Search

Background Hospital administrators are often challenged with overcrowding at hospitals. The study hospital receives referred patients; however, they have to wait in long queues even for getting registered. This was a cause of concern for hospital administrators. The study was undertaken to find an amicable solution to the queues at registration using Queuing Theory. Method This observational and interventional study was carried out in a tertiary care ophthalmic hospital. In the first phase, data of service time and arrival rate was collected. The queuing model was built using the coefficient of variation (CoV) of the observed times. Server utilization for new patient registration was found to be 1.21 and was 0.63 for revisit patients. Scenario-based simulation carried out using free software for optimal utilization of both types of servers. Recommendations made to combine the registration process and to increase one server were implemented. In the second phase, after one year, patient registration data were collected and compared for the number of patients registered using SPSS 17. Results Number of patients registered within the registration timings increased whereas the number of patients registered after the registration timings decreased significantly at 95% CI with a p-value of less than 0.001. Queues finished early and more number of patients were registered in the same time. Conclusion Using queuing theory, the bottleneck of the systems can be identified. Scenario and software-based simulations provide solutions to the problem of queues. The study is an application of Queuing Theory with a focus on efficient resource utilization. It can be replicated in an organization with limited resources facing the challenge of queues.

Radio frequency identification technology has been widely used to verify the presence of items in many applications such as warehouse management and supply chain logistics. In these applications, the challenge of how to timely identify the missing tags (namely tag searching or missing tag identification) is a key focus. Existing missing tag identification solutions have not achieved their full potentials because collision slots have not been well explored. In this paper, we propose an approach named collision resolving based missing tag identification (CR-MTI) to break through the performance bottleneck of existing missing tag identification protocols. In CR-MTI, multiple tags are allowed to respond with different binary strings in a collision slot. Then, the reader can verify them together by using the bit tracking technology and particularly designed string, thereby significantly improve the time efficiency. CR-MTI also reduces the number of messages transmitted by the reader using customized coding. We further explore the optimal parameter settings to maximize the performance of our proposed CR-MTI. Extensive simulation results show that our proposed CR-MTI outperforms prior art in terms of time efficiency, total executive time and communication complexity.

The container selection processing performance analyses huge data with minimal resources and the lowest latency. The elastic management applications execute the containers with the specific hierarchy of virtual processing and machine management. The container that has performance degradation implicit provision of least expensive containers with minimal resources helps to increase the performance of containers. The specific data and stream processing prompts scrutinizing of data through the container selection process through different methodologies. To exterminate the bottleneck problem that selects efficient and required size, processing speed, and its reliability of guiding the batch processing of containers. The extensive neural learning handles container optimality involves a dynamic selection of appropriate containers in cloud service providers. The cloud service providers along with container selection contain batch processing and stream processing allocates efficient container-specific selection appropriately. In the huge data segregation of data processed data that emphasizes multiple data scrutinizing.

In a time where data traffic is booming, Software-defined Networking (SDN) is becoming the most plausible solution to cope with the conventional networks shortcomings. The separation of the control (software) and the data (hardware) planes makes of SDN a technology-enabled to a multitude of new and promising applications and use cases for 5G network technology. Yet, initial SDN deployments considered having a single controller for the entire network. However, this arrangement is deemed to be counterproductive in large-scale and ultra-reliable networks based SDN as the controller might become the systems bottleneck. To address this issue, several multi-controller architectures were proposed. They are of three types: flat, hierarchical or hybrid. For theses architectures to be potent for 5G core networks, efficient path computation and flow tables update mechanisms are needed. In this paper, we propose HybCon, a hybrid SDN-based distributed Cloud architecture having the control plane distributed over different controller types (i.e., Fog, Edge, SDN). When a path computation request is received, HybCon involves some/all of the controllers and uses a multi-priority queueing model along with node parallelism to expedite the path computation and to cut down the synchronization overhead.

Binarized neural networks (BNNs) have drawn significant attention in recent years, owing to great potential in reducing computation and storage consumption. While it is attractive, traditional BNNs usually suffer from slow convergence speed and dramatical accuracy-degradation on large-scale classification datasets. To minimize the gap between BNNs and deep neural networks (DNNs), we propose a new framework of designing BNNs, dubbed Hyper-BinaryNet, from the aspect of enhanced information-flow. Our contributions are threefold: 1) Considering the capacity-limitation in the backward pass, we propose an 1-bit convolution module named HyperConv. By exploiting the capacity of auxiliary neural networks, BNNs gain better performance on large-scale image classification task. 2) Considering the slow convergence speed in BNNs, we rethink the gradient accumulation mechanism and propose a hyper accumulation technique. By accumulating gradients in multiple variables rather than one as before, the gradient paths for each weight increase, which escapes BNNs from the gradient bottleneck problem during training. 3) Considering the ill-posed optimization problem, a novel gradient estimation warmup strategy, dubbed STE-Warmup, is developed. This strategy prevents BNNs from the unstable optimization process by progressively transferring neural networks from 32-bit to 1-bit. We conduct evaluations with variant architectures on three public datasets: CIFAR-10/100 and ImageNet. Compared with state-of-the-art BNNs, Hyper-BinaryNet shows faster convergence speed and outperforms existing BNNs by a large margin.

With the rapid development of 5G, Mobile Edge Computing (MEC) paradigm has emerged to enable various devices or servers at the network edge to contribute their computing capacity for reducing communication delay. A fundamental problem is to preserve satisfactory quality-of-service (QoS) for mobile users in light of densely dispersed wireless communication environment and capacity-constrained MEC nodes. Such user-perceived QoS, typically in terms of the end-to-end delay, is highly vulnerable to both access network bottleneck and communication delay. Previous works primarily focused on optimizing the communication delay through dynamic service placement, while ignoring the critical effect of access network selection on the access delay. This paper studies the problem of jointly optimizing the access network selection and service placement for MEC, with the objective of improving QoS in a cost-efficient manner by judiciously balancing the access delay, communication delay, and service switching cost. We propose an efficient online framework to decompose this long-term time-varying optimization problem into a series of one-shot subproblems. To address the NP-hardness of the one-shot problem, we design a computationally-efficient two-phase algorithm, which achieves a near-optimal solution. Both theoretical analysis on the optimality gap and trace-driven simulations are conducted to validate the efficacy of our proposed solution.

