision (relative error of 6.49%). • The ML model is 130,000 times faster than pure density function theory calculation, Artificial Intelligence; Catalysis; Electrochemistry; Energy Materials

Volumetric video (VV) streaming has drawn an increasing amount of interests recently with the rapid advancements in consumer VR/AR devices and the relevant multimedia and graphics research. While the resource and performance challenges in volumetric video streaming have been actively investigated by the multimedia community, the potential security and privacy concerns with this new type of multimedia have not been studied. We for the first time identify an effective threat model that extracts 3D face models from volumetric videos and compromises face ID-based authentications To defend against such attack, we develop a novel volumetric video security mechanism, namely VVSec, which makes benign use of adversarial perturbations to obfuscate the security and privacy-sensitive 3D face models. Such obfuscation ensures that the 3D models cannot be exploited to bypass deep learning-based face authentications. Meanwhile, the injected perturbations are not perceivable by the end-users, maintaining the original quality of experience in volumetric video streaming. We evaluate VVSec using two datasets, including a set of frames extracted from an empirical volumetric video and a public RGB-D face image dataset. Our evaluation results demonstrate the effectiveness of both the proposed attack and defense mechanisms in volumetric video streaming.

The method of experiencing recreational resources, whether it be through on-site participation in an activity or viewing an image of the resource, might directly affect an individual's preference for, and evaluation of, those resources. In this research note, we explore the effect of three-dimensional (3D) displays, which are now widely available to consumer markets, on an individual's perceptions of degraded trail conditions. We also explore the hypothesis that viewing an educational message about responsible hiking behavior influences perceptions of trail conditions. The effects of imagery type and the educational message were tested through experimentally varying types of trail impacts (muddiness and erosion) and impact severity (minimal and severe) across 20 images presented to individuals in a controlled laboratory setting on a 60-inch, 3D capable LCD monitor. Results indicate neither the use of 3D imagery nor the presence of an educational message had a significant main effect on perceptio...

This research examines public acceptability of regulations to reduce agricultural nutrient runoff and curb Harmful Algal Blooms (HABs). We tested the effects of two novel policy specific beliefs including support for farmers’ autonomy and support for external accountability. We also simultaneously tested the direct and indirect effects of political orientation and environmental worldview through a Direct Effect Model and a Mediation Model using structural equation modelling. Survey data were collected from 729 Ohio residents collected in November 2018. The specific regulatory policy measure we targeted is fines on excessive agricultural runoff. As hypothesized, autonomy beliefs negatively affect, and accountability positively affect support for fines. Both models revealed good fits. the direct effects of environmental worldviews political orientation were not supported. Instead, environmental worldviews indirectly increased support for fines through increased accountability beliefs and diminished autonomy beliefs. From the results, we suggest that when proposing suitable regulations for specific sites, policy makers and interest groups should be aware of differences in public support for farmer autonomy and external accountability, and that such differences are likely rooted in environmental worldviews. The study also suggests a need for coupled ecological and social studies that assess the likelihood of regional agricultural producers voluntarily adopting conservation practices and forecast the effectiveness of potential accountability measures.

With the utilization of electric vehicle on a large scale, its charging behavior is random in space and time, which has an impact on the quality of the power grid and its security and economic operation. In this paper, the factors related to driving rules of different types of vehicles are analyzed, the charging load model is generated correspondingly based on Monte-Carlo simulation. The results show the peak-valley ratio of the load under the different permeability of electric vehicles, and the proposed simulation can effectively simulate the charging load of the electric vehicles.

