Search

The sulfur dioxide blower is a centrifugal blower that transports various gases in the process of acid production with flue gas. Accurate prediction of the outlet pressure of the sulfur dioxide blower is quite significant for the process of acid production with flue gas. Due to the internal structure of the sulfur dioxide blower being complex, its mechanism model is difficult to establish. A novel method combining one-dimensional convolution (Conv1D) and bidirectional gated recurrent unit (BiGRU) is proposed for short-term prediction of the outlet pressure of sulfur dioxide blower. Considering the external factors such as inlet pressure and inlet flow rate of the blower, the proposed method first uses Conv1D to extract periodic and local correlation features of these external factors and the blower's outlet pressure data. Then, BiGRU is used to overcome the complexity and nonlinearity in prediction. More importantly, genetic algorithm (GA) is used to optimize the important hyperparameters of the model. Experimental results show that the combined model of Conv1D and BiGRU optimized by GA can predict the outlet pressure of sulfur dioxide blower accurately in the short term, in which the root mean square error (RMSE) is 0.504, the mean absolute error (MAE) is 0.406, and R-square (R2) is 0.993. Also, the proposed method is superior to LSTM, GRU, BiLSTM, BiGRU, and Conv1D-BiLSTM. [ABSTRACT FROM AUTHOR]

Due to the particularity of the tendon-to-bone interface, the integration of tendon graft and bone after anterior cruciate ligament (ACL) reconstruction is the key to the success of the surgery. This reconstruction requires early on-demand angiogenesis and osteogenesis to promote early and rapid healing of the tendon-to-bone interface. Here, a hypoxic-mimetic 3D porous hydrogel was prepared with a multi-responsive hybrid interpenetrating network to the tendon-to-bone interfacial microenvironment under a highly hypoxic, acidic, reactive oxygen species (ROS), and related enzyme-enriched state. The hypoxia agent DFO can be effectively control-released adaptive to the stimulation of tendon-to-bone interfacial microenvironment, to upregulate HIF-1α, promote the expression of vascular endothelial growth factor, bone morphogenetic protein 2 and runt-related transcription factor 2, thereby promoting rapid early angiogenesis and bone formation on demand. At the same time, it can effectively resist oxidative stress, reduce ROS and the inflammatory response, promote better integration of the tendon-to-bone interface. Furthermore, in vivo experiments combining micro-CT, histological and biomechanical analysis also showed effective promotion of early and rapid tendon-to-bone healing. In conclusion, our study may provide a new therapeutic strategy for the integration of the tendon-to-bone interface. To promote early and rapid healing of the tendon-bone interface by on-demand angiogenesis and osteogenesis, a novel hypoxic-mimetic 3D porous interpenetrating network hydrogel was prepared with controllable release of hypoxia agent DFO multi-responsive to the stimulation of tendon-bone interfacial microenvironment. [Display omitted] • A novel stimuli-responsive hypoxia-mimetic hydrogel was developed for tendon-to-bone healing. • The fabricated hybrid interpenetrating network is multi-responsive to the highly hypoxic, acidic, ROS, and related enzyme-enriched microenvironment. • The DFO can be controllable released adaptive to the tendon-bone interfacial microenvironment, thereby promoting on demandangiogenesis and bone formation. • In vivo results from micro-CT, histological and biomechanical analysis showed effective promotion of early and rapid healing of ACL reconstruction. [ABSTRACT FROM AUTHOR]

Residential displacement by urban regeneration in western economies and passive relocation in eastern countries have attracted the attention of researchers. Over the past decades, Chinese megacities have undergone massive passive relocation. They are reforming their old downtown areas and demolishing substandard housing. The government relocates residents to affordable city-peripheral large-scale residential areas. These residents are called “relocatees”. So far, few studies have explored relocatee and non-relocatee on travel-mode-choice preferences and travel equity in these types of areas with adequate resident samples. To fill this gap, this study conducts a survey in five peripheral large-scale residential areas in Shanghai, uses statistical analysis of individual demographic characteristics and transportation-related decisions of relocatees and non-relocatees, and estimates travel mode choice models for three different groups (relocatee, non-relocatee, and overall samples). Consumer surplus difference is calculated as a measurement of travel equity. Results show that compared to non-relocatees, relocatees are older, poorer, and have a higher mode share of bus and lower mode share of car. Non-relocatees' value of time (¥25.23 per hour) is greater than relocatees' (¥22.62 per hour). As for travel mode preferences, for relocatees, the males tend to choose motorcycle, moped or e-bike, then bicycle. For non-relocatees, the males tend to choose motorcycle and then car. Relocatee has a ¥6.88 per person lower travel quality than non-relocatee. Megacity-periphery development and relocation process have a more negative effect on relocatee’s travel than on other population, from a travel equity perspective. This study contributes to the literatures on travel behavior and equity in megacity peripheral areas in developing countries. The findings point to important implications for the peripheral area’s policy. [ABSTRACT FROM AUTHOR]

A generic evacuation model is developed based on the behavioral characteristics of passengers during bus fires. Inspired by hydrodynamic mechanisms, the proposed particle-based force-driven model is built using force analysis of evacuees and analogs the movement of people as fluid motion. Several modules are developed to characterize evacuees' preferences for varying exits, aggregated behavior of groups, and effects of driver. The Smoothed Particle Hydrodynamics (SPH) methods are applied to solve the proposed model using smoothed kernel functions. Experiments are conducted to calibrate and validate the proposed model based on observed evacuees' behavioral characteristics, such as evacuation time, evacuation speed under different scenarios in terms of bus capacity, and the interaction between evacuees. The results of the study will contribute to a better understanding of bus evacuation options, improved design of bus vehicles to cope with internal fires, and the development of appropriate policies and regulations for bus passenger evacuation in response to fires. The proposed method can also be readily applied to other related fields with minor modifications. • A force-driven human motion model is proposed. • Heterogeneous effects of different evacuees are considered. • The model is calibrated with real-world experiments. [ABSTRACT FROM AUTHOR]