With the rise in popularity and its open system architecture, Android has become vulnerable to malicious attacks. There are several malware detection approaches available to fortify the Android operating system from such attacks. These malware detectors classify target applications based on the patterns found in the features present in the Android applications. As the analytics data continues to grow, it negatively impacts the Android defense mechanism. A large number of irrelevant features has become the performance bottleneck of the detection mechanism. This paper presents a multi-tiered feature selection model, which can discover relevant and significant features for improving the accuracy of malware detection approaches. The proposed method applies five machine learning classification techniques to the selected feature set. This work presents the Optimal Static Feature Set (OSFS) and, Most Important Features (MIFs) discovered with each machine learning approach. Rigorous testing and analysis show that Random Forest classification achieves the highest Accuracy rate of 96.28%.

With the rapid growth of Internet, plenty of applications has been implemented for users to exchange resources with each other over networks. As a result, it has been a critical challenge to efficiently and fairly allocate resource among participating agents. Motivated by the famous BitTorrent system, a BD Mechanism for resource exchange allocation is proposed, which is based on a combinatorial structure called bottleneck decomposition. The mechanism has been shown to lead to market equilibrium, a state optimizing each individual's utility, and to be truthful, that is robust against agents' strategic behaviors. However, the crux on how to compute a bottleneck decomposition of any graph remains untouched. In this paper, we focus on the computation of bottleneck decomposition to fill the blanks and prove that the bottleneck decomposition of a network can be computed in O(n^6\log(nU)) time, where n and U are the number of agents and the maximal amount of resource any agent has in the network. Based on the bottleneck decomposition, a fair allocation in a resource exchange network can be obtained in polynomial time. We further carefully discuss the fairness in resource exchange scenario, showing the derived allocation satisfies several common fairness notions simultaneously.

Workload characteristics are vital for both data center operation and job scheduling in co-located data centers, where online services and batch jobs are deployed on the same production cluster. In this paper, a comprehensive analysis is conducted on Alibaba‘s cluster-trace-v2018 of a production cluster of 4034 machines. The findings and insights are the following: (1) The workload on the production cluster poses a daily cyclical fluctuation, in terms of CPU and disk I/O utilization, and the memory system has become the performance bottleneck of a co-located cluster. (2) Batch jobs including their tasks and derived instances can be approximated as Zipf distribution. However, for all batch jobs with directed acyclic graph dependency, they suffer from co-location with online services since the online services are highly prioritized. (3) The resource usages of containers have similar cyclical fluctuation consistent with the whole cluster, while their memory usages remain approximately constant. (4) The number of batch jobs co-located with online services is dependent on the mispredictions per kilo instructions of online services. In order to guarantee the QoS of online services, when the MPKI of online services rises, the number of batch jobs to be co-located on the same machine should decrease.

With the promising advances in the telecommunication and automotive industries, there are increasing concerns on the reliability of Intelligent Transportation Systems (ITS) applications and standardization efforts. Nowadays, vehicles are empowered with innovative communications and sensing capabilities to accelerate next-generation Connected and Autonomous Vehicles (CAV) or Vehicle-to-Everything (V2X) communications. Furthermore, V2X is based on many radio access technologies including Dedicated Short-Range Communications (DSRC), Cellular-V2X, and millimeter-wave. It is expected that a single technology is incapable to support the diverse QoS and reliability requirements of vehicular safety and non-safety applications. Therefore, recent research works have investigated and developed different approaches to enable multi-radio access technologies (Multi-RAT) in V2X networks to support these multitudes of requirements. In addition, the heterogeneity of diverse radio access technologies and rigidity in their deployment increases the complexity of the network management. Complications may also arise from inefficient resource utilization due to the complexity of an optimum RAT selection or multi-link establishment through Multi-RATs. Therefore, a global monitoring paradigm such as Software-Defined Vehicular Network (SDVN) has been identified as a solution to reduce this bottleneck, by adding flexibility and agility to the vehicular network. This paper extensively reviews recent related works on Multi-RAT V2X networks focusing on the opportunistic selection mechanism of RAT, allocation of radio resources, and simultaneous establishment of multi-link over a heterogeneous network. Consequently, we have outlined the SDVN architecture, layers, features, and significance towards enhancing the performance of such Multi-RAT networks. Finally, several potential challenges with probable solutions and opportunities are discussed.

Human-robot collaborative assembly (HRCA) is vital for achieving high-level flexible automation for mass personalization in todays smart factories. However, existing works in both industry and academia mainly focus on adaptive robot planning, while seldom consider human operators intentions in advance. Hence, it hinders the HRCA transition towards a proactive manner. To overcome the bottleneck, this research proposes a multimodal transfer learning-enabled action prediction approach, serving as the prerequisite to ensure the Proactive HRCA. Firstly, a multimodal intelligence-based action recognition approach is proposed to predict ongoing human actions by leveraging the visual stream and skeleton stream with short-time input frames. Secondly, a transfer learning-enabled model is adapted to transfer learned knowledge from daily activities to industrial assembly operations rapidly for online operator intention analysis. Thirdly, a dynamic decision-making mechanism including the robotic decision and motion control is described to allow mobile robots to assist operators in a proactive manner. Lastly, an aircraft bracket assembly task is demonstrated in the lab environment, and the comparative study result shows that the proposed approach outperforms other state-of-the-art ones for efficient action prediction.

The reverse distribution of renewable energy resources and load centers makes exploring the optimal transmission mode of long-distance and large-scale renewable hydrogen the key to solving the bottleneck of renewable hydrogen development. This study incorporates hydrogen pipeline (HGP), natural gas pipeline (NGP), and Ultra High Voltage (UHV) into an optimal planning model framework and analyzes the optimal transmission mode, quantity, network, and cost of large-scale renewable hydrogen in China. Constructing a sensitivity analysis framework, this study also investigates the optimal transmission mode changes under different scenarios. The results show that the optimal mode of large-scale renewable hydrogen transmission in the province is NGP, and 5.4% of supply level is the critical point to export renewable hydrogen inter-provincially. It switches to the combination of NGP and HGP when the unit transmission costs of these elements decrease to a certain proportion simultaneously or switches to HGP when the unit transmission cost of HGP decreases more than that of NGP. The complementary transmission mode of NGP and UHV is the optimal mode for inter-provincial transmission, and the HGP can be put into use for inter-provincial transmission only when the unit transmission cost of which is reduced to less than 25%. Jilin is the key node in the NGP network, and Tibet and Gansu are the key nodes in UHV network, and the participation or absence of which will have significant impact on the renewable hydrogen transmission system. Only minor adjustments to the transmission technical parameters of NGP or HGP can promote the qualitative overflight of the optimal transmission volume of them so as to achieve the target optimization at the minimum cost.