Reducing harmful algal blooms in Lake Erie, situated between the United States and Canada, requires implementing best management practices to decrease nutrient loading from upstream sources. Bi-national water quality targets have been set for total and dissolved phosphorus loads, with the ultimate goal of reaching these targets in 9-out-of-10 years. Row crop agriculture dominates the land use in the Western Lake Erie Basin thus requiring efforts to mitigate nutrient loads from agricultural systems. To determine the types and extent of agricultural management practices needed to reach the water quality goals, we used five independently developed Soil and Water Assessment Tool models to evaluate the effects of 18 management scenarios over a 10-year period on nutrient export. Guidance from a stakeholder group was provided throughout the project, and resulted in improved data, development of realistic scenarios, and expanded outreach. Subsurface placement of phosphorus fertilizers, cover crops, riparian buffers, and wetlands were among the most effective management options. But, only in one realistic scenario did a majority (3/5) of the models predict that the total phosphorus loading target would be met in 9-out-of-10 years. Further, the dissolved phosphorus loading target was predicted to meet the 9-out-of-10-year goal by only one model and only in three scenarios. In all scenarios evaluated, the 9-out-of-10-year goal was not met based on the average of model predictions. Ensemble modeling revealed general agreement about the effects of several practices although some scenarios resulted in a wide range of uncertainty. Overall, our results demonstrate that there are multiple pathways to approach the established water quality goals, but greater adoption rates of practices than those tested here will likely be needed to attain the management targets.

Hemorrhoids are one of the most common conditions that lead to surgery, and until now surgical hemorrhoidectomy has been the major effective treatment. Post-operative pain from hemorrhoidectomy has been experienced by thousands of patients and remains a major inconvenience of the operation.This study evaluates the clinical efficacy of the pestle needle therapy, an acupoint stimulation method, for relief of post-hemorrhoidectomy pain.This was a single-center, patient-assessor-blinded and randomized controlled trial with 154 patients receiving Milligan hemorrhoidectomy surgery. Eligible patients were randomly assigned to either a treatment group or a control group at a ratio of 1:1. The treatment group received the pestle needle therapy, with manual stimulation at Yaoshu (DU2), Mingmen (DU4), Changqiang (DU1), Chengshan (BL57), Erbai (EX-UE2) and the perianal points (1, 3, 5, 7, 9, and 11o'clock around the lesion); while the control group received a sham treatment with very light pressure. Three sessions of treatment were performed at 30 min, 4 h and 12 h after the surgery, and each lasted for 15 min.The primary outcome was post-operative pain measured with the visual analogue scale (VAS) at 12 h after surgery. The secondary outcomes included the VAS scores measured at 0.5, 2, 4, 6, 8, 24 and 48 h after surgery, the analgesic dose, the time and the VAS score of the patients' first defecation after surgery, as well as the Hamilton Rating Scale for Anxiety (HAMA) evaluated before discharge.The mean pain score of the treatment group was significantly lower than that of the control group (3.10 ± 1.27 vs 4.82 ± 1.29; P 0.001) at 12 h after surgery. Compared with the control group, patients in the treatment group needed a smaller dose of analgesic within the first 24 hours after surgery (P = 0.002); and their HAMA scores before discharge were lower (4.07 ± 2.40 vs 5.10 ± 2.45, P = 0.009). Compared to the treatment group, patients in the control group had a greater time to the first defecation after surgery ([52.34 ± 15.72] h vs [27.08 ± 13.68] h; P 0.001), but there was no difference in their VAS scores at the first defecation (P = 0.092).The pestle needle therapy was effective for relieving pain, reducing anxiety and improving bowel function after hemorrhoidectomy, and it is worthy of clinical application.

Deep Neural Networks are allowing mobile devices to incorporate a wide range of features into user applications. However, the computational complexity of these models makes it difficult to run them effectively on resource-constrained mobile devices. Prior work approached the problem of supporting deep learning in mobile applications by either decreasing model complexity or utilizing powerful cloud servers. These approaches each only focus on a single aspect of mobile inference and thus they often sacrifice overall performance. In this work we introduce a holistic approach to designing mobile deep inference frameworks. We first identify the key goals of accuracy and latency for mobile deep inference and the conditions that must be met to achieve them. We demonstrate our holistic approach through the design of a hypothetical framework called MDInference. This framework leverages two complementary techniques; a model selection algorithm that chooses from a set of cloud-based deep learning models to improve inference accuracy and an on-device request duplication mechanism to bound latency. Through empirically-driven simulations we show that MDInference improves aggregate accuracy over static approaches by over 40% without incurring SLA violations. Additionally, we show that with a target latency of 250ms, MDInference increased the aggregate accuracy in 99.74% cases on faster university networks and 96.84% cases on residential networks., To be published as an invited paper at IC2E'20. 10.5 pages (9.5 text + 1 bibliography) 8 figures, 4 tables

