Search

Read alignment is a time-consuming step in genome sequencing analysis. The most widely used software for read alignment, BWA-MEM, and the recently published faster version BWA-MEM2 are based on the seed-and-extend paradigm for read alignment. The seeding step of read alignment is a major bottleneck contributing ~40% to the overall execution time of BWA-MEM2 when aligning whole human genome reads from the Platinum Genomes dataset. This is because both BWA-MEM and BWA-MEM2 use a compressed index structure called the FMD-Index, which results in high bandwidth requirements, primarily due to its character-by-character processing of reads. For instance, to seed each read (101 DNA base-pairs stored in 37.8 bytes), the FMD-Index solution in BWA-MEM2 requires ~68.5 KB of index data. We propose a novel indexing data structure named Enumerated Radix Tree (ERT) and design a custom seeding accelerator based on it. ERT improves bandwidth efficiency of BWA-MEM2 by 4.5X while guaranteeing 100% identical output to the original software, and still fitting in 64 GB DRAM. Overall, the proposed seeding accelerator implemented on AWS F1 FPGA (f1.4xlarge) improves seeding throughput of BWA-MEM2 by 3.3X. When combined with seed-extension accelerators, we observe a 2.1X improvement in overall read alignment throughput over BWA-MEM2. The software implementation of ERT is integrated into BWA-MEM2 (ert branch: https://github.com/bwa-mem2/bwa-mem2/tree/ert) and is open sourced for the benefit of the research community.

Mobile edge computing (MEC) has been an effective paradigm to support real-time computation-intensive vehicular applications. However, due to highly dynamic vehicular topology, these existing centralized-based or distributed-based scheduling algorithms requiring high communication overhead, are not suitable for task offloading in vehicular networks. Therefore, we investigate a novel service scenario of MEC-based vehicular crowdsourcing, where each MEC server is an independent agent and responsible for making scheduling of processing traffic data sensed by crowdsourcing vehicles. On this basis, we formulate a data-driven task offloading problem by jointly optimizing offloading decision and bandwidth/computation resource allocation, and renting cost of heterogeneous servers, such as powerful vehicles, MEC servers and cloud, which is a mixed-integer programming problem and NP-hard. To reduce high time-complexity, we propose the solution in two stages. First, we design an asynchronous deep Q-learning to determine offloading decision, which achieves fast convergence by training the local DQN model at each agent in parallel and uploading for global model update asynchronously. Second, we decompose the remaining resource allocation problem into several independent subproblems and derive optimal analytic formula based on convex theory. Lastly, we build a simulation model and conduct comprehensive simulation, which demonstrates the superiority of the proposed algorithm.

Convolutional neural networks (CNNs) have achieved significant accuracy improvement in many intelligent applications at the cost of intensive convolution operations and massive data movements. To efficiently deploy CNNs on low power embedded platforms in real time, the depthwise separable convolution has been proposed to replace the standard convolution, especially in lightweight CNNs, which remarkably reduces the computation complexity and model size. However, it is difficult for a general convolution engine to obtain the theoretical performance improvement as the decreased data dependency of depthwise convolution significantly reduces the data reuse opportunity. To address this issue, a flexible and highperformance accelerator based on FPGA is proposed to efficiently process the inference of both large-scale and lightweight CNNs. Firstly, by sharing the activation dataflow between the depthwise convolution and pooling layers, the control logic and data bus of the two layers are reused to maximize the data utilization and minimize the logic overhead. Furthermore, these two layers can be processed either directly after standard convolutions to eliminate the external memory accesses or independently to gain better flexibility. Thirdly, a performance model is proposed to automatically explore the optimal design options of the accelerator. The proposed hardware accelerator is evaluated on Intel Arria 10 SoC FPGA and demonstrates state-of-the-art performance on both large-scale CNNs, e.g., VGG, and lightweight ones, e.g., MobileNet.

Nowadays, many novel blockchain-based architecture and frameworks are proposed to solve issues in computer science and financial service. Smart contracts with blockchain systems, especially consortium blockchain systems, can help to provide many reliable and efficient functions for existing systems like smart grid payments. The novel concept of smart contract as a service is proposed but the difficulty of developing smart contracts on various kinds of blockchain systems are also significantly increasing which brings the additional cost for both developers and infrastructure builders. In this paper, we present an updated cloud-based smart integrated smart contract development system, ChainIDE 2.0, for the ultra-efficient development of blockchain-based smart contracts on multiple kinds of blockchain systems. Not only we stay as the most popular cloud-based developing Integrated Development Environment (IDE) for the Libra blockchain, but also we introduce the consortium blockchain systems such as Ant Financial Open-Chain (Ant OC) and served as the first cloud-based IDE supporting the Ant Financial OpenChain test net. Today, we have served almost 1 million compiled smart contracts which makes us the most popular cloud-based blockchain development IDE in the world.

