哈夫曼树
给定n个权值作为n个叶子节点,构造一棵二叉树,若该树的带权路径长度达到最小,称这样的二叉树为最优二叉树,也称为哈夫曼树(Huffman Tree)。哈夫曼树是带权路径长度最短的树,权值较大的结点离根较近。
哈夫曼树代码实现
示例图:以字符串AAABBCDDDDD为例构建哈夫曼树
哈夫曼节点类
public class HaffmanNode {
int val; //值
int weight;//权值
char ch;//所放字符串
HaffmanNode left;//左孩子
HaffmanNode right;//右孩子
public HaffmanNode(int val, char ch) {
this.val = val;
this.ch = ch;
}
@Override
public String toString() {
return "HaffmanNode [val=" + val + ", weight=" + weight + ", ch=" + ch + "]";
}
}
哈夫曼树实现类
import java.util.Comparator;
import java.util.HashMap;
import java.util.Map;
import java.util.Map.Entry;
import java.util.PriorityQueue;
public class Haffman {
HaffmanNode root;
int count[] = new int[35];//统计字符出现的次数,这里只收集 A~Z的
public void build(char[] arr) {
//先统计字符数组中的字符出现的次数
for (int i = 0; i < arr.length; i++) {
count[arr[i] - 'A']++;
}
//创建优先队列,默认是优先级大的,先出队列,数值越小,优先级别越大
//按照哈夫曼编码最小的两个节点先出,所以修改比较设置
PriorityQueue<HaffmanNode> priorityQueue = new PriorityQueue<>(new Comparator<HaffmanNode>() {
public int compare(HaffmanNode o1, HaffmanNode o2) {
return o1.val - o2.val;
}
});
//新建字符节点,并且将节点按照出现次数作为优先级放入优先队列中
for (int i = 0; i < count.length; i++) {
if (count[i] != 0) {
priorityQueue.offer(new HaffmanNode(count[i], (char) (65 + i)));
}
}
//当优先队列中的节点数大于1时,
while (priorityQueue.size() != 1) {
//取出两个最小的以及第二小的节点
HaffmanNode left = priorityQueue.poll();
HaffmanNode right = priorityQueue.poll();
//创建一个新的节点,放入优先队列中,值为左孩子加右孩子的值
HaffmanNode newHaffmanNode = new HaffmanNode(left.val + right.val, ' ');
newHaffmanNode.left = left;
newHaffmanNode.right = right;
priorityQueue.add(newHaffmanNode);
}
//当优先队列中只剩一个时,此时的节点就是已经构建好的哈夫曼树
root = priorityQueue.poll();
}
}
更新权值
//map存储字符出现的权值
private Map<Character, Integer> map1 = null;
private void getWeight() {
//更新权值
map1 = new HashMap<>();
count(root, 0);
}
//获取权值
public Integer getWeight(char ch) {
if (map1 == null) {
getWeight();
}
return map1.get(ch);
}
public void count(HaffmanNode haffmanNode, int weight) {
//递归更新权值
if (haffmanNode != null && haffmanNode.ch != ' ') {
map1.put(haffmanNode.ch, weight);
}
if (haffmanNode.left != null) {
count(haffmanNode.left, weight + 1);
}
if (haffmanNode.right != null) {
count(haffmanNode.right, weight + 1);
}
}
获取哈夫曼编码
//存储一个字符的哈夫曼编码后的编码集
private Map<Character, String> map = new HashMap<>();
public void bianma() {
bianma(root, "");
}
private void bianma(HaffmanNode node, String str) {
if (node.ch >= 'A' && node.ch <= 'Z') {
//记录字符的哈夫曼编码
map.put(node.ch, str);
}
if (node.left != null) {
//往左侧走,编码为0
bianma(node.left, str + "0");
}
if (node.right != null) {
//往右侧走,编码为1
bianma(node.right, str + "1");
}
}
应用
1.哈夫曼树又称为最优树,树的带权路径长度规定为所有叶子结点的带权路径长度之和,记为WPL。
2.数据压缩