一、01背包
Bone Collector HDU - 2602
Many years ago , in Teddy’s hometown there was a man who was called “Bone Collector”. This man like to collect varies of bones , such as dog’s , cow’s , also he went to the grave …
The bone collector had a big bag with a volume of V ,and along his trip of collecting there are a lot of bones , obviously , different bone has different value and different volume, now given the each bone’s value along his trip , can you calculate out the maximum of the total value the bone collector can get ?
【题解】
import java.util.*;
public class Main {
// 注意坑点。。。每个物品的容量也有可能是0而价值大于0啊。。。
public static void main(String[] args) {
Scanner scanner = new Scanner(System.in);
int T = scanner.nextInt();
for(int t = 0; t < T; t++){
int N = scanner.nextInt();
int V = scanner.nextInt();
int[] value = new int[N];
int[] volume = new int[N];
for(int i = 0; i < N; i++){
value[i] = scanner.nextInt();
}
for(int i = 0; i < N; i++){
volume[i] = scanner.nextInt();
}
int ans = zeroOneBackpack3(V, N, value, volume);
System.out.println(ans);
}
}
// 1. 记忆化搜索(暴力递归超时,改成记忆化搜索)
private static int zeroOneBackpack(int V, int N, int[] value, int[] volume){
int[][] dp = new int[V+1][N];
for(int i = 0; i <= V; i++){
Arrays.fill(dp[i], -1);
}
return f(V, value, volume, N-1, dp);
}
// (关键)容量为V的包放入前i个物品,所获的最大价值
private static int f(int V, int[] value, int[] volume, int i, int[][] dp){
if(i < 0 || V < 0){ // 注意这里V=0时,也有可能放物品,因为物品有可能容量是0但价值大于0
return 0;
}
if(dp[V][i] != -1){
return dp[V][i];
}
if(V < volume[i]){
return dp[V][i] = f(V, value, volume, i-1, dp);
}
return dp[V][i] = Math.max(f(V, value, volume, i-1, dp), f(V-volume[i], value, volume, i-1, dp) + value[i]);
}
// 2. 动态规划(分析暴力递归或记忆化搜索表,找到递推依赖关系,直接改成此)
private static int zeroOneBackpack2(int V, int N, int[] value, int[] volume){
int[][] dp = new int[V+1][N];
for(int k = 0; k <= V; k++){
dp[k][0] = volume[0] > k ? 0 : value[0];
}
for(int j = 1; j < N; j++){
for(int v = 0; v <= V; v++){
if(volume[j] > v){
dp[v][j] = dp[v][j-1];
}else{
dp[v][j] = Math.max(dp[v][j-1], dp[v-volume[j]][j-1]+value[j]);
}
}
}
return dp[V][N-1];
}
// 3.动态规划(状态压缩),分析递推表发现,(i,j)位置的值仅取决于(k,j-1)的值,其中k<=i。
// 参考易误点分析图,压缩为一位数组后,应该采用自下而上递推。
// 自下而上推导,下面的值确定后,不会影响上面的值。
// 自上而下推导,上面的值会影响下面的值。
private static int zeroOneBackpack3(int V, int N, int[] value, int[] volume){
int[] dp = new int[V+1];
for(int k = 0; k <= V; k++){
dp[k] = volume[0] > k ? 0 : value[0];
}
for(int j = 1; j < N; j++){
for(int v = V; v >= volume[j]; v--){
dp[v] = Math.max(dp[v], dp[v-volume[j]]+value[j]);
}
}
return dp[V];
}
}
关于递归改动态规划?
递归函数参数有几个是可变值,就对应几维数组即几维表格(状态确定),然后根据递归方程分析状态转移表,确定动态规划转移方程,再根据递归结束条件对状态进行初始化。
关于状态压缩的时候的易错点分析?