We present the first ASIC accelerator for pair-Hidden-Markov-Model (Pair-HMM) in DNA variant calling, which conventionally requires ∼250T FLOPs per sequenced human genome. Using a hardware-algorithm co-design, we opportunistically replace floating point (FP) multiplication with 20b log-domain addition while employing bound checks to maintain (provable) correct results in downstream processing. FP computation is reduced by 43× on real human genome data. Implemented in 40nm CMOS, the 5.67 mm2 accelerator demonstrates 17.3G cell updates per second (CUPS) throughput, marking a 6.6× improvement over our baseline ASIC implementation and 355× GCUPS/mm2 improvement over a recent FPGA implementation [2].

Event-based cameras are bio-inspired sensors capturing asynchronous per-pixel brightness changes (events), which have the advantages of high temporal resolution and low power consumption compared with traditional frame-based cameras. Despite recent progress in human action recognition on traditional cameras, few solutions on event cameras have been proposed due to its unconventional frameless output. To perform low-power human action recognition on event camera, we propose a multipath deep neural network for action recognition based on event camera outputs. Specifically, a fixed number of asynchronous events are accumulated to form frames for feature extraction. With events encoding dynamic information, we estimate human pose from event frames to encode static information to improve recognition accuracy. The complementary properties between dynamic event frames and static human poses are jointly explored and fused to predict action. Extensive experiments verify the effectiveness of the proposed model with a recognition accuracy of 85.91% on the DHP19 dataset.

We present an accelerator for seed-extension, a critical and computational intensive step in genome sequencing. The accelerator consists of a triangular array of 25×25 custom design processing elements implementing a string-independent automata. It achieves 2.46M reads/s, a ~1800x performance improvement, and 27x smaller silicon footprint compared to a Xeon E5420.

Blockchain has become novel solutions for many traditional issues in computer science and finance. Recently, with the release of Libra blockchain from Facebook, the decentralized finance concepts have another huge development. However, there are already many different blockchain systems existing now and the development of blockchain systems have become more and more complicated since each kind of blockchain development environment will consume time to be built. To make the programming on different blockchain systems more easily, we propose a novel cloud-based solution, namely ChainIDE, for the development of blockchain-based smart contracts on multiple kinds of blockchain systems. With chainIDE, cross-chain developing of smart contracts on different blockchain systems can be easily done without any time consumed by building the environment. Based on the operation statistics in this paper, we served more than 310,000 compilations in the past 30 days which makes us the most popular cloud-based cross-chain development platform in the world.

Mobile edge computing (MEC) has been an emerging paradigm to support low-latency applications in vehicular networks by offloading resources at network edge. However, it is still challenging to apply MEC- based architecture to implement multimedia services due to varying wireless communication, high vehicle mobility and heterogeneous resource integration. In this paper, we investigate adaptive-bitrate (ABR)-based multimedia services (MS) in MEC-based vehicular networks, where each multimedia file is divided into multiple chunks and can be requested at different bitrate levels. Further, MEC servers can satisfy local vehicular requests by integrating heterogeneous cache and communication resources. Based on the above observation, we formulate joint resource optimization (JSO) problem by synthesizing cache placement, wireless bandwidth allocation and chunk quality adaptation. On this basis, we propose a reinforcement- learning-based cache placement (RLCP) algorithm, which determines the optimal offloaded chunks by learning the global knowledge of cache reward in an iterative way. Further, we design an adaptive-quality- based chunk selection (AQCS) algorithm, which can be adaptive to time-varying wireless channel by dynamically adjusting bandwidth allocation and quality level based on real-time service workload. Lastly, we build the simulation model and conduct an extensive performance evaluation, which demonstrates the superiority of proposed algorithms.

Developing effective control strategy to achieve flexible operation of the solvent-based post combustion CO 2 capture (PCC) process is the guarantee for the application of the technology. However, control of the PCC process is challenging due to its complex behavior such as strong couplings and disturbances. This paper proposes an advanced control approach for the solvent-based post combustion CO 2 capture system. On the basis of the conventional PID control structure, decoupling technique is used to overcome the coupling issues among multi-variables and a feedforward controller is used to improve the response speed of the PCC system to the flue gas flow rate change. Moreover, the parameters of the PID controller is tuned through optimization using genetic algorithm. Simulation studies show the advantages of the proposed control strategy.

A large amount of datasets are required for the off-line test of unmanned vehicles. Since the redundancy of different image datasets exists, we apply image style transformation to generate new images, which simplifies the data collection process. In this paper, an image-to-image transformation network is proposed based on the structural information. The proposed network consists of two branches, one for style transfer and the other for semantic segmentation. We use the structural information extracted by semantic segmentation to constrain the transformed image, so that the transformed image can preserve the integrity and structural consistency. We conduct the experiments and comparisons with the baseline methods. The experimental results prove that the proposed method can generate more visually pleasant images for the tasks of day-night transformation, sunny-foggy transformation, and summer-winter transformation.

Grounding electrode line impedance supervision (ELIS) method is used to detect and deal with faults of electrode line in UHVDC project. The reliability of ELIS depends on the difference between the normal impedance and fault impedance. Presently, there is no theoretical analysis about the fault impedance of electrode line, which makes it difficult to evaluate the adaptability of ELIS. This paper studies the fault impedance characteristic under different fault conditions by means of compound mode network. The influence of fault resistance and fault distance on fault impedance is also discussed. Then the adaptability analysis of ELIS is carreid out. The analysis results indicate that the ELIS has dead zone in some special fault points.