In this paper, a heterogeneous spatial-temporal similarity search framework is proposed, in which the datasets come from multiple different asynchronous data sources. Due to measuring error, data loss, and other factors, the similarity search based on single points along a trajectory usually cannot fulfill the accuracy requirements in our heterogeneous case. To address this issue, we introduce a concept of the spatial-temporal cluster of points, instead of single points, which can be identified for each target query. By following this concept, we further design a spectral clustering algorithm to construct the clusters in the pre-processing phase effectively. And the query processing is improved for the accuracy of the search by unifying multiple search metrics. To validate our idea, we also prototype a clustered online spatial-temporal similarity search system, "Osprey", to calculate in parallel the similarity of spatial-temporal sequences in the heterogeneous search on a distributed database. Our empirical study is conducted based on an open dataset, called "T-Drive", and a billion-scale dataset consisting of WiFi positioning records gathered from the urban metro system in Shenzhen, China. The experimental results show that the latency of our proposed system is less than 4s in most cases, and the accuracy is more than 70% when the similarity exceeds 0.5. [ABSTRACT FROM AUTHOR]

A common approach to improving resource utilization in data centers is to adaptively provision resources based on the actual workload. One fundamental challenge of doing this in microservice management frameworks, however, is that different components of a service can exhibit significant differences in their impact on end-to-end performance. To make resource management more challenging, a single microservice can be shared by multiple online services that have diverse workload patterns and SLA requirements. We present an efficient resource management system, namely Erms, for guaranteeing SLAs with high probability in shared microservice environments. Erms profiles microservice latency as a piece-wise linear function of the workload, resource usage, and interference. Based on this profiling, Erms builds resource scaling models to optimally determine latency targets for microservices with complex dependencies. Erms also designs new scheduling policies at shared microservices to further enhance resource efficiency. Experiments across microservice benchmarks as well as trace-driven simulations demonstrate that Erms can reduce SLA violation probability by 5× and more importantly, lead to a reduction in resource usage by 1.6×, compared to state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

As autonomous vehicles (AVs) and human-driven vehicles (HVs) are expected to share the road for the foreseeable future, understanding how to improve the stability of mixed-autonomy platoons is crucial. This paper introduces a novel adaptive control strategy tailored for a specific platoon configuration termed as “1+n+1\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$1+n+1$$\end{document}”, consisting of a leading AV, n intervening vehicles, and a trailing AV. Utilizing vehicle-to-vehicle communication, the trailing AV adapts to real-time traffic states, thereby promoting overall platoon stability. Our analysis, grounded in Lyapunov theory, demonstrates that stabilizing only the trailing vehicle is sufficient to ensure the entire system reaches a stable state. To mitigate the negative effects of sensor and actuator uncertainties, we also introduce a corrective signal framework capable of nullifying adverse inputs. Numerical experiments validate the effectiveness of the proposed strategy in platoon control, which can also be adapted to other platoon configurations. Additionally, the performance of this strategy on macroscopic traffic flow is explored, suggesting substantial potential throughput increases compared with the benchmark strategy of Cooperative Adaptive Cruise Control (CACC). [ABSTRACT FROM AUTHOR]

The prediction of surrounding agent trajectories in heterogeneous traffic environments remains a challenging task for autonomous driving due to several critical issues, such as understanding social interactions among agents and the environment, handling multiclass traffic movements, and generating feasible trajectories in accordance with real‐world rules, all of which hinder prediction accuracy. To address these issues, a new multimodal trajectory prediction framework based on the transformer network is presented in this study. A hierarchical‐structured context‐aware module, inspired by human perceptual logic, is proposed to capture contextual information within the scene. An efficient linear global attention mechanism is also proposed to reduce the computation and memory load of the transformer framework. Additionally, this study introduces a novel auxiliary loss to penalize infeasible off‐road predictions. Empirical results on the Lyft l5kit data set demonstrate the state‐of‐the‐art performance of the proposed model, which substantially enhances the accuracy and feasibility of prediction outcomes. The proposed model also possesses a unique feature, effectively dealing with missing input observations. This study underscores the importance of comprehending social interactions among agents and the environment, handling multiclass traffic movements, and generating feasible trajectories adhering to real‐world rules in autonomous driving. [ABSTRACT FROM AUTHOR]

Objective: To determine factors influence place of death (POD) for end-stage cancer patients and investigate how the healthcare utilization mediates on the effect of socioeconomic status (SES) on POD.Design: A population-based, retrospective study from July 2015 to June 2017.Setting: Yichang, China.Participants: 894 end-stage cancer patients.Main Outcome Measure: POD.Results: Patients of hospital death experience more inpatient hospitalization services (IHS) and emergency department visits. Patients enrolled in the New Rural Cooperative Medical Scheme (OR = 7.60, P < 0.001) and Urban Employee Basic Medical Insurance (OR = 28.0, P < 0.001) have higher rates of hospital death than those in the Urban Resident-based Basic Medical Insurance. Living with spouse (OR = 1.72, P = 0.019) and receiving higher education (OR = 1.92, P = 0.004), increase the likelihood of hospital death by 72% and 92%, respectively. The probability of hospital death will increase by 14% and decrease by 4% per IHS and outpatient services occur, respectively. Outpatient services (Z = -2.28, P < 0.001), and IHS (Z = 2.17, P < 0.001) mediate 1.81% and 1.89%, respectively, of the effect of health insurance on POD. The overall effect of the mediators is non-statistically significant (Z = 0.09, P = 0.825).Conclusion: POD is mainly driven by SES. The relationship between health insurance and POD is partly mediated by outpatient services and IHS, respectively. The results corroborated that hospital and home services should be coherently bridged. Furthermore, benefit packages for end-stage cancer patients could be redesigned. [ABSTRACT FROM AUTHOR]

Federated Learning coordinates many mobile devices to train an artificial intelligence model while preserving data privacy collaboratively. Mobile devices are usually equipped with totally different hardware configurations, leading to various training capabilities. At the same time, the distribution of the local training data is highly heterogeneous across different clients. Randomly selecting the clients to participate in the training process results in poor model performance and low system efficiency. In this paper, we propose GraphCS, a graph-based client selection framework for heterogeneity in Federated Learning. GraphCS first measures the distribution coupling across the clients via the model gradients. After that, it divides the clients into different groups according to the diversity of the local datasets. At the same time, it well estimates the runtime training capability of each client by jointly considering the hardware configuration and resource contention caused by the concurrently running apps. With the distribution coupling information and runtime training capability, GraphCS selects the best clients in order to well balance the model accuracy and overall training progress. We evaluate the performance of GraphCS with mobile devices with different hardware configurations on various datasets. The experiment results show that our approach improves model accuracy up to 45.69%. Meanwhile, it reduces communication and computation overhead 87.35% and 89.48% at best, respectively. Furthermore, GraphCS accelerates the overall training process up to 35×. • We observe that the distribution of the training data can be reflected by the gradient of a machine learning model. • We observe that the training capability is jointly affected by the hardware configuration and resource contention. • we design a framework selects the best clients with diverse local datasets and highest runtime training capability. • We prototype GraphCS on a Federated system consisting of mobile devices with heterogeneous hardware configurations. [ABSTRACT FROM AUTHOR]

