1.txt

This complete matlab for neural network

发信人: yaomc (白头翁&山东大汉), 信区: DataMining
标  题: [合集]请问数据挖掘和大规模并行数据挖掘
发信站: 南京大学小百合站 (Fri Apr  5 13:08:03 2002), 站内信件

yeli (旗手) 于Wed Apr  3 21:03:11 2002)
提到：

这两个方向哪个更好一点。
其中大规模并行数据挖掘是在cluster机上运行的。


ddren (无声的雨) 于Thu Apr  4 10:11:48 2002提到：

最近两年并行数据挖掘是研究热点
不过我看过的论文都是用PVM实现并行的
我不懂什么时cluster机

可否给讲一讲啊？
是聚类机么？


fervvac (高远) 于Thu Apr  4 12:08:23 2002提到：

The originaal question is invalide, A \supset B, :p

As far as I know, the current focus of the DM researchers are on mining sequence
data, mainly because there are many applications.

Another field is mining on the web.

There are about 300 submissions for this year's KDD, so it seems still a hot
topic to do, :p



yeli (旗手) 于Thu Apr  4 16:49:40 2002)
提到：

cluster机器是一种穷人的大型机,是N个处理单元通过网络联起来,不过不使用常见的网络
协议(因为这会带来开销),能达到4500亿次每秒的峰值浮点计算.一般用linux系统.
我们认为随着数据量的增加,对复杂巨系统,数据挖掘应该是大规模运算,不是一台计算机能
实现的.如果能在各个有计算能力的机器上进行,不是更好吗?主要研究挖掘算法的并行处理
以及中间件的构造.

fervvac (高远) 于Fri Apr  5 11:24:52 2002提到：

Just curious about this topic:
1) How are N processors interconnected? Via mainboard or share-memory or share-
disk? What is the network topology? Bus, star or hypercube?
2) MS claimed its NT clusters are good. Is that relevant? If linux is 
recommended, then the cluster is mostly based on software layer?
3) What's the difference between cluster and grid?