132. Palindrome Partitioning II 回文分割
Given a string s, partition s such that every substring of the partition is a palindrome.
Return the minimum cuts needed for a palindrome partitioning of s.
Example:
Input: "aab" Output: 1 Explanation: The palindrome partitioning ["aa","b"] could be produced using 1 cut.
从后往前
首先是记忆化数组pal(至于为什么yao要从后xian向前便利,一会看看是否和记忆化数组有关)
串分割成回文字串最少需要多少步
一个二维数组,来存储从位置i到位置j是否构成回文,用的是记忆化存储,以往当我们进行回文序列查找的时候,当位置i变化的时候,总是经历了重复工作,所以用记忆话数组pal来存储各个状态
vector<vector<bool>>pal(n,vector(n,false))
比如我们到达位置,想知道从i'开始往后是否有回文数组
int j=i; if(j-i<2),也就是说此时要么j和i相邻,要么j和i之间相差一个,那么此时str[i]==str[j], 如果说j-i>2,那么就要要求,从i+1到j-1之间都要构成回文 ,也就是说pal[i+1][j-1]==true for(int j=i;j<n;j++) if(str[i]==str[j]&&(j-i<2||pal[i+1][j-1]) dp[i]=min[dp[i],dp[j+1]+1;
Calculate and maintain 2 DP states:
pal[i][j] , which is whether s[i..j] forms a pal
d[i], which
is the minCut for s[i..n-1]Once we comes to a pal[i][j]==true:
if j==n-1, the string s[i..n-1] is a Pal, minCut is 0, d[i]=0;
else: the current cut num (first cut s[i..j] and then cut the rest
s[j+1...n-1]) is 1+d[j+1], compare it to the exisiting minCut num
d[i], repalce if smaller.
d[0] is the answer.所以就可以写出如下代码段
class Solution { public: int minCut(string s) { if(s.empty()) return 0; int n = s.size(); vector<vector<bool>> pal(n,vector<bool>(n,false)); vector<int> d(n); for(int i=n-1;i>=0;i--) { d[i]=n-i-1; for(int j=i;j<n;j++) { if(s[i]==s[j] && (j-i<2 || pal[i+1][j-1])) { pal[i][j]=true; if(j==n-1) d[i]=0; else if(d[j+1]+1<d[i]) d[i]=d[j+1]+1; } } } return d[0]; } };
从前往后
One typical solution is DP based. Such solution first constructs a two-dimensional bool array isPalin to indicate whether the sub-string s[i..j] is palindrome. To get such array, we need O(N^2) time complexity. Moreover, to get the minimum cuts, we need another array minCuts to do DP and minCuts[i] saves the minimum cuts found for the sub-string s[0..i-1]. minCuts[i] is initialized to i-1, which is the maximum cuts needed (cuts the string into one-letter characters) and minCuts[0] initially sets to -1, which is needed in the case that s[0..i-1] is a palindrome. When we construct isPalin array, we update minCuts everytime we found a palindrome sub-string, i.e. if s[i..j] is a palindrome, then minCuts[j+1] will be updated to the minimum of the current minCuts[j+1] and minCut[i]+1(i.e. cut s[0..j] into s[0,i-1] and s[i,j]). At last, we return minCuts[N].
So the complexity is O(N^2). However, it can be further improved since as described above, we only update minCuts when we find a palindrome substring, while the DP algorithm spends lots of time to calculate isPalin, most of which is false (i.e. not a palindrome substring). If we can reduce such unnecessary calculation, then we can speed up the algorithm. This can be achieved with a Manancher-like solution, which is also given as following.// DP solution class Solution { public: int minCut(string s) { const int N = s.size(); if(N<=1) return 0; int i,j; bool isPalin[N][N]; fill_n(&isPalin[0][0], N*N, false); int minCuts[N+1]; for(i=0; i<=N; ++i) minCuts[i] = i-1; for(j=1; j<N; ++j) { for(i=j; i>=0; --i) { if( (s[i] == s[j]) && ( ( j-i < 2 ) || isPalin[i+1][j-1] ) ) { isPalin[i][j] = true; minCuts[j+1] = min(minCuts[j+1], 1 + minCuts[i]); } } } return minCuts[N]; } };