Federated Learning is designed for multiple mobile devices to collaboratively train an artificial intelligence model while preserving data privacy. Instead of collecting the raw training data from mobile devices to the cloud, Federated Learning coordinates a group of devices to train a shared model in a distributed manner with the training data located on the devices. However, unbalanced data distribution and heterogeneous hardware configurations across different devices badly hurts the performance of collaborative model and severely impacts the overall training progress. Thus, a framework that can well balance the model accuracy and the training progress is urgently required. In this paper, we propose PAGroup, a privacy-aware grouping framework for high-performance Federated Learning. PAGroup intelligently divides the participating clients into different groups through carefully analyzing the privacy requirement of the training data and the solid social relationship of the participating clients. After that, PAGroup conducts data shaping and capability-ware training in order to improve the model performance while accelerating the overall training process. We evaluate the performance of PAGroup with both simulation and hardware testbed. The evaluation results show that PAGroup improves model accuracy up to 21%. Meanwhile, it decreases 81% communication overhead, 29% computation cost and 84% wall-clock time at best comparing with the baselines. • Specific private data that required by a learned model can be exchanged daily in certain social community. • PAGroup can share private date within task-specific groups organized according to strongly connected social community. • Training devices in a group can be viewed as a "super" client to compatibility to existing optimization algorithms. • The impact of local data scale, diversity, and label distribution to the model performance has been theoretically proved. [ABSTRACT FROM AUTHOR]

Federated learning is designed to collaboratively train a shared model based on a large number of mobile devices while preserving data privacy, which has been widely adopted to support different geo-spatial systems. However, two critical issues prevent federated learning to be effectively deployed on resource-constrained devices in large scale. First, federated learning causes high energy consumption which can badly hurt the battery lifetime of mobile devices. Second, leakage of sensitive personal information still occurs during the training process. Thus, a system that can effectively protect the sensitive information while improving the energy efficiency is urgently required for a mobile-based federated learning system. This paper proposes SmartDL, an energy-aware decremental learning framework that well balances the energy efficiency and data privacy in an efficient manner. SmartDL improves the energy efficiency from two levels: (1) global layer, which adopts an optimization approach to select a subset of participating devices with sufficient capacity and maximum reward. (2) local layer, which adopts a novel decremental learning algorithm to actively provides the decremental and incremental updates, and can adaptively tune the local DVFS at the same time. We prototyped SmartDL on physical testbed and evaluated its performance using several learning benchmarks with real-world traces. The evaluation results show that compared with the original federated learning, SmartDL can reduce energy consumption by 75.6–82.4% in different datasets. Moreover, SmartDL achieves a speedup of 2–4 orders of magnitude in model convergence while ensuring the accuracy of the model. [ABSTRACT FROM AUTHOR]

Accurate prediction of short-term origin-destination (OD) matrix is crucial for operations in metro systems. Recently, some deep learning-based models have been proposed for OD matrix forecasting in ride-hailing or high-way scenarios. However, the metro OD matrix forecasting receives less attention, and it has different prior knowledge and complex spatiotemporal contextual setting; for example, the sparse destination distribution and the incomplete OD matrices collection in recent time intervals due to unfinished trips before the predicted time interval. This paper designs a deep learning approach for metro OD matrix prediction by addressing the recent destination distribution availability, augmenting the flow presentation for each station, and digging out the global spatial dependency and multiple temporal scale correlations in the mobility patterns of metro passengers. Specifically, it first proposes to complete the recent OD matrices by combining some empirical knowledge including the historical mobility pattern and arrival time distribution. Then, it learns the complementary spatiotemporal contextual features by embedding methods to enrich the station representation. Finally, it captures global mobility trend of metro passengers at each origin station through aggregating the trend of all other origin stations by self-attention mechanism since the mobility synchronizes among stations from spatial perspective. Three temporal convolutional networks are leveraged to extract three temporal trends in passenger mobility data, i.e., recent trend, daily trend, and weekly trend. Smart card data from Shenzhen and Hangzhou metro systems are utilized to demonstrate the superiority of our model over other competitors. [ABSTRACT FROM AUTHOR]

Currently, most edge-based Deep Reinforcement Learning (Deep-RL) applications have been deployed in the edge network, however, their mainstream studies are still short of adequate considerations on its limited compute and bandwidth resources. In this paper, we investigate the near on-policy of actions taking in distributed Deep-RL architecture, and propose a "hybrid near on-policy" Deep-RL framework, called Coknight , by leveraging a game-theory based DNN partition approach. We first formulate the partition problem into a variant of knapsack problem in device-edge setting, and then transform it into a potential game with a formal proof. Finally, we show the problem is NP-complete whereby an efficient distributed algorithm based on the potential game theory is developed from device perspective to achieve fast and dynamic partitioning. Coknight not only significantly improves the resource efficiency of the Deep-RL but also allows the inference to enforce the scalability of the actor policy. We prototype the framework with extensive experiments to validate our findings. The experimental results show that with the premise of a rapid convergence guarantee, Coknight , compared with Seed-RL , can reduce GPU utilization by 30% while providing large-scale scalability. • Leverage the DNN partition technology in the device-edge settings to accelerate DRL. • Formulate partition problem as a knapsack problem and prove its NP-Completeness. • Transform the problem into a multi-player game and prove it is a potential game. • Design a dynamic algorithm with polynomial time complexity for DNN partition. • Implement the framework to conduct physical tests and large-scale simulation studies. [ABSTRACT FROM AUTHOR]

