重读《Programming Pearls》之八：Algorithm Design Techniques

Problem——最大子序列和:

Find the maximum sum in any contiguous subvector of a vector of n numbers.

扩展问题：最大子矩阵和

 

 

最大子序列和的解法：

Brute Force: O(n^3)

maxsofar = 0 for i = [0, n) for j = [i, n) sum = 0 for k = [i, j] sum += x[k] maxsofar = max(maxsofar, sum) 

 

The Quadratic Algorithms(Dynamic Programming/Memory Searching): O(n^2)

算法一 maxsofar = 0 for i = [0, n) sum = 0 for j = [i, n) sum += x[j] maxsofar = max(maxsofar, sum) 算法二 cumarr[-1] = 0 for i = [0, n) cumarr[i] = cumarr[i-1] + x[i] maxsofar = 0 for i = [0, n) for j = [i, n) sum = 0 sum = cumarr[j] - cumarr[i-1] maxsofar = max(maxsofar, sum) 

 

Divide-and-Conquer Algorithm: O(n log n)

 

float masum3(l, u) { if (l > u) return 0 if (l == u) return max(0, x[l]) m = (l + u) / 2 /* find max crossing to left */ lmax = sum = 0 for (i = m; i >= l; i --) sum += x[i] lmax = max(lmax, sum) rmax = sum = 0 for i = (m, u] sum += x[i] rmax = max(rmax, sum) return max(lmax+rmax, maxsum3(l, m), maxsum3(m+1, u)) } 

 

A Scanning Algorithm（Greedy Algorithm）: O(n)

 

maxsofar = 0 maxendinghere = 0 for i = [0, n) maxendinghere = max(maxendinghere + x[i], 0) maxsofar = max(maxsofar, maxendinghere 

 

 

Principles:

Algorithms design techiques:

1、Save state to avoid recomputation保存状态（空间换时间）

2、Preprocess infomation into data structure预处理

3、Divide-and-Conquer algorithm分治

4、Sanning algorithm贪心

5、Cumulatives（Table）打表

6、Lower bounds。证明自己的算法的复杂度不可能再低（达到下限）

 

 

What Does It Matter?——时间复杂度的影响

 

时空复杂度的估计：

参考

http://baike.baidu.com/view/104946.htm

http://baike.baidu.com/view/540497.htm