Data imbalance issue generally exists in most medical image analysis problems and maybe getting important with the popularization of data-hungry deep learning paradigms. We explore cutting-edge Wasserstein generative adversarial networks (WGANs) to address data problem oversampling on minority classes. The WGAN can estimate underlying distribution a class synthesize more plausible helpful samples for classification model. In this paper, WGAN-based over-sampling technique is applied augment...

The cloud–edge collaborative inference requires efficient scheduling of artificial intelligence (AI) tasks to the appropriate edge devices. Gls DNN latency has become a vital basis for improving efficiency. However, devices exhibit highly heterogeneous due differences in hardware architectures, computing power, etc. Meanwhile, diverse deep neural networks (DNNs) are continuing iterate over time. diversity and DNNs introduces high computational costs measurement methods, while invasive...

The development of deep learning has led to increasing demands for computation and memory, making multi-chiplet accelerators a powerful solution. Multi-chiplet require more precise consideration hardware configurations mapping schemes in terms computation, communication patterns compared monolithic designs, order avoid underutilization performance. However, there is currently lack performance testing methods specifically tailored accelerators. Existing primarily focus on correctness do not...

Mobile devices with limited resources are prevalent as they have a relatively low price. Providing good user experience has been big challenge. This paper found that foreground applications often unexpectedly interfered by background applications’ memory activities. Improving on resource-limited mobile calls for strong collaboration between and process management. proposes framework, Ice, to optimize the devices. With processes will cause frequent refaults in identified frozen accordingly....

Recently, algorithm-hardware co-exploration for neural networks (NNs) has become the key to obtaining high-quality solutions. However, previous efforts FPGAs focus on architecture search (NAS) while lacking hardware (HAS), thus limiting full potential of co-design. Although expanding scope HAS offers performance potential, exponentially increased joint space presents a formidable challenge. To address this, we propose deep and efficient framework, which jointly searches Networks Accelerators...

With the development of deep learning and artificial intelligence, more research apply neural networks to natural language processing tasks. However, while majority these take English corpus as dataset, few studies have been done using Chinese corpus. Meanwhile, Existing algorithms typically regard word or character basic unit but ignore deeper information into character. In linguistic, strokes are who similar letters word. Inspired by recent success at character-level, we delve stroke level...

With the coming concept of 'big data', ability to handle large datasets has become a critical consideration for success industrial organizations such as Google, Amazon, Yahoo! and Facebook. As an important Cloud Computing framework bulk data processing, Hadoop is widely used in these organizations. However, performance MapReduce seriously limited by its stiff configuration strategy. Even single simple job Hadoop, number tuning parameters have be set users. This may easily lead loss due some...

Mobile devices with limited resources are prevalent as they have a relatively low price. Providing good user experience has been big challenge. This paper found that foreground applications often unexpectedly interfered by background applications' memory activities. Improving on resource-limited mobile calls for strong collaboration between and process management. proposes framework, Ice, to optimize the devices. With processes will cause frequent refaults in identified frozen accordingly....

본 논문에서는 하이브리드 궤도회로를 이용한 정위치 정차시스템에 대한 가능성에 대해 연구하였다. 궤도회로는 UHF대역의 RFID시스템을 이용하고 실시간 운영체제를 기반으로 철도차상시스템과의 통신이 가능한 차세대 철도신호시스템의 한 부분이다. 논문에서 사용된 UHF대역 RFID 시스템은 원형편파의 태그와 리더 안테나가 사용되고 일체형 리더기는 송신출력 25dBm에서 우수한 인식회수를 얻었다. 현재 수동운행중인 지하철 차량에 이 시스템을 적용하여 정차를 하게 되면 기존 수동운전에서 기관사의 실수로 역을 지나거나 PSD(Platform Screen Door)와 열차출입문이 Mismatch 되는 현상을 방지할 수 있다. We have studied on the possibility of precise positioning using hybrid Track Circuit system. Hybrid uses which replaces UHF. is a part next generation...

App caching is important for mobile devices, which enables fast switching and state restoration of apps by all the pages in memory. Memory swapping can improve app capability evicting to secondary storage. However, enabling memory could induce jitters interactions, significantly degrades user experience. As a result, storage-based disabled default most devices. This article proposes novel swap framework, SEAL, experience-aware two-level swapping, maximizes benefits minimizes negative impact...