We present an extension of the banker's algorithm to resolve deadlock for programs whose resource-request graph can be modeled as a recursion tree for parallel execution. Our algorithm implements the banker's logic, with the key difference being that some properties of the tree are fully exploited to improve the resource utilization and safety check in deadlock avoidance. For an $n$ n -node tree modeled program making requests to $m$ m types of resources, our recursion-tree based algorithm can obtain a time complexity of $O(mn\log \log n)$ O (m n log log n) on average in safety check while reducing the conservativeness in resource utilization. We reap these benefits by proposing a concept of the resource critical tree and leverage it to localize the maximum claim associated with each node in the tree. To tackle the case when the tree model is not statically known, we relax the definition of a local maximum claim by sacrificing some resource utilization. With this trade-off, the algorithm can resolve the deadlock and achieve more efficient safety checks within time of $O(m\log \log n)$ O (m log log n) . Our empirical studies on a two-dimensional integration problem on sparse grids show that the proposed algorithms can reduce resource utilization conservativeness and improve avoidance performance by minimizing the number of safety checks. [ABSTRACT FROM AUTHOR]

In this paper, we focus on strategy configurations of multiple users to make cloud service reservation. We consider the problem from a game theoretic perspective and formulate it into a non-cooperative game among the multiple cloud users, in which each user is informed with incomplete information of other users. For each user, we design a utility function which combines the net profit with time efficiency and try to maximize its value. We solve the problem by employing variational inequality (VI) theory and prove that there exists a Nash equilibrium solution set for the formulated game. Then, we propose an iterative proximal algorithm (IPA), which is designed to compute a Nash equilibrium solution. The convergence of the IPA algorithm is also analyzed and we find that it converges to a Nash equilibrium if several conditions are satisfied. Finally, we conduct some numerical calculations to verify our theoretical analysis. The experimental results show that our proposed IPA algorithm converges to a stable state very quickly and improves the utilities of all users to certain extent by configuring a proper request strategy. [ABSTRACT FROM AUTHOR]

Federated Learning is a framework that coordinates a large amount of workers to train a shared model in a distributed manner, in which the training data are located on the workers' sides in order to preserve data privacy. There are two challenges in the crowdsourcing of FL, the workers who participant in training need to consume computing and communication resources, so that they are reluctant to participate in the training process if they can not get reasonable rewards. Moreover, there may be attackers who send arbitrary updates to get undeserving compensation or even destroy the model, thus, effective prevention of malicious workers is also critical. An incentive mechanism is urgently required in order to encourage high-quality workers to participate in FL and to punish the attackers. In this paper, we propose FGFL, a blockchain-based incentive governor for Federated Learning. In FGFL, we assess the participants with reputation and contribution indicators. Then the task publisher rewards workers fairly to attract efficient ones while the malicious ones are punished and eliminated. In addition, we propose a blockchain-based incentive management system to manage the incentive mechanism. We evaluate the effectiveness and fairness of FGFL through theoretical analysis and comprehensive experiments. The evaluation results show that FGFL fairly rewards workers according to their corresponding behavior and quality. FGFL increases the system revenue by 0.2% to 3.4% in reliable federations compared with baselines. And in the unreliable scenario where contains attackers, the system revenue of FGFL outperforms the baselines by more than 46.7%. • A fair incentive mechanism for federated learning is proposed. • The mechanism relies on accurately measuring workers' contributions and reputations. • Attacks and cheats should be detected and punished. • A blockchain-based management system is proposed to achieve trusted incentives and training in federated learning. [ABSTRACT FROM AUTHOR]

This study delves into the factors that contribute to the severity of single-vehicle crashes, focusing on enhancing both computational speed and model robustness. Utilizing a mixed logit model with heterogeneity in means and variances, we offer a comprehensive understanding of the complexities surrounding crash severity. The analysis is grounded in a dataset of 39,788 crash records from the UK's STATS19 database, which includes variables such as road type, speed limits, and lighting conditions. A comparative evaluation of estimation methods, including pseudo-random, Halton, and scrambled and randomized Halton sequences, demonstrates the superior performance of the latter. Specifically, our estimation approach excels in goodness-of-fit, as measured by ρ 2 , and in minimizing the Akaike Information Criterion (AIC), all while optimizing computational resources like run time and memory usage. This strategic efficiency enables more thorough and credible analyses, rendering our model a robust tool for understanding crash severity. Policymakers and researchers will find this study valuable for crafting data-driven interventions aimed at reducing road crash severity. • Mixed logit model with heterogeneity in means and variance for crash analysis. • Scrambled and random Halton sequences optimize estimation accuracy. • Study achieves robustness and efficiency in crash severity modeling. [ABSTRACT FROM AUTHOR]

Travel activities influence individual travel location, time, and frequencies. Understanding individual travel regularity under different activities is crucial for individual mobility prediction and urban facilities planning. Existing studies need further improvement as they have not adequately taken into account the impact of activity factors on individual travel patterns. In this paper, we propose an innovative framework for exploring travel regularity and the correlation with urban region attributes at the individual level. The framework's novelty and uniqueness lie in its three layers: (i) A trip pre-processing layer for extracting every complete public transit trip and trip purposes (e.g. commuting activity and non-commuting activity) by combining individual's travel characteristics and public transport network. (ii) An entropy-based travel regularity measurement and K-means based multiple-view clustering layer to assess the extent of individual travel repetition over time for various travel activity and investigate the similarities and differences among users. (iii) A region dependence correlation analysis layer for exploring the correlation between individual travel regularity and regions attributes of two key locations: home and workplace More importantly, by employing the framework, we gained empirical insights based on a large-scale dataset (covering 0.64 million users) collected from public traffic smart cards. For instance, commuters can be categorized into four groups based on the regularity of their commuting activities: Regular Workers (24%), Flex-time Workers (21%), Overtime Workers (35%), and Other Workers (20%). The distribution of these four groups is associated with workplace's administrative characteristics, with a Pearson correlation of 0.503. In addition, individuals can also be classified into two groups based on their travel regularity of non-commuting activities: Limited Visitors (37%) and Active Explorers (63%). The distribution of these individuals is correlated with the coverage ratio of public transit facility in their home locations, showing a Pearson correlation of 0.604. • A framework exploring individual travel regularity under different purpose. • Entropy-Space K-means for travel activity-based individual classification. • Inter-personal and intra-personal variability of individual travel patterns. • Correlation between individual travel regularity and urban region attributes. [ABSTRACT FROM AUTHOR]