Federated Learning (FL) serves privacy-preserving collaborative learning among multiple isolated parties, while retaining their privacy data locally. Cross-device and cross-silo FL have achieved great success in cross-domain applications, in which the scarce communication resource is the primary bottleneck. Driven by the need to combine heterogeneous machines from different parties to build a shared data center, we found intra-domain FL, a new type of FL in which isolated parties collaborate in the shared data center, and strong computational heterogeneity becomes the primary bottleneck. To mitigate the training inefficiency caused by stragglers, this paper proposes an efficient synchronization algorithm ESync, which allows parties to train different iterations locally under the coordination of a novel scheduler State Server. We give the boundaries of weight divergence and optimality gap of ESync, and analyze the trade-off between convergence accuracy and communication efficiency. Extensive experiments are conducted to compare ESync with SSGD, ASGD, DC-ASGD, FedAvg, FedAsync, TiFL and FedDrop under strong computational heterogeneity. Numerical results show that ESync achieves great speed up without loss of accuracy, and therefore demonstrate the effectiveness of ESync in both training efficiency and converged accuracy.

Real-time inference of deep learning models on embedded and energy-efficient devices becomes increasingly desirable with the rapid growth of artificial intelligence on edge. Specifically, to achieve superb energy-efficiency and scalability, efficient parallelization of single-pass deep neural network (DNN) inference on chip multiprocessor (CMP) architectures is urgently required by many time-sensitive applications. However, as the number of processing cores scales up and the performance of cores has grown much fast, the on-chip inter-core data movement is prone to be a performance bottleneck for computation. To remedy this problem and further improve the performance of network inference, in this work, we introduce a communication-aware DNN parallelization technique called CAP, by exploiting the elasticity and noise-tolerance of deep learning algorithms on CMP. Moreover, in the hope that the conducted studies can provide new design values for real-time neural network inference on embedded chips, we also have evaluated the proposed approach on both multi-core Neural Network Accelerators (NNA) chips and general-purpose chip-multiprocessors. Our experimental results show that the proposed CAP can achieve 1.13x-1.65x speedups and 51%-90% on-chip communication energy reduction for different neural networks while maintaining the inference accuracy, compared to baseline approaches.

This paper examines the morning commute problem when both household commuters and individual commuters are considered in a Y-shaped network with two upstream links and a single downstream link. The household parents daily pass through an upstream bottleneck with a limited capacity before a school and drop off their children. Then, they traverse the downstream bottleneck common to both household and individual commuters and arrive at the workplace. We explore the effects of staggering policy, that is, staggering work and school start times, on the distribution of traffic congestion and social welfare. We analytically solve all the equilibrium cases and reveal all the traffic congestion patterns. The results reveal that the staggering policy may be harmful in certain cases. When the demand of individuals is relatively low, the staggering policy may not improve social welfare. When the demand of individuals is high, social welfare can be significantly improved if the schedule gap between the work start time and school start time is optimized. The effects of the staggering policy on system performance are examined. We derive a Pareto frontier, which provides a good candidate set for policymakers when the two system performance measures, that is, the total system cost and the total congestion cost, are considered. Our results show that the optimal staggering policy on system performance depends on the demand distribution of the two groups. When the demand of individuals is high, there exists a unique optimal staggering policy that optimizes system performance. However, when the demand of individuals is low, the optimal staggering policy should be selected from the Pareto frontier. Furthermore, we re-examine the capacity expansion paradox under the staggering policy. Our study shows the capacity expansion at the downstream bottleneck can always reduce the total system cost. However, the paradoxical phenomenon may arise when the capacity of the upstream bottleneck is expanded, but it can be eliminated if the schedule gap is properly designed. Ministry of Education (MOE) Nanyang Technological University This research was funded by National University of Singapore, Singapore Ministry of Education Academic Research FundTier 2 [Grant MOE2017-T2-2-128]; Dalian University of Technology, National Natural Science Foun-dation of China [Grant 71533001]; Nanyang Technological University, Singapore Ministry of Educa-tion Academic Research Fund [Grant MOE2017-T2-2-093]; Dalian University of Technology, ChinaScholarship Council.

The cost of transferring data between the off-chip memory system and compute unit is the fundamental energy and performance bottleneck in modern computing systems. Furthermore, with the advent of emerging data-intensive applications and technology scaling, this bottleneck has continuously increased. To overcome these difficulties, Near Memory Processing (NMP) based on 3D die stacking becomes a potential technology to transform the computation-centric system towards memory-centric system. In this work, we explore the feasibility and efficacy of a NMP architecture based on an emerging Non-Volatile Memory technology (NVM) for data-intensive applications and compare it with the conventional 3D-stacked NMP architecture based on DRAM. We demonstrate the effectiveness of our approach with experimental results.

It's common practice to speed up machine learning (ML) training by distributing it across a cluster of computing nodes. Data-parallel distributed ML (DML) training relieves the pressure of computing node; however, the communication traffic introduced during parameter synchronization becomes bottleneck of DML training. Regarding communication bottleneck, we identify two primary causes: high contention of concurrent communication and large volume of redundant transmission in the push and pull stage while synchronizing parameters. To address these issues, we propose a novel Group Stale Synchronous Parallel (GSSP) scheme, which divides the nodes into groups and coordinates groups to synchronize in a circular order. GSSP mitigates network contention and is proven to converge. We provide analysis of the optimal group's number based on bandwidth and buffer size. For reducing traffic redundancy, we propose a multicast-based scheme, which generates multicast trees by minimizing links overlap and allocates transmission rate to multicast flows by solving the min-max optimization problem. Finally, we conduct extensive simulations to evaluate the performance of our proposals. We simulate parameter transmission of All-Reduce and parameter server in Fat-Tree with traffics trace of ML models. Simulation results show that our proposals provide communication-efficiency for DML training by mitigating contention and reducing redundancy.