PURPOSE To compare the clinical efficacy of disposable multifunctional drainage tube (DMDT)-assisted irrigation with the traditional abscess incision rubber drainage technique in patients with severe multi-space infections in oral and maxillofacial head and neck regions. PATIENTS AND METHODS The data of 74 patients with severe multi-space infections in oral and maxillofacial head and neck regions, who were admitted to the Department of Oral and Maxillofacial Surgery, Central Hospital of Panzhihua City, Sichuan Province, China, between January 2015 and January 2019, were retrospectively studied. According to the treatment method, the patients were divided into 2 groups: the DMDT-assisted irrigation group and the abscess incision rubber drainage group. Cure rate, complications, length of hospitalization, days of antimicrobial use, cost of antimicrobial drugs, total hospitalization cost, number of dressing changes, and patient pain during dressing changes were compared between the 2 groups. RESULTS Of the 74 patients, 38 were treated with the DMDT-assisted irrigation, and 36 with the traditional abscess incision rubber drainage. Compared with the traditional treatment group, the total hospitalization cost of the DMDT-assisted irrigation group is not much different (P = 0.72), but the patients in the DMDT-assisted irrigation group have higher cure rate, fewer complications, less antibiotics cost, shorter hospitalization length and fewer dressing changes than the control group (P

Incorporating environmental, social, and governance (ESG) considerations into systematic investments has drawn numerous attention recently. In this paper, we focus on the ESG events in financial news flow and exploring the predictive power of ESG related financial news on stock volatility. In particular, we develop a pipeline of ESG news extraction, news representations, and Bayesian inference of deep learning models. Experimental evaluation on real data and different markets demonstrates the superior predicting performance as well as the relation of high volatility prediction to stocks with potential high risk and low return. It also shows the prospect of the proposed pipeline as a flexible predicting framework for various textual data and target variables.

Target To construct a method for generating object detection dataset based on the virtual environment. The generated dataset can be used for object detection tasks based on deep learning algorithms. Methods The procedural generation method was used to create the city's virtual environment, and also computer graphics were used for rendering and automatic labeling. Results We constructed a virtual reality environment and collected 1500 images through the virtual environment, including 1307 images containing valid vehicle and pedestrian information, and trained a deep learning model based on this dataset. Conclusions A virtual reality environment is successfully created, and the generated dataset can be used to train deep learning object detection algorithms, and the trained models can also effectively perform object detection in real world.

Objectives. Nonalcoholic fatty liver disease (NAFLD) and hyperlipidemia (HL) are common metabolic disorders due to overnutrition and obesity. NAFLD is often associated with hyperlipidemia. The aim of this study was to identify and compare the erythrocyte membrane lipids profile in NAFLD patients with or without HL.Methods. A total of 112 subjects (with similar age and body mass index) were divided into four groups: (1) normal controls, (2) NAFLD alone, (3) HL alone, and (4) NAFLD combined with HL (NAFLD + HL). Lipid was extracted from the erythrocyte membrane, and lipid profiles of subjects were analyzed by liquid chromatography mass spectrometry (LC-MS).Results. Data sets from 103 subjects were adopted for lipidomic analysis. Significant changes of lipid species were observed in patient groups, especially in the HL group and NAFLD + HL group. The HL group showed increased level of most lipid species, and decreased level of most lipid species was observed in the NAFLD + HL group. The weight percent of myristic acid, stearic acid, erucic acid, and docosahexaenoic acid also showed distinct variation between different groups.Conclusions. NAFLD, HL, and NAFLD + HL all had an impact on lipid profiling of the erythrocyte membrane. The influence of NAFLD alone is less important compared with HL. Some lipids should be highlighted because of their specific role in cell function and systemic metabolism.