Both semi-supervised learning and transfer learning aim at lowering the annotation burden for training models. However, such two tasks are usually studied separately, i.e. most semi-supervised learning algorithms train models from scratch while transfer learning assumes pre-trained models as the initialization. In this work, we focus on a previously-less-concerned setting that further reduces the annotation efforts through incorporating both semi-supervised and transfer learning, where specifically a pre-trained source model is used as the initialization of semi-supervised learning. As those powerful pre-trained models are ubiquitously available nowadays and can considerably benefit various down-streaming tasks, such a setting is relevant to real-world applications yet challenging to design effective algorithms. Aiming at enabling transfer learning under semi-supervised settings, we propose a hierarchical self-regularization mechanism to exploit unlabeled samples more effectively, where a novel self-regularizer has been introduced to incorporate both individual-level and population-level regularization terms. The former term employs self-distillation to regularize learned deep features for each individual sample, and the latter one enforces self-consistency on feature distributions between labeled and unlabeled samples. Samples involved in both regularizers are weighted by an adaptive strategy, where self-regularization effects of both terms are adaptively controlled by the confidence of every sample. To validate our algorithm, exhaustive experiments have been conducted on diverse datasets such as CIFAR-10 for general object recognition, CUB-200-2011/MIT-indoor-67 for fine-grained classification and MURA for medical image classification. Compared with state-of-the-art semi-supervised learning methods including Pseudo Label , Mean Teacher , MixMatch and FixMatch , our algorithm demonstrates two advantages: first of all, the proposed approach adopts a new point of view to tackle problems caused by inadequate supervision and achieves very competitive results; then, it is complementary to these state-of-the-art methods and thus can be combined with them to get additional improvements. Furthermore, our method can also be applied to fully supervised transfer learning and self-supervised learning. We have published our code at https://github.com/SHI-Labs/Semi-Supervised-Transfer-Learning. [Display omitted] We study a very practical and rarely explored problem: fine-tuning with additional unlabeled data. Traditional semi-supervised learning (SSL) cares about the default setting of training from scratch. Labeled data are generally limited in SSL applications, and we propose a novel framework of hierarchical self-regularization, in which unlabeled data are utilized to achieve individual-level knowledge consistency and population-level distribution consistency. Adaptive strategies are imposed to enhance the quality of selected unlabeled data. The proposed framework is performant and orthogonal to existing SSL methods. [ABSTRACT FROM AUTHOR]

Service migration in mobile edge computing is a promising approach to improving the quality of service (QoS) for mobile users and reducing the network operational cost for service providers as well. However, these benefits are not free, coming at costs of bulk-data transfer, and likely service disruption, which could consequently increase the overall service costs. To gain the benefits of service migration while minimizing its cost across the edge nodes, in this paper, we leverage reinforcement learning (RL) method to design a cost-effective framework, called Mig-RL , for the service migration with a reduction of total service costs as a goal in a mobile edge environment. The Mig-RL leverages the infrastructure of edge network and deploys a migration agent through Q-learning to learn the optimal policy with respect to the service migration status. We distinguish the Mig-RL from other existing works in several major aspects. First, we fully exploit the nature of this problem in a modest migration space, which allows us to constrain the number of service replicas whereby a defined state–action space could be effectively handled, as opposed to those methods that need to always approximate a huge state–action space for policy optimality. Second, we advocate a migration policy-base as a cache to save the learning process by retrieving the most effective policy whenever a similar migration pattern is encountered as time goes on. Finally, by exploiting the idea of software defined network, we also investigate the efficient implementation of Mig-RL in mobile edge network. Experimental results based on some real and synthesized access sequences show that Mig-RL , compared with the selected existing algorithms, can substantially minimize the service costs, and in the meantime, efficiently improve the QoS by adapting to the changes of mobile access patterns. • A Q-learning-based framework is proposed for service migration in mobile edges. • A policy-cache-based method is introduced to optimize the learning process of the framework. • A software-defined approach is presented to effectively implement the framework. • A prototype implementation is evaluated with the trace data gathered from Shenzhen taxi system and metro stations. [ABSTRACT FROM AUTHOR]

Computation offloading between the edge and the cloud is an effective way for deployed service to fully utilize the resources at both sides for its QoS improvement and overall cost reduction. Although the offloading problem has been intensively studied in the context of mobile computing, existing algorithms in most cases cannot be effectively migrated to the edge-cloud environment because their inter-partition communication costs are always deemed as symmetric, and their intra-partition communication costs are often ignored, which, though reasonable to the traditional case, are not valid to our settings anymore. In this article, we propose a new algorithmic approach to the offloading problem in the edge-cloud environment, where a heterogeneous model is advocated to incorporate the communication cost between co-resident tasks while considering the asymmetry of communication costs between non-coresident tasks. We prove the offloading problem with respect to this model is NP-hard, and thereby designing an efficient algorithm to obtain a sub-optimal solution. Additionally, we also show that in a homogeneous case when the intra-partition and inter-partition communication costs between any pair of interactive tasks are symmetric, an optimal offloading algorithm can be devised by transforming the problem into a classical min-cut problem. We implemented and evaluated the algorithms by offloading a PageRank-based application in a controlled edge-cloud setting. Our empirical results show that the proposed algorithm for the heterogeneous case is always efficient to find a better offloading scheme, compared with the selected existing algorithms, while for the homogeneous case, the proposed solution can efficiently achieve the optimal strategy. [ABSTRACT FROM AUTHOR]

Background: Disparities in the utilization, expenditures, and quality of care by insurance types have been well documented. Such comparisons have yet to be investigated in end-of-life (EOL) settings in China, where public insurance covers over 95% of the Chinese population. This study examined the associations between health insurance and EOL care in the last six months of life: outpatient visits, emergency department (ED) visits, inpatient services, intensive care unit (ICU) admissions, expenditures, and place of death among the cancer patients.Methods: A total of 398 patients diagnosed with cancer who survived more than 6 months after diagnosis and died from July 2015 to June 2017 in urban Yichang, China, were included. Descriptive analysis and multivariate regression models were used to investigate the bivariate and independent associations, respectively, between health insurance with EOL healthcare utilization, expenditures and place of death.Results: Urban Employee Basic Medical Insurance (UEBMI) beneficiaries visited EDs more frequently than Urban Resident-based Basic Medical Insurance (URBMI) and New Rural Cooperative Medical Scheme (NRCMS) beneficiaries (marginal effects [95% Confidence Interval]: 2.15 [1.81-2.48] and 1.92 [1.59-2.26], respectively). NRCMS and UEBMI beneficiaries had more hospitalizations than URBMI beneficiaries (1.01 [0.38-1.64] and 0.71 [0.20-1.22], respectively). Compared to URBMI beneficiaries, NRCMS beneficiaries and UEBMI beneficiaries had ¥15,722 and ¥43,241 higher expenditures. Similarly, UEBMI beneficiaries were most likely to die in hospitals, followed by NRCMS (UEBMI vs. NRCMS: 0.23 [0.11-0.36]) and URBMI (UEBMI vs. URBMI: 0.67 [0.57-0.78]) beneficiaries.Conclusions: The disproportionately lower utilization of EOL care among NRCMS and URBMI beneficiaries, compared to UEBMI beneficiaries, raised concerns regarding quality of EOL care and financial burdens of NRCMS and URBMI beneficiaries. Purposive hospice care intervention might be warranted to address EOL care for these beneficiaries in China. [ABSTRACT FROM AUTHOR]