二、完全背包
寒冰王座 HDU - 1248
【题解】
import java.util.*;
public class Main {
public static void main(String[] args) {
Scanner scanner = new Scanner(System.in);
int[] volume = {150, 200, 350};
int[] value = {150, 200, 350};
int T = scanner.nextInt();
while (T-- > 0){
int n = scanner.nextInt();
int ans = n - completeBackpack4(n, volume, value);
System.out.println(ans);
}
}
// 1.普通递归(容易改成记忆化搜索,略)
private static int completeBackpack(int V, int[] volume, int[] value){
return f1(V, volume, value, volume.length-1);
}
// (关键)类比于01背包,容量为V,放入前i种物品,产生的最大价值,每种物品最多可以用V/volume[i]次
// (关键)f(V, i) = max(f(V-k*volume[i], i-1) + k*value[i]), k取值:[0,V/volume[i]]
private static int f1(int V, int[] volume, int[] value, int i){
if(i < 0 || V < 0){
return 0;
}
int ans = 0;
for(int k = 0; k*volume[i] <= V; k++){
ans = Math.max(ans, f1(V-k*volume[i], volume, value, i-1) + k*value[i]);
}
return ans;
}
// 2.动态规划
private static int completeBackpack2(int V, int[] volume, int[] value){
int[][] dp = new int[V+1][volume.length];
for(int k = 0; k <= V; k++){
dp[k][0] = k/volume[0]*value[0];
}
for(int j = 1; j < volume.length; j++){
for(int v = 0; v <= V; v++){
int ans = 0;
for(int p = 0; p*volume[j] <= v; p++){
ans = Math.max(ans, dp[v-p*volume[j]][j-1] + p*value[j]);
}
dp[v][j] = ans;
}
}
return dp[V][volume.length-1];
}
// 3.动态规划(状态压缩)
private static int completeBackpack3(int V, int[] volume, int[] value){
int[] dp = new int[V+1];
for(int k = 0; k <= V; k++){
dp[k] = k/volume[0]*value[0];
}
for(int j = 1; j < volume.length; j++){
for(int v = V; v >= 0; v--){
int ans = 0;
for(int p = 0; p*volume[j] <= v; p++){
ans = Math.max(ans, dp[v-p*volume[j]] + p*value[j]);
}
dp[v] = ans;
}
}
return dp[V];
}
// 4.转成01背包(同时用二进制优化,采用01背包最终的优化版本)
// 虽说某种物品数量无限,但总容量空间有限。可以把物品拆分成0,2^0, 2^1, 2^2,……, 这些数字组合可在
// 总容量允许的范围内取得任意值。如果不二进制优化直接转为01背包,物品的数量大大增加。。。
// 注意这里和多重背包不同,因为物品数量无限,所以可以一直无穷的二进制下去,人家用不用不管,反正人
// 家有的是啊。。。所以没错啊。。。但多重背包就必须那么保守的处理了啊。。因为你数量有限啊。。。
private static int completeBackpack4(int V, int[] volume, int[] value){
int[] dp = new int[V+1];
for(int k = 0; k <= V; k++){
dp[k] = k/volume[0]*value[0];
}
for(int j = 1; j < volume.length; j++){
for(int p = 0; p*volume[j] <= V; p = (p==0) ? 1 : (p <<= 1)) { // 二进制优化
for (int v = V; v >= p*volume[j]; v--) {
dp[v] = Math.max(dp[v], dp[v - p*volume[j]] + p*value[j]);
}
}
}
return dp[V];
}
// 5.公式推导优化。。。
// (1) f(V, i) = max(f(V-k*volume[i], i-1) + k*value[i]), k取值:[0,V/volume[i]]
// (2) f(V, i) = max(f(V, i-1), f(V-k*volume[i], i-1) + k*value[i]), k取值:[1, V/volume[i]]
// (3) f(V, i) = max(f(V, i-1), max(f(V-volume[i] - k*volume[i], i-1) + k*value[i]) + value[i]), k取值:[0, V/volume[i]-1]
// (1)带入(3)得到(4)
// (4) f(V, i) = max(f(V, i-1), f(V-volume[i], i) + value[i])
// 由上述(4)式容易写出递归函数或动态规划,下面只给出动态规划经过状态压缩的最终形式
private static int completeBackpack5(int V, int[] volume, int[] value){
int[] dp = new int[V+1];
for(int k = 0; k <= V; k++){
dp[k] = k/volume[0]*value[0];
}
for(int j = 1; j < volume.length; j++){
for(int v = volume[j]; v <= V; v++){
dp[v] = Math.max(dp[v], dp[v-volume[j]] + value[j]);
}
}
return dp[V];
}
}
三、多重背包
和完全背包解法基本相同。要么直接转成01背包,要么模仿01背包的递归思路写出类似代码,然后改成记忆化搜索或动态规划再状态压缩。
