我正在分析一种算法,该算法给出方阵的“峰值”的位置(这意味着该值的邻居小于或等于该值)。 有问题的算法效率很低,因为它会从位置(0,0)开始移动到大于该数字的邻居,从而一一检查值。这是代码:
def algorithm(problem, location = (0, 0), trace = None):
# if it's empty, it's done!
if problem.numRow <= 0 or problem.numCol <= 0: #O(1)
return None
nextLocation = problem.getBetterNeighbor(location, trace) #O(1)
#This evaluates the neighbor values and returns the highest value. If it doesn't have a better neighbor, it return itself
if nextLocation == location:
# If it doesnt have a better neighbor, then its a peak.
if not trace is None: trace.foundPeak(location) #O(1)
return location
else:
#there is a better neighbor, go to the neighbor and do a recursive call with that location
return algorithm(problem, nextLocation, trace) #O(????)
我知道最好的情况是峰值位于(0,0),并且确定以下情况是最坏的情况(使用10x10矩阵):
problem = [
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 10],
[34, 35, 36, 37, 38, 39, 40, 41, 0, 11],
[33, 0, 0, 0, 0, 0, 0, 42, 0, 12],
[32, 0, 54, 55, 56, 57, 0, 43, 0, 13],
[31, 0, 53, 0, 0, 58, 0, 44, 0, 14],
[30, 0, 52, 0, 0, 0, 0, 45, 0, 15],
[29, 0, 51, 50, 49, 48, 47, 46, 0, 16],
[28, 0, 0, 0, 0, 0, 0, 0, 0, 17],
[27, 26, 25, 24, 23, 22, 21, 20, 19, 18]]
请注意,它基本上使算法呈螺旋状运行,并且必须评估59个位置。
因此,问题是:在这种情况下,我如何获得时间复杂度?为什么? 我知道除递归外所有操作都是O(1),我迷路了
答案 0 :(得分:1)
对于您在示例中显示的大小为[m,n],
的任意矩阵,我们可以按以下方式分解由该算法(A)生成的给定矩阵的遍历:
n-1
个元素到元素8,m-1
个元素从9改为17,n-1
个元素,m-3
元素从27改为33,n-3
个元素从34个增加到40个,m-5
个元素从41改为45,n-5
个元素从46变为50,m-7
个元素这时,模式应该很清楚,因此可以建立以下最坏情况的重复关系:
T(m,n) = T(m-2,n-2) + m-1 + n-1
T(m,n) = T(m-4,n-4) + m-3 + n-3 + m-1 + n-1
...
T(m,n) = T(m-2i,n-2i) + i*m + i*n -2*(i^2)
其中i是迭代次数,并且仅当m-2i
和n-2i
都大于0时,这种重复才会继续。
WLOG我们可以假设m>=n
,因此该算法在m-2i>0
或m>2i
或im / 2迭代期间继续。因此,为i重新插入,我们得到:
T(m,n) = T(m-m,n-m) + m/2*m + m/2*n -2*((m/2)^2)
T(m,n) = 0 + m^2/2 + m*n/2 -2*((m^2/4))
T(m,n) = 0 + m^2/2 + m*n/2 -2*((m^2/4))
T(m,n) = m*n/2 = O(m*n)