Evaluation on urban commercial centers' attractiveness not only benefits business strategy making and location choice, but also helps traffic management and urban planning. Traditionally, it is studied using questionnaires and field research, which are labor-intensive and time-consuming. To overcome these problems when evaluating the urban commercial centers' attractiveness, massive data analytics with datasets from taxi traffic, population, area, and road networks are adopted in this paper. Taking fifteen commercial centers at Shenzhen as a case study, a Cyber-Physical-Social System is built up to deal with these massive data for statistical analysis. An "Attractiveness Degree Model" is proposed to describe the degree to which customers desire to visit a commercial center. Then attractiveness thematic maps are drawn. Results show that YiTianJiaRiGuangChang has the highest attractiveness degree even though it has a small size and low commercial value. The attractiveness degree rankings are corroborated by annual customer satisfaction survey from Shenzhen Retail Business Association. Attractiveness thematic maps show that about 50–65% visits by taxis are within 5 km range. These results can be applied to support market analysis, urban planning, traffic management, and related areas. • Commercial center is precisely defined geographically and sociologically. • Attractiveness degree is the proposed metric to evaluate the attractiveness. • A Cyber-Physical-Social System is implemented for massive data processing. • Statistical analysis and thematic maps shed new insights into attractiveness. • Commercial centers are classified into four representative types in China. [ABSTRACT FROM AUTHOR]

Background: Previous studies showed inconsistent results on risk of increased outpatient visits for cause-specific diseases associated with ambient carbon monoxide (CO).Methods: Daily data for CO exposure and outpatient visits for all-causes and five specific diseases in Yichang, China from 1st January 2016 to 31st December 2017 were collected. Generalised additive models with different lag structures were used to examine the short-term effects of ambient CO on outpatient visits. Potential effect modifications by age, sex and season were examined.Results: A total of 5,408,021 outpatient visits were recorded. We found positive and statistically significant associations between CO and outpatient visits for multiple outcomes and all the estimated risks increased with longer moving average lags. An increase of 1 mg/m3 of CO at lag06 (a moving average of lag0 to lag6), was associated with 24.67% (95%CI: 14.48, 34.85%), 21.79% (95%CI: 12.24, 31.35%), 39.30% (95%CI: 25.67, 52.92%), 25.83% (95%CI: 13.91, 37.74%) and 19.04% (95%CI: 8.39, 29.68%) increase in daily outpatient visits for all-cause, respiratory, cardiovascular, genitourinary and gastrointestinal diseases respectively. The associations for all disease categories except for genitourinary diseases were statistically significant and stronger in warm seasons than cool seasons.Conclusion: Our analyses provide evidences that the CO increased the total and cause-specific outpatient visits and strengthen the rationale for further reduction of CO pollution levels in Yichang. Ambient CO exerted adverse effect on respiratory, cardiovascular, genitourinary, gastrointestinal and neuropsychiatric diseases especially in the warm seasons. [ABSTRACT FROM AUTHOR]

In this paper we consider a new caching model that enables data sharing for network services in a cost-effective way. The proposed caching algorithms are characterized by using monetary cost and access information to control the cache replacements, instead of exploiting capacity-oriented strategies as in traditional approaches. In particular, given a stream of requests to a shared data item with respect to a homogeneous cost model, we first propose a fast off-line algorithm using dynamic programming techniques, which can generate an optimal schedule within $O(mn)$ time-space complexity by using cache, migration as well as replication to serve a $n$ -length request sequence in a $m$ -node network, substantially improving the previous results. Furthermore, we also study the online form of this problem, and present an 3-competitive online algorithm by leveraging an idea of anticipatory caching. The algorithm can serve an online request in constant time and is space efficient in $O(m)$ as well, rendering it more practical in reality. We evaluate our algorithms, together with some variants, by conducting extensive simulation studies. Our results show that the optimal cost of the off-line algorithm is changed in a parabolic form as the ratio of caching cost to transfer cost is increased, and the online algorithm is less than 2 times worse in most cases than its optimal off-line counterpart. [ABSTRACT FROM AUTHOR]

Background: Home blood pressure telemonitoring (HBPT) has great potential in improving blood pressure (BP) control among patients with hypertension. However, the longitudinal use trajectories of HBPT have not been identified yet. In addition, there has been a lack of understanding of the relationship between developmental trajectories of HBPT and BP control over time. The primary goal of this study was to identify the longitudinal trajectories of using HBPT among hypertensive patients and to explore the relationship between longitudinal trajectories of HBPT use patterns and BP control.Methods: A total of 122 hypertensive patients were enrolled consecutively in Xiling, Huayan, Baisha and Xueyuan communities in Yichang City, Hubei Province, China. Each patient was provided with a portable monitoring device which has unlimited data service at the time of enrollment. Socio-demographics (e.g. name, age, sex, marital status) were collected at baseline. Real-time data including systolic and diastolic blood pressure were automatically uploaded to cloud platform through devices. Latent class growth analysis was conducted to determine the latent trajectory of HBPT use. Joint trajectory method was used to correlate the longitudinal trajectories of HBPT utilization and BP control status.Results: Five trajectories were identified which are persistently low (47.1%), moderate with decreasing (23.9%), sharply decreasing (11.2%), high with decreasing (11.3%) and persistently high with increasing (6.6%). There was no statistically significant difference among 5 trajectories in the baseline survey in terms of age, marital status, BP (both SBP and DBP) and BP control status. However, there was a strong positive correlation between the HBPT utilization pattern and BP control status over time.Conclusions: The latent trajectories of HBPT utilization were identified in our study. However, no predictors of trajectory membership were identified. Nevertheless, we have demonstrated that HBPT was to some extent positively correlated with improved BP control, and this correlation still needs to be further proved. [ABSTRACT FROM AUTHOR]