悼念512汶川大地震遇难同胞——珍惜现在,感恩生活 HDU - 2191
【题解】
package test;
import java.util.*;
public class Main {
public static void main(String[] args) {
Scanner scanner = new Scanner(System.in);
int T = scanner.nextInt();
while (T-- > 0){
int n = scanner.nextInt();
int m = scanner.nextInt();
int[] volume = new int[m];
int[] value = new int[m];
int[] count = new int[m];
for(int i = 0; i < m; i++){
volume[i] = scanner.nextInt();
value[i] = scanner.nextInt();
count[i] = scanner.nextInt();
}
int ans = multipleBackpack4(n, m, volume, value, count);
System.out.println(ans);
}
}
// 1.记忆化搜索
private static int multipleBackpack(int V, int N, int[] volume, int[] value, int[] count){
int[][] dp = new int[V+1][N];
for(int i = 0; i <= V; i++){
Arrays.fill(dp[i], -1);
}
return f(V, volume, value, count, N-1, dp);
}
// (关键)类比于01背包,容量为V,放入前i种物品,产生的最大价值,每种物品最多可以用count[i]次
// (关键)f(V, i) = max(f(V-k*volume[i], i-1) + k*value[i]), k取值:[0,Math.min(count[i], V/volume[i])]
private static int f(int V, int[] volume, int[] value, int[] count, int i, int[][] dp){
if(i < 0 || V < 0){
return 0;
}
if(dp[V][i] != -1){
return dp[V][i];
}
int ans = 0;
for(int k = 0; k <= Math.min(count[i], V/volume[i]); k++){
ans = Math.max(ans, f(V-k*volume[i], volume, value, count, i-1, dp) + k*value[i]);
}
return dp[V][i] = ans;
}
// 2.动态规划
private static int multipleBackpack2(int V, int N, int[] volume, int[] value, int[] count){
int[][] dp = new int[V+1][N];
for(int k = 0; k <= V; k++){
dp[k][0] = Math.min(k/volume[0], count[0])*value[0];
}
for(int j = 1; j < N; j++){
for(int v = 0; v <= V; v++){
int ans = 0;
for(int p = 0; p <= Math.min(v/volume[j], count[j]); p++){
ans = Math.max(ans, dp[v-p*volume[j]][j-1] + p*value[j]);
}
dp[v][j] = ans;
}
}
return dp[V][N-1];
}
// 3.动态规划(状态压缩)
private static int multipleBackpack3(int V, int N, int[] volume, int[] value, int[] count){
int[] dp = new int[V+1];
for(int k = 0; k <= V; k++){
dp[k] = Math.min(k/volume[0], count[0])*value[0];
}
for(int j = 1; j < N; j++){
for(int v = V; v >= 0; v--){
int ans = 0;
for(int p = 0; p <= Math.min(v/volume[j], count[j]); p++){
ans = Math.max(ans, dp[v-p*volume[j]] + p*value[j]);
}
dp[v] = ans;
}
}
return dp[V];
}
// 4.转成01背包问题(同时使用二进制优化,并使用01背包的最终优化版本)
// 举例,V=10,某种物品可拆成1,2,4,(10-7)=3,为啥这么拆之后确能凑出[1, 10]所有数,待证明
private static int multipleBackpack4(int V, int N, int[] volume, int[] value, int[] count){
List<Integer> volumeList = new ArrayList<>();
List<Integer> valueList = new ArrayList<>();
int total = 0;
for(int i = 0; i < N; i++){
int j = 0;
for(; (1<<(j+1)) - 1 < count[i]; j++){
total++;
volumeList.add(volume[i]*(1<<j));
valueList.add(value[i]*(1<<j));
}
int remain = count[i] - (1<<j) + 1;
if(remain <= 0){
continue;
}
total++;
volumeList.add(volume[i]*remain);
valueList.add(value[i]*remain);
}
int[] valueNew = new int[valueList.size()];
int[] volumeNew = new int[volumeList.size()];
for(int i = 0; i < valueList.size(); i++){
valueNew[i] = valueList.get(i);
volumeNew[i] = volumeList.get(i);
}
return zeroOneBackpack3(V, total, valueNew, volumeNew);
}
private static int zeroOneBackpack3(int V, int N, int[] value, int[] volume){
int[] dp = new int[V+1];