We explore the opportunities and challenges of model execution caching, a nascent research area that promises to improve the performance of cloud-based deep inference serving. Broadly, model execution caching relies on servers that are geographically close to the end-device to service inference requests, resembling a traditional content delivery network (CDN). However, unlike a CDN, such schemes cache execution rather than static objects. We identify the key challenges inherent to this problem domain and describe the similarities and differences with existing caching techniques. We further introduce several emergent concepts unique to this domain, such as memory-adaptive models and multi-model hosting, which allow us to make dynamic adjustments to the memory requirements of model execution.

Infrastructure-as-a-Service (IaaS) cloud platforms rent computing resources with different cost and availability tradeoffs. For example, users may acquire virtual machines (VMs) in the spot market that are cheap, but can be unilaterally terminated by the cloud operator. Because of this revocation risk, spot servers have been conventionally used for delay and risk tolerant batch jobs. In this paper, we develop risk mitigation policies which allow even interactive applications to run on spot servers. Our System, SpotCheck is a derivative cloud platform, and provides the illusion of an IaaS platform that offers always-available VMs on demand for a cost near that of spot servers, and supports unmodified applications. SpotCheck’s design combines virtualization-based mechanisms for fault-tolerance, and bidding and server selection policies for managing the risk and cost. We implement SpotCheck on EC2 and show that it i) provides nested VMs with 99.9989 percent availability, ii) achieves upto 2-5 $\times$ cost savings compared to using on-demand VMs, and iii) eliminates any risk of losing VM state.

Geographically distributed cloud platforms are well suited for serving a geographically diverse user base. However, traditional cloud provisioning mechanisms that make local scaling decisions are not adequate for delivering the best possible performance for modern web applications that observe both temporal and spatial workload fluctuations. We propose GeoScale, a system that provides geo-elasticity by combining model-driven proactive and agile reactive provisioning approaches. GeoScale can dynamically provision server capacity at any location based on workload dynamics. We conduct a detailed evaluation of GeoScale on Amazon’s geo-distributed cloud and show up to 40% improvement in the 95th percentile response time when compared to traditional elasticity techniques.

Online applications that serve global workload have become a norm and those applications are experiencing not only temporal but also spatial workload variations. In addition, more applications are hosting their backend tiers separately for benefits such as ease of management. To provision for such applications, traditional elasticity approaches that only consider temporal workload dynamics and assume well-provisioned backends are insufficient. Instead, in this article, we propose a new type of provisioning mechanisms—geo-elasticity, by utilizing distributed clouds with different locations. Centered on this idea, we build a system called DBScale that tracks geographic variations in the workload to dynamically provision database replicas at different cloud locations across the globe. Our geo-elastic provisioning approach comprises a regression-based model that infers database query workload from spatially distributed front-end workload, a two-node open queueing network model that estimates the capacity of databases serving both CPU and I/O-intensive query workloads and greedy algorithms for selecting best cloud locations based on latency and cost. We implement a prototype of our DBScale system on Amazon EC2’s distributed cloud. Our experiments with our prototype show up to a 66% improvement in response time when compared to local elasticity approaches.