The aim of crowd counting is to estimate the number of people in images by leveraging the annotation of center positions for pedestrians' heads. Promising progresses have been made with the prevalence of deep Convolutional Neural Networks. Existing methods widely employ the Euclidean distance (i.e., $L_2$ loss) to optimize the model, which, however, has two main drawbacks: (1) the loss has difficulty in learning the spatial awareness (i.e., the position of head) since it struggles to retain the high-frequency variation in the density map, and (2) the loss is highly sensitive to various noises in crowd counting, such as the zero-mean noise, head size changes, and occlusions. Although the Maximum Excess over SubArrays (MESA) loss has been previously proposed to address the above issues by finding the rectangular subregion whose predicted density map has the maximum difference from the ground truth, it cannot be solved by gradient descent, thus can hardly be integrated into the deep learning framework. In this paper, we present a novel architecture called SPatial Awareness Network (SPANet) to incorporate spatial context for crowd counting. The Maximum Excess over Pixels (MEP) loss is proposed to achieve this by finding the pixel-level subregion with high discrepancy to the ground truth. To this end, we devise a weakly supervised learning scheme to generate such region with a multi-branch architecture. The proposed framework can be integrated into existing deep crowd counting methods and is end-to-end trainable. Extensive experiments on four challenging benchmarks show that our method can significantly improve the performance of baselines. More remarkably, our approach outperforms the state-of-the-art methods on all benchmark datasets., ICCV 2019 Oral

In this paper, we propose a framework for 3D traffic scenes construction based on semantic segmentation using convolutional neural networks (CNNs). Firstly, the segmentation network, whose architecture is constructed with encoder-decoder model and a hall module, divides traffic scenes into different parts: Road, sky, vehicle and other regions. Furthermore, we generate spatio-temporal graph models and construct 3D traffic scenes according to semantic segmentation results. The applications for scene simulation are then developed. The experimental results on the KITTI dataset demonstrate the effectiveness of the proposed framework.

In this paper, we propose a fast two-cycle level set algorithm of contour tracking for foreground objects based on superpixel segmentation. Contour tracking method includes three stages: superpixel segmentation, interactive feature extraction and foreground tracking. In the first stage, we preprocess the image by the method of superpixel segmentation. In the second phase, the work of defining and building the feature pool is based on the complexity of the image. Finally, on the basis of the first two stages, the fast two-cycle level set method can be used to track the contour of the target. The experimental results show the effectiveness of the proposed method.

In the process of constant angular speed grinding and constant linear speed grinding for camshaft workpiece, there are problems in which the speed, acceleration and jerk of each linkage servo system change drastically. This paper presents a polynomial motion law curve of tectonic spindle (rotary axis), designs a single-objective evaluation function with the rigidity requirement of the servo system of X-C linkage axis, and uses the particle swarm optimization (PSO) to solve the optimal solution of the objective function. The results show that the mechanical impact caused by excessive acceleration during the grinding process is greatly reduced, and the quality of high efficiency precision grinding of camshaft is greatly improved.

In order to effectively classify the increasingly massive load data, the method of applying support vector machine to load classification is studied. By comparing and selecting the C-SVC classification and the RBF kernel function, the optimization method of the grid search is used to find the optimal parameter combination of the LIBSVM classifier and establish the model. The visualization accuracy and index function are used to observe the classification accuracy of each class. Operations such as weighting the unbalanced samples are conducted to adjust the accuracy gap between large and small samples. Experimental results show that the method is feasible in load classification.

In the system architecture design phase, there may be flaws introduced by the access control design. The flaws make the system design vulnerable. Vulnerabilities that are not detected during the design phase will be taken to the subsequent phases of the system development. This reduces system security and leads to more repair cost. In order to detect the access control flaws in the design phase as early as possible, we propose the security analysis method using Architecture Analysis and Design Language (AADL). This method analyzes the access control policies between components. It can analyze and detect the access control flaws in the AADL architecture model. For describing the access control policies, vulnerability model annex is created. We use a case study, based on a module from an aircraft intelligent information system, to introduce the application of modeling and analysis methods.

Boiler combustion optimization is an effective method to improve the boiler efficiency and re fired units. This paper proposes a closed-loop combustion optimization control method based on the dyn by least-squares support vector machine trained by unscented kalman filter(UKF-LSSVM). The kernel parameters ($\alpha_{\backslash}b$) are continuously updated through unscented kalman filter, and support vectors are $\iota$ update strategy. Simulation results show that the proposed adaptive model with the UKF-LSSVM met time-varying characteristics of a boiler combustion system. Based on the dynamic model of the combusti the predictive control and the nonlinear optimization algorithm to realize the closed-loop adjustment of To validate the proposed approach, a series of field tests were conducted on a 1000MW ultra-superc. - Results show that the method can significantly reduce NOx emission and the fluctuation of reheat st maintaining the boiler efficiency.