While our recent deep fuzzy classifier DSA-FC, which stacks adversarial interpretable Takagi-Sugeno-Kang (TSK) fuzzy sub-classifiers, shares its promising classification, its training speed will become very slow and even intolerable for large-scale datasets, due to successive training on all training samples with their random gradient based updates along each layer of its stacked structure. In order to circumvent this bottleneck issue, a fast training algorithm FTA is developed in this study by downsizing fuzzy rules with the proposed gradient guided learning for each sub-classifier at each layer of DSA-FC on large-scale datasets. The core of FTA is to assure fast training of each sub-classifier at each layer of DSA-FC, which first generates first-order smooth gradient guided information by means of the proposed top-k fuzzy rules selected from all fuzzy rules in each sub-classifier, and then quickly updates the current inputs in terms of such information, which will be taken as the inputs of the sub-classifier at the next layer. Our theoretical analysis reveals that the proposed gradient guided learning indeed enhances the generalization capability of a deep fuzzy classifier with or without adversarial attacks on outputs. Experimental results on large datasets demonstrate that FTA indeed trains the deep fuzzy classifier DSA-FC quickly with enhanced generalization capability.

Decentralized learning has gained great popularity to improve learning efficiency and preserve data privacy. Each computing node makes equal contribution to collaboratively learn a Deep Learning model. The elimination of centralized Parameter Servers (PS) can effectively address many issues such as privacy, performance bottleneck and single-point-failure. However, how to achieve Byzantine Fault Tolerance in decentralized learning systems is rarely explored, although this problem has been extensively studied in centralized systems. In this paper, we present an in-depth study towards the Byzantine resilience of decentralized learning systems with two contributions. First, from the adversarial perspective, we theoretically illustrate that Byzantine attacks are more dangerous and feasible in decentralized learning systems: even one malicious participant can arbitrarily alter the models of other participants by sending carefully crafted updates to its neighbors. Second, from the defense perspective, we propose UBAR, a novel algorithm to enhance decentralized learning with Byzantine Fault Tolerance. Specifically, UBAR provides a Uniform Byzantine-resilient Aggregation Rule for benign nodes to select the useful parameter updates and filter out the malicious ones in each training iteration. It guarantees that each benign node in a decentralized system can train a correct model under very strong Byzantine attacks with an arbitrary number of faulty nodes. We conduct extensive experiments on standard image classification tasks and the results indicate that UBAR can effectively defeat both simple and sophisticated Byzantine attacks with higher performance efficiency than existing solutions.

Existing domain adaptation (DA) methods generally assume that different domains have identical label space, and the training data are only sampled from a single domain. This unrealistic assumption is quite restricted for real-world applications, since it neglects the more practical scenario, where the source domain can contain the categories that are not shared by the target domain, and the training data can be collected from multiple modalities. In this article, we address a more difficult but practical problem, which recognizes RGB images through training on RGB-D data under the label space inequality scenario. There are three challenges in this task: 1) source and target domains are affected by the domain mismatch issue, which results in that the trained models perform imperfectly on the test data; 2) depth images are absent in the target domain (e.g., target images are captured by smartphones), when the source domain contains both the RGB and depth data. It makes the ordinary visual recognition approaches hardly applied to this task; and 3) in the real world, the source and target domains always have different numbers of categories, which would result in a negative transfer bottleneck being more prominent. Toward tackling the above challenges, we formulate a deep model, called visual-depth matching network (VDMN), where two new modules and a matching component can be trained in an end-to-end fashion jointly to identify the common and outlier categories effectively. The significance of VDMN is that it can take advantage of depth information and handle the domain distribution mismatch under label inequality simultaneously. The experimental results reveal that VDMN exceeds the state-of-the-art performance on various DA datasets, especially under the label inequality scenario.

Drug metabolite identification is a bottleneck of drug metabolism studies due to the need for time-consuming chromatographic separation and structural confirmation. Ion mobility-mass spectrometry (IM-MS), on the other hand, separates analytes on a rapid (millisecond) time scale and enables the measurement of collision cross section (CCS), a unique physical property related to an ion's gas-phase size and shape, which can be used as an additional parameter for identification of unknowns. A current limitation to the application of IM-MS to the identification of drug metabolites is the lack of reference CCS values. In this work, we assembled a large-scale database of drug and drug metabolite CCS values using high-throughput in vitro drug metabolite generation and a rapid IM-MS analysis with automated data processing. Subsequently, we used this database to train a machine learning-based CCS prediction model, employing a combination of conventional 2D molecular descriptors and novel 3D descriptors, achieving high prediction accuracies (0.8-2.2% median relative error on test set data). The inclusion of 3D information in the prediction model enables the prediction of different CCS values for different protomers, conformers, and positional isomers, which is not possible using conventional 2D descriptors. The prediction models, dmCCS, are available at https://CCSbase.net/dmccs_predictions.

Three-dimensional (3D) imaging, such as micro-computed tomography (micro-CT), is increasingly being used by organismal biologists for precise and comprehensive anatomical characterization. However, the segmentation of anatomical structures remains a bottleneck in research, often requiring tedious manual work. Here, we propose a pipeline for the fully-automated segmentation of anatomical structures in micro-CT images utilizing state-of-the-art deep learning methods, selecting the ant brain as a test case. We implemented the U-Net architecture for 2D image segmentation for our convolutional neural network (CNN), combined with pixel-island detection. For training and validation of the network, we assembled a dataset of semi-manually segmented brain images of 94 ant species. The trained network predicted the brain area in ant images fast and accurately; its performance tested on validation sets showed good agreement between the prediction and the target, scoring 80% Intersection over Union (IoU) and 90% Dice Coefficient (F1) accuracy. While manual segmentation usually takes many hours for each brain, the trained network takes only a few minutes. Furthermore, our network is generalizable for segmenting the whole neural system in full-body scans, and works in tests on distantly related and morphologically divergent insects (e.g., fruit flies). The latter suggest that methods like the one presented here generally apply across diverse taxa. Our method makes the construction of segmented maps and the morphological quantification of different species more efficient and scalable to large datasets, a step toward a big data approach to organismal anatomy.

