当前位置:网站首页>动态规划(区间dp)
动态规划(区间dp)
2022-07-03 10:16:00 【Dαīsч】
(一)、基础
1、特点:
(1)、合并:将两个或多个部分进行整合,能将问题分解为能两两合并的形式
(2)、求解:对整个问题设最优值,枚举合并点,将问题分解为左右两个部分,最后合并两个部分的最优值得到原问题的最优值
(3)、复杂度:大多为 O ( n 3 ) O(n^3) O(n3),因此区间 d p dp dp的数据范围一般在 100 100 100左右
2、套路:第一层循环枚举区间长度,第二层循环枚举区间左端点,右端点也固定下来,第三层循环枚举区间的分割点
状态表示: d p [ i ] [ j ] dp[i][j] dp[i][j]表示区间 [ i , j ] [i,j] [i,j]能得到的最大(小)价值
状态转移: d p [ i ] [ j ] = m a x ( d p [ i ] [ j ] , d p [ i ] [ k ] + d p [ k + 1 ] [ j ] + c o s t ) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+cost) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+cost)
(二)、例题
1、石子合并
题意:在一个环上有 n n n个石子 a 1 , a 2 . . . . . . a n a_1,a_2......a_n a1,a2......an,进行 n − 1 n-1 n−1次合并操作,每次操作将相邻的两堆合并成一堆,能获得新的一堆中的石子数量的和的得分,求最大得分和最小得分( 1 ≤ n ≤ 100 1\leq n\leq 100 1≤n≤100)
题解:考虑不在环上的情况,那么我们可以把两堆石子合并看出成两个区间合并,代价即为两个区间内的石子数量的和
状态表示: d p [ i ] [ j ] dp[i][j] dp[i][j]代表区间 [ i , j ] [i,j] [i,j]石子堆能得到的最大(小)得分
状态转移方程: d p [ i ] [ j ] = m a x ( d p [ i ] [ j ] , d p [ i ] [ k ] + d p [ k + 1 ] [ j ] + s u m ( i , j ) ) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+sum(i,j)) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+sum(i,j))
为了处理环,我们把n个石子堆变为 2 ∗ n 2*n 2∗n个石子堆,这样对 2 ∗ n 2*n 2∗n个石子堆进行区间 d p dp dp,然后取从 i i i往后 n n n个石子堆能得到的最大值即可
int dp1[maxn][maxn], dp2[maxn][maxn];
int a[maxn], pre[maxn];
void solve()
{
int n;
cin >> n;
for (int i = 1; i <= n; i++)
{
cin >> a[i];
}
for (int i = n + 1; i <= 2 * n; i++)
{
a[i] = a[i - n];
}
n *= 2;
for (int i = 1; i <= n; i++)
{
pre[i] = pre[i - 1] + a[i];
}
memset(dp1, 0x3f, sizeof(dp1));
for (int i = 1; i <= n; i++)
{
dp1[i][i] = 0;
}
for (int sz = 1; sz < n; sz++)
{
for (int i = 1, j; (j = i + sz) <= n; i++)
{
for (int k = i; k < j; k++)
{
dp1[i][j] = min(dp1[i][j], dp1[i][k] + dp1[k + 1][j]);
dp2[i][j] = max(dp2[i][j], dp2[i][k] + dp2[k + 1][j]);
}
dp1[i][j] += pre[j] - pre[i - 1];
dp2[i][j] += pre[j] - pre[i - 1];
}
}
n /= 2;
int ans1 = inf, ans2 = 0;
for (int i = 1; i <= n; i++)
{
ans1 = min(ans1, dp1[i][i + n - 1]);
ans2 = max(ans2, dp2[i][i + n - 1]);
}
cout << ans1 << '\n' << ans2 << '\n';
}
2、能量项链
题意:有一能量珠,串前一颗能量珠的头标记为 m m m,尾标记为 r r r,后一颗能量珠的头标记为 r r r,尾标记为 r r r,则聚合后释放的能量为 m ∗ n ∗ r m*n*r m∗n∗r,新产生的珠子的头标记为 m m m,尾标记为 n n n,给定 n n n个数, 第 i i i个数为第 i i i颗珠子的头标记( 1 ≤ i ≤ m 1\leq i\leq m 1≤i≤m)例如 2 3 5 10 2\ 3\ 5\ 10 2 3 5 10对应的石子就是 ( 2 , 3 ) ( 3 , 5 ) ( 5 , 10 ) ( 10 , 2 ) (2,3)(3,5)(5,10)(10,2) (2,3)(3,5)(5,10)(10,2),如果合并第一个和最后一个石子那么获得能量是 10 ∗ 2 ∗ 3 10*2*3 10∗2∗3,求将所有珠子成一个能得到的最大能量( 4 ≤ n ≤ 100 4\leq n\leq 100 4≤n≤100)
题解:类似于石子合并
状态表示: d p [ i ] [ j ] dp[i][j] dp[i][j]代表区间 [ i , j ] [i,j] [i,j]能获得的最大能量
状态转移: d p [ i ] [ j ] = m a x ( d p [ i ] [ j ] , d p [ i ] [ k ] + d p [ k + 1 ] [ j ] + a [ i − 1 ] ∗ a [ k ] ∗ a [ j ] ) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+a[i-1]*a[k]*a[j]) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+a[i−1]∗a[k]∗a[j])
ll dp[maxn][maxn];
ll a[maxn];
void solve()
{
int n;
cin >> n;
for (int i = 1; i <= n; i++)
{
cin >> a[i];
a[n + i] = a[i];
}
a[0] = a[n];
n *= 2;
for (int sz = 2; sz <= n; sz++)
{