Summary This study proposes a 300-mA external capacitor-free low-dropout (LDO) regulator for system-on-chip and embedded applications. To achieve a full-load range from 0 to 300 mA, a two-scheme (a light-load case and a heavy-load case) operation LDO regulator with a novel control circuit is proposed. In the light-load case (0–0.5 mA), only one P-type metal–oxide–semiconductor input-pair amplifier with a 10-pF on-chip capacitor is used to obtain a load current as low as 0. In the heavy-load case (0.5 to 300 mA), both P-type metal–oxide–semiconductor and N-type metal–oxide–semiconductor differential input-pair amplifiers with an assistant push-pull stage are utilized to improve the stability of the LDO regulator and achieve a high slew rate and fast-transient response. Measurements show an output voltage of 3.3 V and a full output load range from 0 to 300 mA. A line regulation of 1.66 mV/V and a load regulation of 0.0334 mV/mA are achieved. The measured power-supply rejection ratio at 1 kHz is −65 dB, and the measured output noise is only 34 μV. The total active chip size is approximately 0.4 mm2 with a standard 0.5 μm complementary metal–oxide–semiconductor process. Copyright © 2017 John Wiley & Sons, Ltd.

Distributed clouds offer a choice of data center locations for providers to host their applications. In this paper, we consider distributed clouds that host virtual desktops which are then accessed by users through remote desktop protocols. Virtual desktops have different levels of latency-sensitivity, primarily determined by the actual applications running and affected by the end users’ locations. In the scenario of mobile users, even switching between 3G and WiFi networks affects the latency-sensitivity. We design VMShadow, a system to automatically optimize the location and performance of latency-sensitive VMs in the cloud. VMShadow performs black-box fingerprinting of a VM’s network traffic to infer the latency-sensitivity and employs both ILP and greedy heuristic based algorithms to move highly latency-sensitive VMs to cloud sites that are closer to their end users. VMShadow employs a WAN-based live migration and a new network connection migration protocol to ensure that the VM migration and subsequent changes to the VM’s network address are transparent to end-users. We implement a prototype of VMShadow in a nested hypervisor and demonstrate its effectiveness for optimizing the performance of VM-based desktops in the cloud. Our experiments on a private as well as the public EC2 cloud show that VMShadow is able to discriminate between latency-sensitive and insensitive desktop VMs and judiciously moves only those that will benefit the most from the migration. For desktop VMs with video activity, VMShadow improves VNC’s refresh rate by 90% by migrating virtual desktop to the closer location. Transcontinental remote desktop migrations only take about 4 min and our connection migration proxy imposes 13 μs overhead per packet.

In weather forecast products, stakeholder engagement in the research-to-operations (R2O) transition process has been increasingly valued yet it is far from being standardized. Engagement at multipl...

In this work, we look at how to effectively manage and utilize deep learning models at each edge location, to provide performance guarantees to inference requests. We identify challenges to use these deep learning models at resource-constrained edge locations, and propose to adapt existing cache algorithms to effectively manage these deep learning models.

This paper presents a compensation technique for a power supply rejection (PSR) improved external capacitor-less low drop-out (LDO) regulator. A replica circuit is used to cancel the power supply noise generated by the gate-source parasitic capacitor of the pass transistor. This design was fabricated in a 0.18 µm bipolar-CMOS-DMOS (BCD) technology with a power supply of 1.8 V. The active core chip area is 0.023 mm2, and the entire proposed LDO consumes 65 µA of quiescent current. It has a drop-out voltage of 200 mV, and the maximum load current is 60 mA. The measured PSR has a maximum -22 dB enhancement compared with a conventional uncompensated LDO when delivering a current of 60 mA.

Tourist congestion at hot spots has been a major management concern for UNESCO World Heritage Sites and other iconic protected areas. A growing number of heritage sites employ technologies, such as cameras and electronic ticket-checking systems, to monitor user levels, but data collected by these monitoring technologies are often under-utilized. In this study, we illustrated how to integrate data from hot spots by camera-captured monitoring and entrance counts to manage use levels at a World Heritage Site in Southeastern China. 6,930 photos of a congestion hotspot (scenic outlook on a trail) were collected within the park at a 10-minute interval over 105 days from January to November 2017. The entrance counts were used to predict daily average and maximum use level at the hotspots. Results showed that the average use level at the congestion hotspot did not exceed the use limit mandated by the park administration agency. However, from 9:20 am to 12:00 pm, the use level at hotspots exceeded visitor preferred use level. Visitor use level was significantly higher at the hotspot during a major Chinese "Golden Week". The daily entrance counts significantly predicted the average and maximum use level at the hotspot. Based on our findings, park managers can achieve the management goals by permitting the corresponding number of visitors passing the entrances. The gap manifested the complexities in visitor capacity management at high-use World Heritage Sites and other protected areas and calls for innovative monitoring and management strategies.