The lack of publicly available, large, and unbiased datasets is a key bottleneck for the application of machine learning (ML) methods in synthetic chemistry. Data from electronic laboratory notebooks (ELNs) could provide less biased, large datasets, but no such datasets have been made publicly available. The first real-world dataset from the ELNs of a large pharmaceutical company is disclosed and its relationship to high-throughput experimentation (HTE) datasets is described. For chemical yield predictions, a key task in chemical synthesis, an attributed graph neural network (AGNN) performs as good or better than the best previous models on two HTE datasets for the Suzuki and Buchwald-Hartwig reactions. However, training of the AGNN on the ELN dataset does not lead to a predictive model. The implications of using ELN data for training ML-based models are discussed in the context of yield predictions.

High-level synthesis (HLS) is a well established framework used to translate high level algorithmic behaviours into hardware designs. Despite the enduring research efforts, a major prevailing bottleneck of HLS is the large gap between the design and verification processes. At present, RTL simulation is the primary platform used for HLS design verification. Although most of the state-of-art RTL simulators provide an abstracted user friendly platform for verification, they are undesirably slow and sometimes incomprehensible to non-FPGA experts to debug. The alternative software simulators (C simulation) introduced by commercial HLS tools render faster simulation, but are not cycle accurate and are not capable to estimate design performance. In this paper, we introduce an automatic cycle accurate simulation tool, FastSim, that manipulates certain unique features of HLS design to extract a concise, well indented and debug friendly C behaviour from the synthesized RTL. Our simulation tool ensures RTL correctness, provides cycle accuracy, accurate performance estimation and renders on an average around 300 times faster simulation compared to RTL simulators and comparable performance to that of software C simulators.

The processing and analysis of large-scale journey trajectory data is becoming increasingly important as vehicles become ever more prevalent and interconnected. Mapping these trajectories onto a road network is a complex task, largely due to the inevitable measurement error generated by GPS sensors. Past approaches have had varying degrees of success, but achieving high accuracy has come at the expense of performance, memory usage, or both.In this paper, we solve these issues by proposing a map matching algorithm based on Hidden Markov Models (HMM). The proposed method is shown to be more efficient when compared against a traditional HMM based map matching method, whilst maintaining high accuracy and eschewing any requirements for CPU-intensive and memory-expensive pre-processing. The proposed algorithm offers a method for significantly accelerating transition-probability calculations using instances of high data-availability, which have previously been a large bottleneck in map matching algorithm performance. It is shown that this can be accomplished with the application of road-network segmentation combined with a spatially-aware heuristic. Experiments are performed using two different datasets, with over 9 hours of GPS samples. We show that the proposed framework is able to offer a reduction in run-time of over 90% with no significant effect on the algorithm's accuracy when compared against the traditional HMM approach.

We observe that data access and processing takes a significant amount of time in large-scale deep learning training tasks (DLTs) on image datasets. Three factors contribute to this problem: (1) the massive and recurrent accesses to large numbers of small files; (2) the repeated, expensive decoding computation on each image, and (3) the frequent communication between computation nodes and storage nodes. Existing work has addressed some aspects of these problems; however, no end-to-end solutions have been proposed. In this article, we propose DIESEL+, an all-in-one system which accelerates the entire I/O pipeline of deep learning training tasks. DIESEL+ contains several components: (1) local metadata snapshot; (2) per-task distributed caching; (3) chunk-wise shuffling; (4) GPU-assisted image decoding and (5) online region-of-interest (ROI) decoding. The metadata snapshot removes the bottleneck on metadata access in frequent reading of large numbers of files. The per-task distributed cache across the worker nodes of a DLT task to reduce the I/O pressure on the underlying storage. The chunk-based shuffle method converts small file reads into large chunk reads, so that the performance is improved without sacrificing the training accuracy. The GPU-assisted image decoding and the online ROI method minimize the image decoding workloads and reduce the cost of data movement between nodes. These techniques are seamlessly integrated into the system. In our experiments, DIESEL+ outperforms existing systems by a factor of two to three times on the overall training time.

Face detection generally requires prior boxes and an extra non-maximum suppression(NMS) post-processing in modern deep learning methods. However, anchor design and anchor matching strategy significantly affect the performance of face detectors, so we have to spend a lot of time on anchor designing for different business scenarios. The other issue is that NMS cannot be easily parallelized and it may become a bottleneck of detection speed. In this paper, we propose a simple yet efficient pure convolutional neural network face detection method, named dual-branch center face detector(DBCFace for short), which solve face detection via a dual branch fully convolutional framework without extra anchor design and NMS. Extensive experiments are conducted on four popular face detection benchmarks, including AFW, PASCAL face, FDDB, and WIDER FACE, demonstrating that our method is comparable with state-of-the-art methods while the speed is faster.

The past decades have witnessed the tremendous success of Conventional Magnetic Recording (CMR) based Hard Disk Drives (HDDs) in data storage. To eliminate the bottleneck of CMR based HDDs in providing higher areal density, an emerging Interlaced Magnetic Recording (IMR) is capable of achieving higher areal density with limited changes to disk makeup. Nevertheless, existing approaches for IMR based HDDs may suffer serious read and write performance degradation as compared with CMR based HDDs. Thus, this paper presents a device-level solution, namely MAGIC translation layer, which aims at MAkinG IMR based HDDs perform like CMR based HDDs in terms of comparable access performance. Specifically, not merely trying to improve the performance of raw IMR based HDDs, this work, for the first time, moves one step forward to minimize the performance gap between IMR and CMR based HDDs. Technically, by 1) fully utilizing two special CMR-like potentials of IMR and 2) gracefully trading the sequential access performance as space usage increases, MAGIC minimizes track rewriting overheads to achieve CMR-like performance. Our results reveal that MAGIC not only improves the write performance compared with existing designs, but also has potential to approach read and write performance of CMR based HDD.