for (int i = 1, j; (j = i + sz - 1) <= n; i++)
{
for (int k = i; k < j; k++)
{
dp[i][j] = max(dp[i][j], dp[i][k] + dp[k + 1][j] + a[i-1] * a[k] * a[j]);
}
}
}
ll ans = 0;
for (int i = 1; i <= n / 2; i++)
{
ans = max(ans, dp[i][i + n / 2 - 1]);
}
cout << ans << '\n';
}
3、括号匹配对数
题意:寻找括号匹配对数,不可重复使用
题解:
状态表示: d p [ i ] [ j ] dp[i][j] dp[i][j]表示区间 [ i , j ] [i,j] [i,j]内有多少对匹配的括号
状态转移:两种转移 d p [ i ] [ j ] = d p [ i + 1 ] [ j − 1 ] + m a t c h ( i , j ) dp[i][j]=dp[i+1][j-1]+match(i,j) dp[i][j]=dp[i+1][j−1]+match(i,j)从稍小的区间转移,并且判断边界是否匹配
d p [ i ] [ j ] = m a x ( d p [ i ] [ j ] , d p [ i ] [ k ] + d p [ k + 1 ] [ j ] + m a t c h ( k , k + 1 ) ) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+match(k,k+1)) dp[i][j]=max(dp[i][j],dp[i][k]+dp[k+1][j]+match(k,k+1))从分界点转移,并且判断分界点是否匹配
int dp[maxn][maxn];
char s[maxn];
int match(int a, int b)
{
return s[a] == '(' && s[b] == ')' || s[a] == '[' && s[b] == ']';
}
void solve()
{
while (cin >> s + 1)
{
if (s[1] == 'e' && s[2] == 'n' && s[3] == 'd' && s[4] == '\0')
break;
memset(dp, 0, sizeof(dp));
int n = strlen(s + 1);
for (int sz = 2; sz <= n; sz++)
{
for (int i = 1, j; (j = i + sz - 1) <= n; i++)
{
dp[i][j] = dp[i + 1][j - 1] + match(i, j);
for (int k = i; k < j; k++)
{
dp[i][j] = max(dp[i][j], dp[i][k] + dp[k + 1][j] + match(k, k + 1));
}
}
}
cout << dp[1][n] * 2 << '\n';
}
}
边栏推荐
- php服务器 与redis交互大量CLOSE_WAIT分析
- I, a tester from a large factory, went to a state-owned enterprise with a 50% pay cut. I regret it
- MATLAB提取不规则txt文件中的数值数据(简单且实用)
- Unity移动端游戏性能优化简谱之 画面表现与GPU压力的权衡
- Oracle收回权限 & 创建角色
- 行业唯一!法大大电子合同上榜36氪硬核科技企业
- 如何成为一名高级数字 IC 设计工程师(1-4)Verilog 编码语法篇:表达式
- Google Earth Engine(GEE)——GHSL 全球人口网格数据集250米分辨率
- 按键切换:按F1-F12都需要按Fn
- one hot 独热码
猜你喜欢
The element form shows the relationship between elementary transformation and elementary matrix
Tencent micro app to get wechat user information
基于I2C协议的驱动开发
用了这么久线程池,你真的知道如何合理配置线程数吗?
After a month, I finally got Kingdee offer! Share tetrahedral Sutra + review materials
The testing department of the company came to the king of the Post-00 roll, and the veteran exclaimed that it was really dry, but
C语言 AES加解密
数据库增量备份 - DB INCR DB FULL
活动预告 | 直播行业“内卷”,以产品力拉动新的数据增长点
Software testing e-commerce projects that can be written into your resume, don't you come in and get it?
随机推荐
Clion debug
解决undefined reference to `__aeabi_uidivmod‘和undefined reference to `__aeabi_uidiv‘错误
ORACLE 11G 单机冷备数据库
Lecture 1 number field
高精度室内定位技术,在智慧工厂安全管理的应用
程序员的创业陷阱:接私活
读书笔记:《心若菩提》 曹德旺
How to clean up v$rman_ backup_ job_ Details view reports error ora-02030
图解网络:什么是虚拟路由器冗余协议 VRRP?
The manuscript will be revised for release tonight. But, still stuck here, maybe what you need is a paragraph.
Analysis of JMM memory model
Kotlin's use of the no Arg compiler plug-in in gradle
[vtk] interpretation of source code comments of vtkwindowedsincpolydatafilter
进程与线程
项目管理精华读书笔记(七)
面试题总结(2) IO模型,集合,NIO 原理,缓存穿透,击穿雪崩
程序进程管理工具-go supervisor
Leetcode 46: full arrangement
2021 postgraduate entrance examination mathematics 2 linear algebra
Test what the leader should do