Distributed training frameworks, like TensorFlow, have been proposed as a means to reduce the training time of deep learning models by using a cluster of GPU servers. While such speedups are often desirable---e.g., for rapidly evaluating new model designs---they often come with significantly higher monetary costs due to sublinear scalability. In this paper, we investigate the feasibility of using training clusters composed of cheaper transient GPU servers to get the benefits of distributed training without the high costs. We conduct the first large-scale empirical analysis, launching more than a thousand GPU servers of various capacities, aimed at understanding the characteristics of transient GPU servers and their impact on distributed training performance. Our study demonstrates the potential of transient servers with a speedup of 7.7X with more than 62.9% monetary savings for some cluster configurations. We also identify a number of important challenges and opportunities for redesigning distributed training frameworks to be transient-aware. For example, the dynamic cost and availability characteristics of transient servers suggest the need for frameworks to dynamically change cluster configurations to best take advantage of current conditions., Comment: Accepted to ICAC'19. 11 pages, 8 figures, 5 tables

ObjectiveTo demonstrate different entities of appendicitis and causal association between microbiota and different types of appendicitis through studying cluster/outbreak, and providing guidance to find new cluster/outbreak of appendicitis and the epidemiological evidences of infectious etiology of appendicitis.Data SourcesPubMed, Embase, CNKI, WanFang, VIP, CBM from their establishment to Jan, 2019, and the references lists from retrieved reports.Study EligibilityReports on cluster/outbreak of appendicitis and reports of case series occurring in cluster/outbreak worldwide according to CDC’s definition of cluster/outbreak.Data Extraction and SynthesisTwo researchers independently assessed report quality and extracted data according to Moose. We used random effect model for meta-analysis by Meta-Analyst ß3.13 software. Study-level assessment was conducted according to investigation methods introduced by Reingold and outcome-level assessment by GRADE system. We selected outcome measures before data collection began.ResultsWe included 10 clusters/outbreaks of appendicitis from China and USA with total 626 patients. We demonstrated two entities, type 1 appendicitis (455 patients) and type 2 appendicitis (151patients). 20 patients left were unclassified type. For type 1 appendicitis, Natural history showed progression from a non-perforated appendicitis to perforated appendicitis as described traditionally. More than 88% of patients had elevated body temperature, WBC and neutrophil percentage. For type 2 appendicitis, natural history showed that only a few patients developed into phlegmonous appendicitis (6.9%,) or acute gangrenous appendicitis (1.4%) and no perforation or periappendicular abscess. More than 78% of patients had normal body temperature, WBC and NP. The patients’ time of type 1 appendicitis is shorter than that of type 2 appendicitis. Type 2 appendicitis had different histological features from type 1 appendicitis and was associated with fusobacteria. 9 of 10 cluster/outbreak occurred in group living unity such as school and camps, and many of them showed features of infectious diseases. The bodies of evidence were high quality in Meta analysis.ConclusionCluster/outbreak of appendicitis is more often than expected worldwide and occurred in group living unity. Sporadic perforated appendicitis and non-perforated appendicitis may be not two different entities, but different stages of a same entity, which is inconsistent with modern classification of appendicitis. Type 2 appendicitis is a new entities. Studying cluster/outbreak is a new method in finding of new entity and causal association between microbiota and different types of appendicitis. Epidemiological evidence supported infectious etiology of appendicitis.

