k-Means, a simple but effective clustering algorithm, is widely used in data mining, machine learning and computer vision community. k-Means algorithm consists of initialization of duster centers and iteration. The in...
详细信息
ISBN:
(纸本)9781479983346
k-Means, a simple but effective clustering algorithm, is widely used in data mining, machine learning and computer vision community. k-Means algorithm consists of initialization of duster centers and iteration. The initial duster centers have a great impact on duster result and algorithm efficiency. More appropriate initial centers of k-Means can get closer to the optimum solution, and even much quicker convergence. In this paper, we propose a novel clustering algorithm, Kmms, which is the abbreviation of k-Means and Mean Shift. It is a density based algorithm. Experiments show our algorithm not only costs less initialization time compared with other density based algorithms, but also achieves better clustering quality and higher efficiency. And compared with the popular k-Means++ algorithm, our method gets comparable accuracy, mostly even better. Furthermore, we parallelize Kmms algorithm based on OPenMP from both initialization and iteration step and prove the convergence of the algorithm.
Although the total variation based gradient descent (TV-GD) algorithm has revealed a good performance for photoacoustic imaging (PAl), fast or real-time imaging remains a challenge. In this paper, the data dependencie...
详细信息
ISBN:
(纸本)9781479923465
Although the total variation based gradient descent (TV-GD) algorithm has revealed a good performance for photoacoustic imaging (PAl), fast or real-time imaging remains a challenge. In this paper, the data dependencies that exist in the TV-GD algorithm were exploited, and a general expression was then for the first time derived to unroll the inner loop that occupied the majority of the entire running time of the algorithm. All the terms consisting of the measurement matrices or the under-sampled data sets were then extracted and preprocessed rather than being calculated along with reconstruction. For implementation, we accessed the JACKET toolbox to parallelize the execution of the matrix-vector multiplications and the vector additions generated by the general expression itself. The under-sampled dataset with 30, 60, 90 and 120 projections were adopted to reconstruct a 128x128 Shepp-Logan Phantom. The simulation results revealed a minimum reconstruction time of 0.64s in the case of the 60-view data, and a maximum speedup of 69X from the 120-view dataset.
In digitalization era, credit card fraud detection is of high significance to financial organizations. This paper discussed about credit card fraud detection by parallelizing of Negative Selection Algorithm on the Clo...
详细信息
ISBN:
(纸本)9781467364904;9781467364898
In digitalization era, credit card fraud detection is of high significance to financial organizations. This paper discussed about credit card fraud detection by parallelizing of Negative Selection Algorithm on the Cloud computing platform. We present performance evaluation of running the algorithm on the cloud by MapReduce framework and show it's dramatically results on real world financial data. We argue that, for the fraud detection rate, False Negative rate, fraud catching rate (True Positive rate) and false alarm rate (False Positive rate), Cost and Hit rate that are the best metrics for a desirable credit card fraud detection system.
SECloud is a automatic platform to deal with the resource-intensive and laborintensive nature of high-quality software analysis. SECloud parallelizes symbolic execution in computing cloud to cope with path explosion. ...
详细信息
ISBN:
(纸本)9781479932610
SECloud is a automatic platform to deal with the resource-intensive and laborintensive nature of high-quality software analysis. SECloud parallelizes symbolic execution in computing cloud to cope with path explosion. To our knowledge, SECloud is the first binary analyzing software that scales to large clusters of machines and can automatically test real-world softwares (e.g., Squirrel, Aeon, Socat, Aspell, Atphttpd) effectively. It utilizes the technology of loop elision and state merging to reduce the executing path state explosion. SECloud can also diagnose incomplete software patches by analyzing the difference pathes caused by patches. SECloud offers a flexible testing service according to the software testing task. It runs on computing clouds, like Amazon EC2, and takes advantage of the the flexible computing resource of cloud. Our experiment results show that SECloud can achive 3 to 4 orders of magnitude speedup comparing a state-of-the-art symbolic execution engine (e.g., S2E).
Simulating turbulent liquids with breaking waves and splashes is among the most desired features in fluid animation. Lagrangian methods such as Smoothed Particle Hydrodynamics method (SPH) are a promising way to captu...
详细信息
Simulating turbulent liquids with breaking waves and splashes is among the most desired features in fluid animation. Lagrangian methods such as Smoothed Particle Hydrodynamics method (SPH) are a promising way to capture such properties. However, the Particle-based liquid surface simulation has not been applied very well since its consumption is way too large. This paper derives the governing equations in SPH approaches and parallelizes the dynamics-based surface simulation with the MapReduce program models which apply the SPH approach in Cloud Computing. Compared to the serial methods, this approach obtained a 3.11 times speedup on the experimental platform.
A digital signal processor (DSP), which is a special-purpose microprocessor, is designed to achieve higher performance on DSP applications. Because most DSP applications contain many nested loops and permit a very hig...
详细信息
A digital signal processor (DSP), which is a special-purpose microprocessor, is designed to achieve higher performance on DSP applications. Because most DSP applications contain many nested loops and permit a very high degree of parallelism, the DSP multiprocessor has a suitable architecture to execute these applications. Unfortunately, conventional scheduling methods used on DSP multiprocessors allocate only one operation to each DSP every time unit, even if the DSP includes several function units that can operate in parallel. Obviously they cannot achieve full function unit utilization. Hence, in this paper, we propose a two-level scheduling method (TSM) to overcome this common failing. TSM contains two approaches, which integrates unimodular transformations, loop tiling technique, and conventional methods used on single DSP, Besides introducing algorithm, we also use an analytic module to analyze its preliminary performance. Based on our analyses the TSM can achieve shorter execution time and more scalable speedup results. In addition, the TSM causes less memory access and synchronization overheads, which are usually negligible in the DSP multiprocessor architecture. (C) 2004 Elsevier Inc. All rights reserved.
暂无评论