For distributed training, the communication overhead for parameter synchronization is heavy in the network. Data aggregation can efficiently alleviate network overheads. However, existing works on data aggregation are...
详细信息
Controlled thermonuclear fusion has always been a dream pursued by mankind. However, the physical processes of controlled thermonuclear fusion are complex, requiring numerical simulations with high performance computi...
Controlled thermonuclear fusion has always been a dream pursued by mankind. However, the physical processes of controlled thermonuclear fusion are complex, requiring numerical simulations with high performance computing, and the amount of data generated by the physical processes on spatial, temporal and temperature scales is too large to be captured, managed, processed and collated in a reasonable time frame by mainstream software tools to achieve more aggressive fusion physical design. The data are too large to be captured, managed, processed, and collated into more aggressive targets for fusion physical design in a reasonable time by mainstream software tools. At the same time, the failure of fusion ignition can be caused by the distortion of various key physical quantities, and only by decomposing the process step by step and clarifying the changes of key physical quantities in the fusion physics process, can an effective mechanism be formed to prevent the distortion of key physical quantities from causing ignition failure in experimental physics. Big data in collaboration with artificial intelligence and high performance computing to drive the physical design of fusion is a novel avenue. By data acquisition with and pre-processing, this allows the creation of small sample libraries and deep learning. With supervised learning function convergence, incorporating solid/fluid computational methods, further network layering and cell expansion, the parameters of the physical model conforming to Lawson’s criterion will become experimental physical parameters, and we find the new approach in Data capacity, Model combination approach, Type of material, Calculation speed, Optimisation of design iteration times, etc. are superior to the traditional approach.
As the application scenarios of convolutional neural network (CNN) become more and more complex, the general CNN accelerator based on matrix multiplication has become a new research focus. The existing mapping methods...
详细信息
Stream clustering is an important data mining technique to capture the evolving patterns in real-time data streams. Today's data streams, e.g., IoT events and Web clicks, are usually high-speed and contain dynamic...
详细信息
ISBN:
(纸本)9781728170022
Stream clustering is an important data mining technique to capture the evolving patterns in real-time data streams. Today's data streams, e.g., IoT events and Web clicks, are usually high-speed and contain dynamically-changing patterns. Existing stream clustering algorithms usually follow an online-offline paradigm with a one-record-at-a-time update model, which was designed for running in a single machine. These stream clustering algorithms, with this sequential update model, cannot be efficiently parallelized and fail to deliver the required high throughput for stream clustering. In this paper, we present DistStream, a distributed framework that can effectively scale out online-offline stream clustering algorithms. To parallelize these algorithms for high throughput, we develop a mini-batch update model with efficient parallelization approaches. To maintain high clustering quality, DistStream's mini-batch update model preserves the update order in all the computation steps during parallel execution, which can reflect the recent changes for dynamically-changing streaming data. We implement DistStream atop Spark Streaming, as well as four representative stream clustering algorithms based on DistStream. Our evaluation on three real-world datasets shows that DistStream-based stream clustering algorithms can achieve sublinear throughput gain and comparable (99%) clustering quality with their single-machine counterparts.
Reputation has been widely used in the energy field in recent years. However, their reputation mechanisms are usually centralized even if they are designed for distributed energy systems, which could cause vulnerabili...
详细信息
ISBN:
(纸本)9781665407359
Reputation has been widely used in the energy field in recent years. However, their reputation mechanisms are usually centralized even if they are designed for distributed energy systems, which could cause vulnerability to single point failures. This paper explores the design of blockchain-based distributed reputation for a cap-and-trade carbon emission system. The blockchain technology is adopted to achieve distributed management of reputation scores and realize a peer-to-peer carbon trading market. Simulation experiments are carried out to demonstrate the influence of the proposed reputation rules on reputation scores. In addition, a case study shows how reputation affects the results of carbon trading. As far as we know, this paper is one of the few works that incorporate distributed reputation in a carbon emission system.
The proceedings contain 91 papers. The topics discussed include: a personal distributed real-time collaborative system;a customized reinforcement learning based binary offloading in edge cloud;efficient post-quantum i...