• A general two-stage cGANs framework is proposed to learn hash codes based on the pairwise label information. • In the first stage, the conditional information is generated via a general Bayesian framework, which has several advantages over traditional methods. The first advantage is that it preserves semantic information of original data samples. Besides, it can be represented by a much lower dimensional vector, which is much more flexible than other representations. • In order to utilize both labeled and unlabeled data samples, a simple semi-supervised conditional generative adversarial hashing method is presented in the second stage. Both pairwise-based cross entropy loss and adversarial loss are introduced to make full use of these data samples. • We conduct extensive experiments to verify the effectiveness of our method. The results on three benchmark datasets have shown the superiority of our method. Hashing has been widely exploited in recent years due to the rapid growth of image and video data on the web. Benefiting from recent advances in deep learning, deep hashing methods have achieved promising results with supervised information. However, it is usually expensive to collect the supervised information. In order to utilize both labeled and unlabeled data samples, many semi-supervised hashing methods based on Generative Adversarial Networks (GANs) have been proposed. Most of them still need the conditional information, which is usually generated by the pre-trained neural networks or leveraging random binary vectors. One natural question about these methods is that how can we generate a better conditional information given the semantic similarity information? In this paper, we propose a general two-stage conditional GANs hashing framework based on the pairwise label information. Both the labeled and unlabeled data samples are exploited to learn hash codes under our framework. In the first stage, the conditional information is generated via a general Bayesian approach, which has a much lower dimensional representation and maintains the semantic information of original data samples. In the second stage, a semi-supervised approach is presented to learn hash codes based on the conditional information. Both pairwise based cross entropy loss and adversarial loss are introduced to make full use of labeled and unlabeled data samples. Extensive experiments have shown that the propose algorithm outperforms current state-of-the-art methods on three benchmark image datasets, which demonstrates the effectiveness of our method. [ABSTRACT FROM AUTHOR]

• Proposed Bayesian robit model for outlier handling in crash data. • Replaced logit/probit link with heavy-tailed Student's t distribution. • Improved posterior estimation with sandwich algorithm and data augmentation. • Tested on tunnel crash dataset, showing superior efficiency and robustness. Traffic crash datasets are often marred by the presence of anomalous data points, commonly referred to as outliers. These outliers can have a profound impact on the results obtained through the application of traditional methods such as logit and probit models, commonly used in the domain of traffic safety analysis, resulting in biased and unreliable estimates. To mitigate this issue, this study introduces a robust Bayesian regression approach, the robit model, which utilizes a heavy-tailed Student's t distribution to replace the link function of these thin-tailed distributions, effectively reducing the influence of outliers on the analysis. Furthermore, a sandwich algorithm based on data augmentation is proposed to enhance the estimation efficiency of posteriors. The proposed model is rigorously tested using a dataset of tunnel crashes, and the results demonstrate its efficiency, robustness, and superior performance compared to traditional methods. The study also reveals that several factors such as night and speeding have a significant impact on the injury severity of tunnel crashes. This research provides a comprehensive understanding of the outliers treatment methods in traffic safety studies and offers valuable recommendations for the development of appropriate countermeasures to effectively prevent severe injuries in tunnel crashes. [ABSTRACT FROM AUTHOR]

Graphic processing units (GPUs) are widely used for general-purpose computing—so-called GPGPU computing. GPUs feature a large number of architecture parameters, resulting in a huge design space. To quickly explore this design space and identify the optimum architecture for a group of widely used computing kernels, it is critical to know how important each parameter is and how strongly these parameters interact with each other. This paper proposes an ensemble-learning-based approach, called quantifying the importance and interaction of Gpgpu architecture parameters (QIG), to quantify the importance of architecture parameters and their interactions with respect to performance. QIG employs a stochastic gradient boosted regression tree to construct performance models using performance data from a random set of GPU architectures. Leveraging these models, QIG observes the impact of each architecture parameter on performance, and calculates its importance and interaction intensity with other parameters. Using 25 widely used GPGPU kernels, we demonstrate that QIG accurately ranks the importance and interaction of GPU architecture parameters while the previously proposed Plackett–Burman design does not. Moreover, we show that QIG leads to a substantially more accurate performance model compared to prior work, including Starchart and approaches using artificial neural networks and supported vector machines: average error of 4.2% for QIG versus 23+% for prior work. Finally, QIG reveals a number of interesting insights for GPU architectures running GPGPU workloads. [ABSTRACT FROM PUBLISHER]

We present a network I/O allocation framework, called Raccoon, for workload-aware VM scheduling algorithm to facilitate hybrid I/O workloads in virtual environments. Raccoon combines the strengths of paravirtual I/O and SR-IOV techniques to not only minimize the network latency, but also optimize the bandwidth utilization for workload-aware VM scheduling. In Raccoon, a limited number of VFs in SR-IOV are granted to I/O-intensive VMs while the paravirtual Network Interface Cards (vNICs) are allocated to other non-I/O-intensive VMs as the default resources. With this design, Raccoonprovides latency reduction and bandwidth guarantee under the premise that I/O-intensive VMs will always be granted the VFs to facilitate their I/O operations. The types of workloads in each VM are identified at runtime by modified XenMon. By leveraging the ACPI Hotplug technique, Raccoon can adaptively plugin and plugout the SR-IOV VFs upon the changes of VM requirements so that an efficient I/O workload-aware VM scheduling algorithm can be implemented based on the bonding driver technique. The experimental results reveal that Raccoon can combine the benefits of para-virtual I/O and SR-IOV techniques to improve the overall performance of virtualized platforms with VMs that have diverse I/O workloads. [ABSTRACT FROM AUTHOR]