Tourist congestion at hot spots has been a major management concern for UNESCO World Heritage Sites and other iconic protected areas. A growing number of heritage sites employ technologies, such as cameras and electronic ticket-checking systems, to monitor user levels, but data collected by these monitoring technologies are often under-utilize. In this study, we illustrated how to integrate data from hot spots by camera-captured monitoring and entrance counts to manage use levels at a World Heritage Site in southeastern China. 6,930 photos of a congestion hotspot (scenic outlook on a trail) were collected within the park at a 10-minute interval over 105 days from January to November 2017. The entrance counts were used to predict daily average and maximum use level at the hotspot Average use level at the congestion hotspot did not exceed the use limit mandated by the Chinese park administration agency. However, from 9:20 am to 12:00 pm, the use level at hotspots exceeded visitor preferred use level. Visitor use level was significantly higher at the hotspot during a major Chinese “golden week” holiday. The daily entrance counts significantly predicted the average and maximum use level at the hotspot. Based on our findings, we recommend that the number of visitors entering the gate on each day should be less than 28,764 for the hotspots to meet use level mandates, while less than 6,245 to meet visitor preference. The gap manifested the complexity in visitor capacity management at high-use World Heritage Sites and other protected areas and calls for innovative monitoring and management strategies.

Deep learning models are increasingly used for end-user applications, supporting both novel features such as facial recognition, and traditional features, e.g. web search. To accommodate high inference throughput, it is common to host a single pre-trained Convolutional Neural Network (CNN) in dedicated cloud-based servers with hardware accelerators such as Graphics Processing Units (GPUs). However, GPUs can be orders of magnitude more expensive than traditional Central Processing Unit (CPU) servers. These resources could also be under-utilized facing dynamic workloads, which may result in inflated serving costs. One potential way to alleviate this problem is by allowing hosted models to share the underlying resources, which we refer to as multi-tenant inference serving. One of the key challenges is maximizing the resource efficiency for multi-tenant serving given hardware with diverse characteristics, models with unique response time Service Level Agreement (SLA), and dynamic inference workloads. In this paper, we present Perseus, a measurement framework that provides the basis for understanding the performance and cost trade-offs of multi-tenant model serving. We implemented Perseus in Python atop a popular cloud inference server called Nvidia TensorRT Inference Server. Leveraging Perseus, we evaluated the inference throughput and cost for serving various models and demonstrated that multi-tenant model serving led to up to 12% cost reduction., Comment: 8 pages, 5 figures, and 6 tables. In proceedings of International Conference on Cloud Engineering (IC2E) 2020

Topological magnetic states such as skyrmions have widely been investigated as a suitable candidate for memory bits in next generation data storage device applications including random access memory (RAM) or race-track memory as well as spin-torque nano-oscillators etc. Energy efficient multistate switching of these states based on electric control of magnetization involves the electric field tuning of perpendicular magnetic anisotropy (PMA). By using micromagnetic simulation study, we demonstrate the strain-mediated electric manipulation of various topological magnetic states in nanodisc with [Pt/Co/Ta]n trilayer stacks residing on a piezoelectric substrate. A stability phase diagram is derived for the representation of regions of the lowest energy states against the applied strain as a function of disc dimensions. Utilizing this derived phase diagram, various electric field-induced strain-mediated switching scenarios including non-volatile reversible switching of skyrmion into skyrmionium, 3π state and helical stripe domain state, the creation of skyrmionium from uniform single-domain magnetic state and its annihilation and volatile reversible switching between skyrmionium and vortex are achieved. Moreover, an extended phase diagram based on the estimated critical values of the applied strain for different domain structures within the nanodisc is also built and further implemented in designing a multistate sequential switching between various magnetic states triggered by a continuously varying strain pulse. These findings hold a potential for contributing in the development of low-power high-density multistate magneto-electric or magneto-elastic memory and logic device applications involving simple architectural concept applicable to future microelectronic integrated device strategies.

