并行算法

Improvement on the parallel algorithm of association rule

首先简单介绍了相关规则及其并行开采算法的一些基本情况，然后指出了现有算法在分布式异构数据库中不能有效利用计算资源和造成信息丢失的问题。

The simulation results show that it is an effective genetic algorithm with advantages over common ones, and has better parallelism.

仿真结果表明，此算法是有效的，优于普通的遗传算法，具有较高的并行性。

With the result of real image data experiments, it can be pointed out that it is a simple, less memory, parallelizable and real time algorithm.

使用此算法对两组真实图像序列中运动小目标进行检测，根据对检测结果的分析，指出此算法可有效地完成检测任务，并且具有运算简单、存储量小、可并行实现、即时性好等特点。

In this paper, an advanced geneticalgorithm, which is based on the inter-task partial ordering relation and task depth, ispresented, and the implementing method of concurrent design sub-tasks scheduling based onthis algorithm is represented.

针对任务调度问题，本文提出了一种基于任务偏序图和子任务深度值的改进的遗传算法，实现了基于该算法的并行设计任务的优化调度。

The 3-dimentional intricate MMIC model with encapsulation and loss effects and the MLFMA numerical recipes combined with pre-conditioning and parallel technology are mainly studied.

主要研究三维模型下、考虑了封装和损耗效应的复杂微波毫米波集成电路的MLFMA分析模型；结合先进的条件数预处理技术的MLFMA算法；结合并行计算技术和MLFMA算法。

This paper has presented an optimization algorithm to fulfill the search. Compared with the original exhaustive aerial search, this method is based on direction preelection mechanism and coarse-granularity master-slave parallel memory model.

本文对原来的穷举串行搜索算法作改进，提出一种基于方向预选和粗粒度－主从记忆并行模型的优化算法。

As compared with the conventional motion estimation methods, the proposed algorithm based on the multi-level simultaneous wavelet transform avoids the propagation of errors between the decomposed levels, thus improving the measurement accuracy.

同传统运动估计算法相比，这种多层并行小波变换运动测量算法避免了估计中的误差传播，从而提高了测量精度。

This paper presents a multigrid parallel algorithm of one dimensional virtual boundary ransack forecast, and discusses its convergence.

对网格方程组作多重网格区域分裂并行计算，虚拟边界预报提出一维搜索算法，并讨论相应的收敛性；通过 GS迭代的性质，构造区间压缩算法，证明了其收敛性。

First, the recursive predictive errorlearning algorithm for recurrent neural networks is introduced, and its stability is demonstrated.

摘要针对递归神经网络BP学习算法收敛慢的缺陷，提出一种新的递归神经网络快速并行学习算法。

The main content of the dissertation is designing and realizing the total scheme and frame of system software. The module of the input and output is designed by using the generally I/O of the multichannel buffered serial port to simulated the communication of I~2C bus. The Ping-Pong buffer is brought forward to resolve the problem of the data rate matching. The logic switch function is designed in the programmed logic devices. The data transmission rate is improved by using the DMA and the dual buffer mechanism. The system program self-loading is realized by the design of the self-leading program. Lastly, the dissertation analyzes the characteristic of LLL video and experimentalize with some typical image processing algorithm.

主要对系统软件的总体方案与框架结构进行设计实现，内容包括利用通用串行口模拟I~2C通讯实现输入／输出模块的设计，提出乒乓缓冲结构来解决数据的速度匹配问题，利用可编程逻辑器设计完成逻辑切换控制功能，利用DMA并行传输方式并设计双缓冲结构提高数据的传输速度，设计自引导程序来实现系统程序自加载，最后深入探讨了微光视频图像的特征说明了几种典型的微光图像处理算法的原理，参考这些处理算法设计能在高速视频图像处理器上实现的软件方法，并做了一部分相关的实验研究。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。