Parallel file system (PFS) is commonly used in high-end computing systems. With the emergence of solid state drives (SSDs), hybrid PFS, which consists of both HDD and SSD servers, provides a practical I/O system solution for data-intensive applications. However, most existing data layout schemes are inefficient for hybrid PFS due to their unawareness of server heterogeneities and workload changes in different parts of a file. In this study, we propose a heterogeneity-aware region-level data layout scheme, HARL, to improve the data distribution of a hybrid PFS. HARL first divides a file into fine-grained, varying sized regions according to the workload features of an application, then determines appropriate file stripe sizes on servers for each region based on the performance of heterogeneous servers. Furthermore, to further improve the performance of a hybrid PFS, we propose a dynamic region-level layout scheme, HARL-D, which creates multiple replicas for each region and redirects file requests to the proper replicas with the lowest access costs at the runtime. Experimental results of representative benchmarks and a real application show that HARL can greatly improve I/O system performance, and demonstrate the advantages of HARL-D over HARL. [ABSTRACT FROM AUTHOR]

In this paper, we consider to improve scientific workflows in cloud environments where data transfers between tasks are performed via provisioned in-memory caching as a service, instead of relying entirely on slower disk-based file systems. However, this improvement is not free since services in the cloud are usually charged in a “pay-as-you-go” model. As a consequence, the workflow tenants have to estimate the amount of memory that they would like to pay. Given the intrinsic complexity of the workflows, it would be very hard to make an accurate prediction, which would lead to either oversubscription or undersubscription, resulting in unproductive spending or performance degradation. To address this problem, we propose a concept of minmax memory claim (MMC) to achieve cost-effective workflow computations in in-memory cloud computing environments. The minmax-memory claim is defined as the minimum amount of memory required to finish the workflow without compromising its maximum concurrency. With the concept of MMC, the workflow tenants can achieve the best performance via in-memory computing while minimizing the cost. In this paper, we present the procedure of how to find the MMCs for those workflows with arbitrary graphs in general and develop optimal efficient algorithms for some well-structured workflows in particular. To further show the values of this concept, we also implement these algorithms and apply them, through a simulation study, to improve deadlock resolutions in workflow-based workloads when memory resources are constrained. [ABSTRACT FROM PUBLISHER]

Data analytics is at the core of the supply chain for both products and services in modern economies and societies. Big data workloads, however, are placing unprecedented demands on computing technologies, calling for a deep understanding and characterization of these emerging workloads. In this paper, we propose ShenZhen Transportation System (SZTS), a novel big data Hadoop benchmark suite comprised of real-life transportation analysis applications with real-life input data sets from Shenzhen in China. SZTS uniquely focuses on a specific and real-life application domain whereas other existing Hadoop benchmark suites, such as HiBench and CloudRank-D, consist of generic algorithms with synthetic inputs. We perform a cross-layer workload characterization at the microarchitecture level, the operating system (OS) level, and the job level, revealing unique characteristics of SZTS compared to existing Hadoop benchmarks as well as general-purpose multi-core PARSEC benchmarks. We also study the sensitivity of workload behavior with respect to input data size, and we propose a methodology for identifying representative input data sets. [ABSTRACT FROM AUTHOR]

Hadoop is a widely-used implementation framework of the MapReduce programming model for large-scale data processing. Hadoop performance however is significantly affected by the settings of the Hadoop configuration parameters. Unfortunately, manually tuning these parameters is very time-consuming, if at all practical. This paper proposes an approach, called RFHOC, to automatically tune the Hadoop configuration parameters for optimized performance for a given application running on a given cluster. RFHOC constructs two ensembles of performance models using a random-forest approach for the map and reduce stage respectively. Leveraging these models, RFHOC employs a genetic algorithm to automatically search the Hadoop configuration space. The evaluation of RFHOC using five typical Hadoop programs, each with five different input data sets, shows that it achieves a performance speedup by a factor of 2.11 $\times$ on average and up to 7.4 $\times$ over the recently proposed cost-based optimization (CBO) approach. In addition, RFHOC's performance benefit increases with input data set size. [ABSTRACT FROM PUBLISHER]

Contactless smart card systems have gained universal prevalence in modern metros. In addition to its original goal of ticketing, the large amount of transaction data collected by the smart card system can be utilized for many operational and management purposes. This paper investigates an important problem: how to extract spatiotemporal segmentation information of trips inside a metro system. More specifically, for a given trip, we want to answer several key questions: How long does it take for a passenger to walk from the station gantry to the station platform? How much time does he/she wait for the next train? How long does he/she spend on the train? How long does it take to transfer from one line to another? This segmentation information is important for many application scenarios such as travel time prediction, travel planning, and transportation scheduling. However, in reality, we only assume that only each trip's tap-in and tap-out time can be directly obtained; all other temporal endpoints of segments are unknown. This makes the research very challenging. To the best of our knowledge, we are the first to give a practical solution to this important problem. By analyzing the tap-in/tap-out event pattern, our intuition is to pinpoint some special passengers whose transaction data can be very helpful for segmentation. A novel methodology is proposed to extract spatiotemporal segmentation information: first, for nontransfer trips, by deriving the boarding time between the gantry and the platform, and then, for with-transfer trips, by deriving the transfer time. Evaluation studies are based on large-scale real-system data of the Shenzhen metro system, which is one of the largest metro systems in China and serves millions of passengers daily. Onsite investigations validate that our algorithm is accurate and that the average estimation error is only around 15%. [ABSTRACT FROM AUTHOR]

Graphics processing units (GPU), due to their massive computational power with up to thousands of concurrent threads and general-purpose GPU (GPGPU) programming models such as CUDA and OpenCL, have opened up new opportunities for speeding up general-purpose parallel applications. Unfortunately, pre-silicon architectural simulation of modern-day GPGPU architectures and workloads is extremely time-consuming. This paper addresses the GPGPU simulation challenge by proposing a framework, called GPGPU-MiniBench, for generating miniature, yet representative GPGPU workloads. GPGPU-MiniBench first summarizes the inherent execution behavior of existing GPGPU workloads in a profile. The central component in the profile is the Divergence Flow Statistics Graph (DFSG), which characterizes the dynamic control flow behavior including loops and branches of a GPGPU kernel. GPGPU-MiniBench generates a synthetic miniature GPGPU kernel that exhibits similar execution characteristics as the original workload, yet its execution time is much shorter thereby dramatically speeding up architectural simulation. Our experimental results show that GPGPU-MiniBench can speed up GPGPU architectural simulation by a factor of 49$\times$ on average and up to 589 $\times$ , with an average IPC error of 4.7 percent across a broad set of GPGPU benchmarks from the CUDA SDK, Rodinia and Parboil benchmark suites. We also demonstrate the usefulness of GPGPU-MiniBench for driving GPU architecture exploration. [ABSTRACT FROM PUBLISHER]