Light detection and ranging (LiDAR) plays an indispensable role in autonomous driving technologies, such as localization, map building, navigation and object avoidance. However, due to the vast amount of data, transmission and storage could become an important bottleneck. In this article, we propose a novel compression architecture for multi-line LiDAR point cloud sequences based on clustering and convolutional long short-term memory (LSTM) networks. LiDAR point clouds are structured, which provides an opportunity to convert the 3D data to 2D array, represented as range images. Thus, we cast the 3D point clouds compression as a range image sequence compression problem. Inspired by the high efficiency video coding (HEVC) algorithm, we design a novel compression framework for LiDAR data that includes two main techniques: intra-prediction and inter-prediction. For intra-frames, inspired by the depth modeling modes (DMM) adopted in 3D-HEVC, we develop a clustering-based intra-prediction technique, which can utilize the spatial structure characteristics of point clouds to remove the spatial redundancy. For inter-frames, we design a prediction network model using convolutional LSTM cells. The network model is capable of predicting future inter-frames using the encoded intra-frames. As a result, temporal redundancy can be removed. Experiments on the KITTI dataset demonstrate that the proposed method achieves an impressive compression ratio (CR), with 4.10% at millimeter precision, which means the point clouds can compress to nearly 1/25 of their original size. Additionally, compared with the well-known octree, Google Draco, and MPEG TMC13 methods, our algorithm yields better performance in compression ratio.

The bottleneck between the processor and memory is the most significant barrier to the ongoing development of efficient processing systems. Therefore, a research effort begun to shift from processor-centric architectures to memory-centric architectures. Various in-memory processor architectures have been proposed to break this barrier to pave the way for ever-demanding memory-bound applications. Associative in-memory processing is a successful candidate for truly in-memory computing, in which processor and memory are combined in the same location to eliminate the expensive data access costs. The architecture exhibits an unmatched advantage for data-intensive applications due to its memory-centric design principles. On the other hand, this advantage can be revealed fully by an efficient design methodology. This study puts further progressive effort by proposing a hardware/software design methodology for associative in-memory processors. The methodology aims to decrease energy consumption and area requirement of the processor architecture specifically programmed to perform a given task. According to the evaluation of nine different benchmarks, such as fast Fourier transform and multiply-accumulate, the proposed design flow accomplishes an average 7% reduction in memory area and 18% savings in total energy consumption.

A practical min-sum algorithm is associated with tree-based comparison units for the check-node operation, being a major bottleneck in designing low-cost and energy-efficient low-density parity-check (LDPC) decoders. In this paper, we present a cost-effective LDPC decoder architecture by changing its internal computing resolution for the power-hungry check-node processing. The proposed mixed-resolution comparison offers significant advantages in terms of both area and energy, while achieving error-correcting performance within 0.3 dB of the previous normalized min-sum (NMS) algorithm for a (1644, 1408) quasi-cyclic LDPC code of the 5G New Radio specifications. Compared to the baseline NMS architecture, the proposed decoder in a 65-nm CMOS technology reduces the hardware complexity and the power consumption by 28.4% and 23.1%, respectively, enhancing the area efficiency by more than 88.2%.

Serverless computing is becoming a promising paradigm for Distributed Deep Neural Network (DDNN) training in the cloud, as it allows users to decompose complex model training into a number of $functions$ without managing virtual machines or servers. Though provided with a simpler resource interface ( $i.e.,$ function number and memory size), inadequate function resource provisioning (either under-provisioning or over-provisioning) easily leads to $unpredictable$ DDNN training performance in serverless platforms. Our empirical studies on AWS Lambda indicate that, such $unpredictable performance$ of serverless DDNN training is mainly caused by the resource bottleneck of Parameter Servers (PS) and small local batch size. In this paper, we design and implement $\lambda DNN$ , a cost-efficient function resource provisioning framework to provide predictable performance for serverless DDNN training workloads, while saving the budget of provisioned functions. Leveraging the PS network bandwidth and function CPU utilization, we build a $lightweight$ analytical DDNN training performance model to enable our design of $\lambda DNN$ resource provisioning strategy, so as to guarantee DDNN training performance with serverless functions. Extensive prototype experiments on AWS Lambda and complementary trace-driven simulations demonstrate that, $\lambda DNN$ can deliver predictable DDNN training performance and save the monetary cost of function resources by up to 66.7%, compared with the state-of-the-art resource provisioning strategies, yet with an acceptable runtime overhead.

The Internet of Things (IoT) is a highly integrated application of the advanced information technology, which is expected to bring convenience for daily life and improve the efficiency of industrial production. Owing to the limitation of battery capacity and the broadcast nature of IoT nodes, IoT networks face the bottleneck of energy shortage and security vulnerability. In recent years, the emerging technologies of wireless powered communication network (WPCN) and physical layer security (PLS) are regarded as potential solutions to allow IoT nodes to harvest energy from radio frequency (RF) and ensure the secure data delivery. How to efficiently allocate energy for IoT devices to improve throughput while guaranteeing secure data transmission is a challenging problem. In this paper, we consider a WPCN with the existence of an eavesdropper, who is trying to eavesdrop the data transmitted from a certain node to the hybrid sink (H-sink). In the proposed system, the nodes first harvest energy from the H-sink, then transmit the information to the H-sink and generate interference to the eavesdropper. We first formulate the sum-throughput maximization problem as the non-linear optimization problem and find its closed-form solution by Lagrangian method. We further design an efficient algorithm to derive the optimal numerical results to make up for the defect that the closed-form solution may not meet the explicit constraints. Furthermore, we propose a simple and reasonable method, ratio method (RM), based on the observation of the optimal solution, and make a comparison between the proposed method and the most common method, same interference power method (SIPM).