The Smith-Waterman algorithm, which produces the optimal local alignment between pairwise sequences, is universally used as a key component in bioinformatics fields. It more sensitive than heuristic approaches, but also time-consuming. To speed up Single-Instruction Multiple-Data (SIMD) instructions have been to parallelize algorithm by leveraging data parallel strategy. However, SIMD-based (SW) algorithms show limited scalability. Moreover, recent next-generation sequencing machines...

With the development of deep learning and artificial intelligence, neural networks are increasingly being applied for natural language processing tasks. However, majority research on focuses alphabetic languages. Few studies have paid attention to characteristics ideographic languages, such as Chinese language. In addition, existing algorithms typically regard words or characters basic units while ignoring information contained within deeper architecture characters. language, each character...

With the development of NAND flash technology, hybrid SSDs with high-density and low-cost memory have become mainstream existing SSD architecture. In this architecture, two modes can be dynamically switched, such as single-level cell (SLC) mode quad-level (QLC) mode. Based on evaluations analysis multiple real devices, paper presents interesting findings. They demonstrate that coordination between flash-modes is not well-designed in architectures. This proposes HyFlex, which redesigns...

Sequence alignment algorithms are a basic and critical component of many bioinformatics fields. With rapid development sequencing technology, the fast growing reference database volumes longer length query sequence become new challenges for alignment. However, have prohibitively high time space complexity. In this paper, we present DSA, scalable distributed system that employs Apache Spark to process sequences data in horizontally environment, leverages parallel strategy based on Single...

Handwritten Chinese character recognition (HCCR) is an important research field of pattern recognition, which has attracted extensive studies during the past decades. Recently convolutional neural network (CNN) based methods have achieved state-of-the-art performance for handwritten recognition. Nevertheless, still limited to be effectively used in actual environment due large-scale vocabulary and great diversity handwriting style. In this paper, we constructed a service on network, tries...

철도신호시스템에서 열차위치 검지기능은 선로의 레일을 전기회로의 일부분으로 사용하여 차륜에 의해 단락되어 열차의 유무를 검지하는 궤도회로, RFID와 차륜센서, GPS 등과 같은 여러 가지 방식으로 구현 연구가 되고 있다. 하이브리드 궤도회로는 안테나와 리더기를 차량에 설치하고, 태그를 침목위에 설치하여, 안테나에서 태그에 저장된 절대위치정보를 제어장치에 전송하여 열차위치를 인식하는 RFID 방식의 궤도회로이다. 열차위치검지기능에서 태그의 인식율은 열차운행의 안전에 직접적인 영향을 주게 되므로 고신뢰도를 요구한다. 본 논문에서는 방향각을 갖는 이용한 태그인식율의 향상에 관한 연구를 하였다. Track circuit is a simple electrical device which lies in the connection of two rails by wheels and axle locomotives rolling stock to short out an circuit, used...

IMS (IP Multimedia Subsystem) is an emerging architectural framework that delivers a number of multimedia services -- ranging from voice/video over LTE, interactive gaming and many more in operational LTE network. Network operators are embracing cloud-based to meet increasing traffic demand. They can easily cost-efficiently implement applications while ensuring superior end-user experiences through always-on real-time engagement. In this paper, we reveal cannot provide session-level...

Because of its high resolution, low cost, small volume, power dissipation and less conversion time consumption, the direct digital synthesizer (DDS) method has been applied more in fields frequency synthesis signal generation. However, only a limited number precise signals can be synthesized by traditional DDS, for reason that accumulator modulus is fixed, tuning word must integer. In this paper, DDS using compound proposed, which improves accuracy at any points on premise guaranteeing...

With the development of cloud computing and internet, e-Commerce, e-Business corporate world revenue are increasing with high rate. These areas require scalable consistent databases. NoSQL databases such as HBase has been proven to scalability well performance on platforms. However, inevitable special data few increment frequent access leads hotspot unbalanced accessing distribution between storage servers. Due their properties, these often cannot be stored in multiple tables. Some nodes...