ISBN:
(纸本)9781728190747
The proceedings contain 91 papers. The topics discussed include: a personal distributed real-time collaborative system;a customized reinforcement learning based binary offloading in edge cloud;efficient post-quantum identity-based encryption with equality test;D2D-enabled reliable data collection for mobile crowd sensing;deep spatio-temporal attention model for grain storage temperature forecasting;secure and verifiable data access control scheme with policy update and computation outsourcing for edge computing;proactive content caching for internet-of-vehicles based on peer-to-peer federated learning;a similarity clustering-based deduplication strategy in cloud storage systems;and DyRAC: cost-aware resource assignment and provider selection for dynamic cloud workloads.
In recent years, virtual reality, big data and cloud computing technologies have all made rapid development, and they are also more and more closely combined with the apparel industry. The apparel digital design and v...
详细信息
In recent years, virtual reality, big data and cloud computing technologies have all made rapid development, and they are also more and more closely combined with the apparel industry. The apparel digital design and virtual fitting techniques are playing increasingly important roles in the field of apparel design, manufacturing and sales. The three-dimensional digital design and virtual fitting make the designers or the users have multi-view and more real impressions and experiences of the apparel. But 3D display is difficult to be popularized in apparel industry because of its high cost. The cost of two-dimensional display is low and easy to popularize. However, 2D system can only provide images of a specific perspective, lacking a multi-perspective experience. If the supplier is required to provide 2D images of several specific perspectives, the personalized needs of users cannot be satisfied. Therefore, how to make 2D system according to an object image and users' needs to provide the object images of other perspectives is an urgent problem to be solved. In this work, we propose a novel view synthesis pipeline based on an enhanced Pix2Pix neural network for a novel view (Pix2Pix-V). It consists of four parts: 1) a new method of generating the labeled virtual collar image data based on its 3D model, and the 2D images can be obtained on any perspective; 2) Pix2Pix-V neural network; 3) a Pix2Pix-V training workflow for predicting the novel view; 4) transfer learning is used in predicting similar views of virtual collars.
Knitwear is the second largest product category in the traditional clothing industry. The traditional clothing product supply chain is difficult to meet consumer's demand of fast fashion. AI and industrial Interne...
详细信息
Knitwear is the second largest product category in the traditional clothing industry. The traditional clothing product supply chain is difficult to meet consumer's demand of fast fashion. AI and industrial Internet technology, together with intelligent knitting equipments make C2M flexible supply chain possible through digitization. distributed cloud factory is built as regional node of production supported by digital twin and parallel intelligence. Personalized needs of end consumers can be quickly responded, and the cost is close to traditional large-scale assembly line. parallel manufacturing applies industrial robots, Internet of things, cloud computing, AR and AI technologies comprehensively to build virtual factory in Cloud that carries out decision-making and management simultaneously, with highly automated production equipments in workshop. Meanwhile, it combines with smart knitting factories to build a global distributed collaborative production management platform based on parallel manufacturing architecture system through virtual reality interaction, distributed manufacturing, managed and controlled through Cloud. Intelligent upgrading of traditional knitting industry can effectively improve labor productivity, reduce production cost and number of employees.
parallel programming is one of the most effective approaches to handle complex problems regarding time complexity by reducing computation time, by getting the most of the capacity of the processors and shared-memory o...
详细信息
The fast development of acceleration architectures and applications has made heterogeneous computing the norm for high-performance computing. The cost of high volume data movement to the accelerators is an important b...
详细信息
ISBN:
(纸本)9783030576752;9783030576745
The fast development of acceleration architectures and applications has made heterogeneous computing the norm for high-performance computing. The cost of high volume data movement to the accelerators is an important bottleneck both in terms of application performance and developer productivity. Memory management is still a manual task performed tediously by expert programmers In this paper, we develop a compiler analysis to automate memory management for heterogeneous computing. We propose an optimization framework that casts the problem of detection and removal of redundant data movements into a partial redundancy elimination (PRE) problem and applies the lazy code motion technique to optimize these data movements. We chose OpenMP as the underlying parallel programming model and implemented our optimization framework in the LLVM toolchain. We evaluated it with ten benchmarks and obtained a geometric speedup of 2.3x, and reduced on average 50% of the total bytes transferred between the host and GPU.
暂无评论