CNNs outperform traditional machine learning algorithms across a wide range of applications. However, their computational complexity makes it necessary to design efficient hardware accelerators. Most CNN accelerators focus on exploring dataflow styles that exploit computational parallelism. However, potential performance speedup from sparsity has not been adequately addressed. The computation and memory footprint of CNNs can be significantly reduced if sparsity is exploited in network evaluations. To take advantage of sparsity, some accelerator designs explore sparsity encoding and evaluation on CNN accelerators. However, sparsity encoding is just performed on activation or weight and only in inference. It has been shown that activation and weight also have high sparsity levels during training. Hence, sparsity-aware computation should also be considered in training. To further improve performance and energy efficiency, some accelerators evaluate CNNs with limited precision. However, this is limited to the inference since reduced precision sacrifices network accuracy if used in training. In addition, CNN evaluation is usually memory-intensive, especially in training. In this paper, we propose SPRING, a SParsity-aware Reduced-precision Monolithic 3D CNN accelerator for trainING and inference. SPRING supports both CNN training and inference. It uses a binary mask scheme to encode sparsities in activation and weight. It uses the stochastic rounding algorithm to train CNNs with reduced precision without accuracy loss. To alleviate the memory bottleneck in CNN evaluation, especially in training, SPRING uses an efficient monolithic 3D NVM interface to increase memory bandwidth. Compared to GTX 1080 Ti, SPRING achieves 15.6X, 4.2X and 66.0X improvements in performance, power reduction, and energy efficiency, respectively, for CNN training, and 15.5X, 4.5X and 69.1X improvements for inference.

Automatically solving math word problems is a critical task in the field of natural language processing. Recent models have reached their performance bottleneck and require more high-quality data for training. We propose a novel data augmentation method that reverses the mathematical logic of math word problems to produce new high-quality math problems and introduce new knowledge points that can benefit learning the mathematical reasoning logic. We apply the augmented data on two SOTA math word problem solving models and compare our results with a strong data augmentation baseline. Experimental results show the effectiveness of our approach.

Quick, early, and precise detection is important for diagnosis to control the spread of COVID-19 infection. Artificial Intelligence (AI) technology could certainly be used as a modulating tool to ease the detection, and help with the preventive steps further. Convolutional neural networks (CNNs) have achieved state-of-the-art performance in many visual recognition tasks. Nevertheless, most of these state-of-the-art networks highly rely on the availability of a high amount of labeled data, being an essential step in supervised machine learning tasks. Conventionally, this manual, mundane, and time-consuming process of annotating images is done by humans. Learning to localize or detect COVID-19 infection masks in our specific case study typically requires the collection of CT scan data that has been labeled with bounding boxes or similar annotations, which generally is limited. A technique that could perform such learning with much less annotations, and transfer the learned proposals that are algorithm-driven to generate more synthetic annotated samples would be helpful & quite valuable. We present such a technique inspired by weakly trained mask region based convolutional neural networks (R-CNN) architecture for localization, in which the number of images with their pixel-level masks can be a small proportion of the total dataset, and then further improvise CNNs by inversely generating dense annotations on-the-go using an algorithmic-based computational approach. We focus on alleviating the bottleneck associated with deep learning models needing annotated data for training in an intuitive reverse engineering fashion through this work. Our proposed solution can certainly provide the prospect of automated labeling on-the-fly, thereby reducing much of the manual work. As a result, one can quickly train a precise COVID-19 infection detector with the leverage of autonomous frame-by-frame machine generated annotations. The model achieved mean precision accuracy (%) of 0.99, 0.931, and 0.8 for train, validation, and test set, respectively. The results demonstrate that the proposed method can be adopted in a clinical setting for assisting radiologists, and also our fully autonomous approach can be generalized to any detection/recognition tasks at ease.

For hyperspectral image classification tasks, obtaining sufficient labeled samples is usually difficult, time-consuming and expensive. To address the aforementioned issue, by transferring the labeled sample information of a relevant source domain to the unlabeled target domain, a hyperspectral image classification method based on domain adversarial broad adaptation network (DABAN) is proposed. Firstly, the bottleneck adaptation module composed of a bottleneck layer and a domain adaptation layer is constructed and introduced to the domain adversarial neural network, thus the domain adversarial adaptation network (DAAN) is designed. By simultaneously performing domain adversarial learning, reducing both the marginal distribution difference and second-order statistic difference between two domains, the distributions of the source and target domains are aligned. Then the conditional distribution adaptation regularization term based on maximum mean discrepancy is embedded into a broad learning system to obtain the conditional adaptation broad network (CABN). On the one hand, CABN can perform the class-level distribution adaptation on the domain-invariant features extracted by DAAN. On the other hand, the representation ability of the domain-invariant features expanded by CABN can be further enhanced. Experimental results on ten real hyperspectral data pairs show that, compared with the existing mainstream methods, DABAN can effectively utilize relevant source domain information to assist in improving classification accuracy of target domain.

With the increased model size of convolutional neural networks (CNNs), overfitting has become the main bottleneck to further improve the performance of networks. Currently, the weighting regularization methods have been proposed to address the overfitting problem and they perform satisfactorily. Since these regularization methods cannot be used in all the networks and they are usually not flexible enough in different phases of the training and test processes, this article proposes a multiscale conditional (MSC) regularization method. MSC divides the intermediate features into different scales and then generates new data for each scale features, respectively. In addition, the new data are generated by employing the information from two conditions: 1) each sample feature and 2) each layer pattern. Finally, a self-identity structure is proposed to supplement the features with the generated data. Therefore, MSC can adaptively and efficiently generate much finer and individualized data to make the entire regularization more flexible. Furthermore, MSC is more general and can be applied to all kinds of networks through the proposed self-identity structure. The experimental results on all the benchmark datasets showed that the proposed MSC regularization method achieves the best performances in all the networks.

With the vast expansion of data loads across numerous safety and infotainment applications, the software-defined paradigm has become a critical component for 5G-enabled and intelligent vehicular networks. By transforming the network’s rigid hardware devices into programmable units, SDVN offers a broad view of the entire network status and a standard interface between heterogeneous wireless access technologies. However, deploying a centralized control unit carries several problems, such as bottleneck issues and densification issues. A distributed control plane is a potential alternative to a centralized control plane. Still, it poses many questions about the optimal location of local control units and the number of control units required for a given network structure. In this article, we propose an efficient proactive controller deployment and assignment strategy for distributed 5G-enabled software-defined Vehicular Networks based on predictive vehicles’ traffic flow information and communication delays between switch-enabled roadside units. A thorough performance evaluation is conducted under different vehicular network’s mobility scenarios. We assess the performance of the proposed method in terms of communication latency and load on the subsequent control units and their assigned set of switches. The proposed proactive clustering method is compared to different forms of controller placement methods. The results showed that the proposed approach reduced the network’s communication delays while maintaining a balanced distributed load among controllers over the simulation time.