Within a shifting climate of renewable energy options, technology innovations in the energy sector are vital in combating fossil-fuel-driven climate change and economic growth. To enter this market dominated by fossil fuels, renewable energy innovations need to overcome significant barriers related to cost, relative advantages compared to fossil fuels, and policy incentive programs. A better understanding of the innovation diffusion of new technologies in establishing the renewable energy industry can aid policy makers in designing and implementing other renewable energy support programs and improving adoption rates within existing programs. This study assessed industry leaders’ perceptions through semi-structured interviews. We explored the innovation diffusion process of wood pellet residential heating technology, as well as policy needs and barriers within this industry that are hindering successful long-term diffusion and sustainability. We show that while there is high potential to the wood pellet industry in terms of local resources and overall advantages to fossil fuels, it can be difficult to achieve sustainable economic growth with current cost barriers and further policy programs and incentives are needed in addition to improved communication to reduce adoption barriers for wood pellet technology.

Raw material supply and trade changes are the main factors influencing the development of furniture manufacturing in Heilongjiang Province, China. In view of the dynamic association between indicators, wood processing is considered an essential raw material for furniture production, while import-export volumes indicate the changes in demand for furniture. When the Granger causality tests between indicators are performed, they suggest that wood processing has an obvious influence on furniture manufacturing. Through the Vector Auto-Regression model and the impulse response function, it is found that the interaction between wood processing and furniture making is strong. The relationship between furniture making and furniture import and export produces an expanding driving role or inhibiting effect, due to the uncertainty of import-export supply and demand. Finally, suggestions on promoting stable growth of Heilongjiang furniture making are proposed according to the analysis of the results.

Damage can be identified using generalized flexibility matrix based methods, by using the first natural frequency and the corresponding mode shape. However, the first mode is not always appropriate to be used in damage detection. The contact interface of rod-fastened-rotor may be partially separated under bending moment which decreases the flexural stiffness of the rotor. The bending moment on the interface varies as rotating speed changes, so that the first- and second-modal parameters obtained are corresponding to different damage scenarios. In this paper, a structural damage detection method requiring single nonfirst mode is proposed. Firstly, the system is updated via restricting the first few mode shapes. The mass matrix, stiffness matrix, and modal parameters of the updated system are derived. Then, the generalized flexibility matrix of the updated system is obtained, and its changes and sensitivity to damage are derived. The changes and sensitivity are used to calculate the location and severity of damage. Finally, this method is tested through numerical means on a cantilever beam and a rod-fastened-rotor with different damage scenarios when only the second mode is available. The results indicate that the proposed method can effectively identify single, double, and multiple damage using single nonfirst mode.

Despite its clinical promise, photodynamic therapy (PDT) suffers from a key drawback associated with its oxygen-dependent nature, which limits its effective use against hypoxic tumors. Moreover, both PDT-mediated oxygen consumption and microvascular damage further increase tumor hypoxia and, thus, impede therapeutic outcomes. In recent years, numerous investigations have focused on strategies for overcoming this drawback of PDT. These efforts, which are summarized in this review, have produced many innovative methods to avoid the limits of PDT associated with hypoxia.

Our study examined relationships among pellet mills, bulk delivery companies, and high-efficiency pellet boiler equipment firms in northern New England as they relate to homeowner satisfaction, using social network analysis and the concept of supply chain management. The continual growth of supply and demand for automated pellet heating requires a careful match between innovative technologies and homeowner needs; these involve multiple factors and require collaboration among firms. Using interview data with managers from pellet mills, bulk delivery companies, and equipment firms in Maine, New Hampshire and Vermont, we found fifteen firms that are connected through both a transaction network and an informal business interaction network. The networks were characterized by short paths and no obvious sign of centralization. Network statistics reported for each network included density, clustering coefficient, and degree-, closeness- and betweenness- centrality. Most firms in the supply networks shared custome...