for(int k = 0; k <= V; k++){
dp[k] = volume[0] > k ? 0 : value[0];
}
for(int j = 1; j < N; j++){
for(int v = V; v >= volume[j]; v--){
dp[v] = Math.max(dp[v], dp[v-volume[j]]+value[j]);
}
}
return dp[V];
}
}
四、二维费用背包
474. 一和零
【代码1】超时—递归
class Solution {
public int findMaxForm(String[] strs, int m, int n) {
return f(strs, m, n, strs.length-1);
}
private int f(String[] strs, int m, int n, int i){
if(i < 0){
return 0;
}
int c0 = 0;
for(int k = 0; k < strs[i].length(); k++){
if(strs[i].charAt(k) == '0'){
c0++;
}
}
int c1 = strs[i].length() - c0;
if(m < c0 || n < c1){
return f(strs, m, n, i-1);
}
return Math.max(f(strs, m, n, i-1), f(strs, m-c0, n - c1, i-1) + 1);
}
}
【代码2】通过—记忆化搜索
class Solution {
public int findMaxForm(String[] strs, int m, int n) {
int[][][] dp = new int[m+1][n+1][strs.length];
for(int i = 0; i <= m; i++){
for(int j = 0; j <= n; j++){
for(int k = 0; k < strs.length; k++){
dp[i][j][k] = -1;
}
}
}
int[] cnt0 = new int[strs.length];
int[] cnt1 = new int[strs.length];
for(int i = 0; i < strs.length; i++){
int c0 = 0;
for(int k = 0; k < strs[i].length(); k++){
if(strs[i].charAt(k) == '0'){
c0++;
}
}
int c1 = strs[i].length() - c0;
cnt0[i] = c0;
cnt1[i] = c1;
}
return f(cnt0, cnt1, m, n, strs.length-1, dp);
}
// 二维费用背包问题。m个0,n个1,前i个字符串,最多能消耗掉多少个字符串
private int f(int[] cnt0, int[] cnt1, int m, int n, int i, int[][][] dp){
if(i < 0){
return 0;
}
if(m >= 0 && n >= 0 && dp[m][n][i] != -1){
return dp[m][n][i];
}
if(m < cnt0[i] || n < cnt1[i]){
return dp[m][n][i] = f(cnt0, cnt1, m, n, i-1, dp);
}
return dp[m][n][i] = Math.max(f(cnt0, cnt1, m, n, i-1, dp), f(cnt0, cnt1, m-cnt0[i], n - cnt1[i], i-1, dp) + 1);
}
}
【代码3】通过—动态规划
class Solution {
public int findMaxForm(String[] strs, int m, int n) {
int[] cnt0 = new int[strs.length];
int[] cnt1 = new int[strs.length];
for(int i = 0; i < strs.length; i++){
int c0 = 0;
for(int k = 0; k < strs[i].length(); k++){
if(strs[i].charAt(k) == '0'){
c0++;
}
}
int c1 = strs[i].length() - c0;
cnt0[i] = c0;
cnt1[i] = c1;
}
int[][][] dp = new int[m+1][n+1][strs.length];
for(int i = 0; i <= m; i++){
for(int j = 0; j <= n; j++){
dp[i][j][0] = (i >= cnt0[0] && j >= cnt1[0]) ? 1 : 0;
}
}
for(int k = 1; k < cnt0.length; k++){
for(int i = 0; i <= m; i++){
for(int j = 0; j <= n; j++){
if(i >= cnt0[k] && j >= cnt1[k]){
dp[i][j][k] = Math.max(dp[i][j][k-1], dp[i-cnt0[k]][j-cnt1[k]][k-1] + 1);
}else{
dp[i][j][k] = dp[i][j][k-1];
}
}
}
}
return dp[m][n][cnt0.length-1];
}
}
【代码4】通过—动态规划(状态压缩)
class Solution {
public int findMaxForm(String[] strs, int m, int n) {
int[] cnt0 = new int[strs.length];
int[] cnt1 = new int[strs.length];
for(int i = 0; i < strs.length; i++){
int c0 = 0;
for(int k = 0; k < strs[i].length(); k++){
if(strs[i].charAt(k) == '0'){
c0++;
}
}
int c1 = strs[i].length() - c0;
cnt0[i] = c0;
cnt1[i] = c1;
}
int[][] dp = new int[m+1][n+1];
for(int i = 0; i <= m; i++){
for(int j = 0; j <= n; j++){
dp[i][j] = (i >= cnt0[0] && j >= cnt1[0]) ? 1 : 0;
}
}
for(int k = 1; k < cnt0.length; k++){
for(int i = m; i >= cnt0[k]; i--){
for(int j = n; j >= cnt1[k]; j--){
dp[i][j] = Math.max(dp[i][j], dp[i-cnt0[k]][j-cnt1[k]] + 1);
}
}
}
return dp[m][n];
}
}