Wireless interference identification (WII) is one of the most promising technologies in anti-interference communication systems and non-cooperative communication systems. Recently, deep learning (DL) based methods have achieved striking performance for WII. However, existing works have not researched hardware-friendly network quantization for WII, which is essential for deploying model on resource-constrained devices. In this paper, we optimize binarized neural networks (BNNs) for WII, which are the most extreme form of quantization by constraining weights and activations to be binary value. Specifically, to overcome the difficulty in propagating gradient during back-propagation due to the non-differentiable quantization function, we introduce a novel approximation for gradients calculation, which bridges the accuracy gap between the BNN and the full-precision counterpart. Additionally, to solve the bottleneck of serious performance degradation of BNNs, we propose two techniques to minimize quantization noise and create multiple routes to update the parameters of BNNs, resulting in further improvement on performance. Experiments demonstrate that the proposed BNN achieves accuracy near the full-precision counterpart with substantially reduced resource consumption for WII.

Modern Internet of Things (IoT)-assisted networks deploy many critical computing devices in the edge of the network so as to mitigate the bottleneck of network latency and bandwidth outage. Such computing devices have become attractive targets of attackers, as they derive critical insights that can often regulate the underlying industrial processes. Security administrators have to analyze vulnerable configurations, and evaluate feasible hardening options to secure such systems. In this article, we investigate the security threats to the edge devices in the IoT-assisted networks because of the inherent vulnerabilities present in the IoT devices, and formulate a novel graphical security model to conduct vulnerability-based risk assessment in such networks. We further propose a set of combinatorial optimization techniques for security hardening. For securing the target devices, we propose a randomized algorithm to enforce isolation of potential target devices from the attackers. The algorithm produces the least number of vulnerabilities, the removal of which protects the targets. For densely connected networks, we propose another algorithm to identify a minimal set of vulnerabilities to be hardened so as to keep potential targets below admissible levels of risk. The approach includes the formulation of an integer linear program, which is relaxed to solve in polynomial time, followed by a round-off technique to find the actual solution.

Communication has been considered as a major bottleneck in large-scale decentralized training systems since participating nodes iteratively exchange large amounts of intermediate data with their neighbors. Although compression techniques like sparsification can significantly reduce the communication overhead in each iteration, errors caused by compression will be accumulated, resulting in a severely degraded convergence rate. Recently, the error compensation method for sparsification has been proposed in centralized training to tolerate the accumulated compression errors. However, the analog technique and the corresponding theory about its convergence in decentralized training are still unknown. To fill in the gap, we design a method named ECSD-SGD that significantly accelerates decentralized training via error-compensated sparsification. The novelty lies in that we identify the component of the exchanging information in each iteration (i.e., the sparsified model update) and make targeted error compensation over the component. Our thorough theoretical analysis shows that ECSD-SGD supports arbitrary sparsification ratio and achieves the same convergence rate as the non-sparsified decentralized training methods. We also conduct extensive experiments on multiple deep learning models to validate our theoretical findings. Results show that ECSD-SGD outperforms all the start-of-the-art sparsified methods in terms of both the convergence speed and the final generalization accuracy.

As a powerful tool for data streams processing, the vast majority of existing evolving intelligent systems (EISs) learn prediction models from data in a supervised manner. However, high-quality labelled data can be difficult to obtain in many real-world classification applications concerning data streams, though unlabelled data is plentiful. To overcome the labelling bottleneck and construct a stronger classification model, a novel semi-supervised EIS is proposed in this paper. After being primed with a small amount of labelled data, the proposed method is capable of continuously self-developing its system structure and self-updating the meta-parameters from unlabelled data streams chunk-by-chunk in a non-iterative, exploratory manner by exploiting a novel pseudo-labelling strategy. Thanks to its transparent prototype-based structure and human-understandable reasoning process, the proposed method can provide users high explainability and interpretability while achieving great classification precision. Experimental investigation demonstrates the superior performance of the proposed method.

Unmanned aerial vehicles (UAVs) have found widespread commercial, civilian, and military applications. Wireless communication has always been one of the core technologies for UAV. However, the communication capacity is becoming a bottleneck for UAV to support more challenging application scenarios. The heavily-occupied sub-6 GHz frequency band is not sufficient to meet the ultra high-data-traffic requirements. The utilization of the millimeter-wave (mmWave) frequency bands is a promising direction for UAV communications, where large antenna arrays can be packed in a small area on the UAV to perform three-dimensional (3D) beamforming. On the other hand, UAVs serving as aerial access points or relays can significantly enhance the coverage and quality of service of the terrestrial mmWave cellular networks. In this paper, we provide a comprehensive survey on mmWave beamforming enabled UAV communications and networking. The technical potential of and challenges for mmWave-UAV communications are presented first. Then, we provide an overview on relevant mmWave antenna structures and channel modeling. Subsequently, the technologies and solutions for UAV-connected mmWave cellular networks and mmWave-UAV ad hoc networks are reviewed, respectively. Finally, we present open issues and promising directions for future research in mmWave beamforming enabled UAV communications and networking., Accepted by IEEE Communications Surveys and Tutorials

Visible light communication (VLC) is a promising communication technology that benefits from several advantages over other wireless communication techniques. However, the capacity of VLC systems is limited by the low modulation bandwidth of light-emitting diodes (LEDs) used as transmitters. To circumvent this bottleneck, non-orthogonal multiple access (NOMA) is popularly used in VLC systems. However, due to high inter-user interference at the receivers, NOMA suffers from poor outage performance, especially if the number of users is large. We propose to boost the outage performance of a NOMA-based VLC system by incorporating multi-user cooperative diversity over radio frequency (RF) links, thereby transforming it into a hybrid VLC-RF system. Such a system with cooperative NOMA (CoNOMA) has enhanced reliability compared to a conventional non-cooperative NOMA system. We formulate the expressions for the outage probability of the proposed system and analyze the outage performance for various power allocation schemes used in NOMA. We study the effects of increasing the number of users and the target data rate on the system’s overall outage performance. We also model the reliability of the multi-user CoNOMA system using Markov chains and evaluate the various reliability parameters for each user. We show that the outage performance and reliability of the VLC system are significantly enhanced by employing cooperative diversity using CoNOMA.