P*******b 发帖数: 1001 | |
v******k 发帖数: 808 | 2 in case of any large-scale computation that could be done distributively
for instance, counting number of edges in a directed graph in a 1M+ node
graph, one can simply put each two nodes to a mapper and let hadoop take
care of it; its obviously O(n^2) effort but works for large-scale data set
HIH |
M********5 发帖数: 715 | 3 I am wondering whether the tradeoff for space is a possible answer. |
s*********t 发帖数: 1663 | 4 O(n)空间占太多?
【在 P*******b 的大作中提到】 : thanks
|
I**A 发帖数: 2345 | 5 Give some scenarios where you might favor O(n^2) algorithm over a O(nlg(n))
i. If O(nlg(n)) requires space, while O(n^2) doesn’t
ii. If O(nlg(n)) is difficult to understand and to implement, while O(n^2
) is easier
iii. If O(n^2) has a general running time less than O(nlg(n)) instead
iv. If O(n^2) is easy to serialize, while O(nlg(n)) not
【在 P*******b 的大作中提到】 : thanks
|