贪心算法
1)Greedy algorithm
称之为贪心算法或者贪婪算法,核心思想是
- 将寻找最优解的问题分为若干个步骤
- 每一步骤都采用贪心原则,选取当前最优解
- 因为未考虑所有可能,局部最优的堆叠不一定得到最终解最优
贪心算法例子
Dijkstra
java
while (!list.isEmpty()) {
// 选取当前【距离最小】的顶点
Vretex curr = chooseMinDistVertex(list);
// 更新当前顶点到相邻顶点距离
updateNeighboursDish(curr);
// list集合中移除当前顶点
list.remove(curr);
// 标记当前顶点已被访问
curr.visited = true;
}
- 未能找到最短路径:存在负边 情况,得不到正确解;
- 贪心原则会认为本次已找到该顶点的最短路径,使得该顶点赋为已访问 ;
- 与之对比,Bellman-Ford算法并未考虑局部距离最小顶点,而是每次处理所有边 ,不会出错,当然效率不如Dijkstra算法;
prim
java
while (!list.isEmpty()) {
// 选取当前【距离最小】的顶点
Vretex curr = chooseMinDistVertex(list);
// 更新当前顶点到相邻顶点距离
updateNeighboursDish(curr);
// list集合中移除当前顶点
list.remove(curr);
// 标记当前顶点已被访问
curr.visited = true;
}
prim与Dijkstra的区别在于,根据距离选取最小顶点不同,Dijkstra算法是一个累加距离 ,而prim算法中距离是跟相邻顶点间的距离 。
KrusKal
java
List<String> list = new ArrayList<>();
DisjoinSet set = new DisjoinSet(size);
while (list.size() < size - 1) {
// 选取当前【距离最短】的边
Edge poll = queue.poll();
int i = set.find(poll.start);
int j = set.find(poll.end);
// 判断两个集合是否相交
if (i != j) {// 未相交
list.add(poll);
set.union(i, j); // 相交操作
}
}
其他贪心算法例子
- 选择排序、堆排序
- 拓扑排序
- 并查集和中的union by size 和union by height
- 哈夫曼编码
- 钱币找零
- 任务编排
- 近似算法
零钱兑换ⅡLeetcode518
递归实现
java
public class Leetcode518 {
public int change(int[] coins, int amount) {
return coinChange(coins, 0, amount, new LinkedList<>(), true);
}
public int coinChange(int[] coins, int index, int amount, LinkedList<Integer> stack, boolean first) {
if (!first) {
stack.push(coins[i]);
}
int sum = 0;
if (amount == 0) {
System.out.printlin(stack);
sum = 1;
} else if (amount < 0) {
System.out.println(stack)
sum = 0;
} else {
for(int i = index; i < coins.length; i++) {
sum += coinChange(coins, i, amount - coins[i], stack, false);
}
}
if (!stack.isEmpty()) {
stack.pop();
}
return sum;
}
}
零钱兑换Leetcode322
递归实现
java
public class Leetcode322 {
static int min = -1;
public int change(int[] coins, int amount) {
coinChange(coins, 0, amount, new AtomicInteger(-1), new LinkedList<Integer>(), true);
return min;
}
public void coinChange(int[] coins, int index, int amount, AtomicInteger count, LinkedList<Integer> stack, boolean first) {
if (!first) {
stack.push(coins[i]);
}
count.incrementAndGet(); // count++;
int sum = 0;
if (amount == 0) {
System.out.println(stack);
if (min == -1) {
min = min.get();
} else {
min = Math.min(min, count.get());
}
} else {
for(int i = index; i < coins.length; i++) {
sum += coinChange(coins, i, amount - coins[i], count, stack, false);
}
}
count.decrementAndGet(); // count--;
if (!stack.isEmpty()) {
stack.pop();
}
return sum;
}
public static void main(String[] args) {
int[] coins = {5, 2, 1};
Leetcode322 leetcode = new Leetcode322();
System.out.printlin(leetcode.coinChange(coins, 5));
}
}
贪心实现
java
public class Leetcode322{
public int coinChange(int[] coins, int amount) {
// 前提是coins是降序排序
int reminder = amount;
int count;
for(int coin : coins) {
while (reminder > coin) {
reminder -= coin;
count++;
}
if (reminder == coin) {
reminder -= coin;
count++;
break;
}
}
if (reminder > 0) {
return -1;
} else {
return count;
}
}
}
但是这个代码放到Leetcode上跑,有些测试用例是不能通过。
动态规划实现
使用动态规划求解,如下面代码
java
public class Leetcode322{
public int coinChange(int[] coins, int amount) {
int[] dp = new int[amount + 1];
Arrays.fill(dp, amount + 1);
dp[0] = 0;
for(int coin : coins) {
for(int j = coin; j < amount + 1; j++) {
dp[j] = Math.min(dp[j], dp[j - coin] + 1);
}
}
int r = dp[amount];
return r > amount ? -1 : r;
}
}
哈夫曼编码
Huffman树构建过程
- 统计出现频次字符,放入优先级队列
- 每次出对两个频次最低元素(小顶堆),
- 当队列中只有一个元素时,构建Huffman树完成
java
public class Huffman{
static class Node{
Character ch;
int freq;
Node left;
Node right;
String code;
public Node(Character ch) {
this.ch = ch;
}
public Node(int freq, Node left, Node right) {
this.freq = freq;
this.left = left;
this.right = right;
}
int getFreq() {
return freq;
}
boolean isLeaf() {
return node.left == null;
}
@Override
public void toString() {
return "Node{" +
"ch=" + ch +
", freq=" + freq +
"}";
}
}
String str;
Node root;
HashMap<Character, Node> map = new HashMap<>();
public Huffman(String str) {
this.str = str;
char[] chars = str.toCharArray();
// 1.统计频次
for(char ch : chars) {
if (!map.containsKey(ch)) {
map.put(ch, new Node(ch));
} else {
Node node = map.get(ch);
node.freq++;
}
//方法引用
// Node node = map.computeIfAbsent(ch, Node :: new);
// node.frea++;
}
for(Node node : map.values()) {
System.out.println(node.toString());
}
2.构造树
PriorityQueue<Node> queue = new PriorityQueue<>(
Comparator.ComparingInt(Node::getFreq)
);
queue.offerAll(map.values());
while (queue.size() >= 2) {
Node x = queue.poll();
Node y = queue.poll();
int f = x.freq + y.freq;
queue.offer(new Node(f, x, y));
}
root = queue.poll();
System.out.println(root);
// 功能3 计算每个字符编码 //功能4 字符串编码后占几个bit
int sum = dfs(root, new StringBuilder()); // 得到字符串编码后占的bit
for(Node node : map.values()) {
System.out.printlin(node);
}
System.out.println(sum);
}
public int dfs(Node node, StringBuilder sb) {
int sum = 0;
if (node.isLeaf()) {
//编码
node.node = sb.toString();
sum = sb.length() * node.freq;
// System.out.println(sb.toString());
} else {
sum += dfs(node.left, sb.append("0"));
sum += dfs(node.right, sb.append("1"));
}
if (sb.length() > 0) {
sb.deleteCharAt(sb.length() - 1);
}
return sum;
}
public String encode() {
char[] chars = str.toCharArray();
StringBuilder sb = new StringBuilder();
for(char ch : chars) {
sb.append(map.get(ch).code);
}
return sb.toString();
}
public String decode(String str) {
int i = 0;
char[] chars = str.toCharArray();
StringBuilder sb = new StringBuilder();
Node node = root;
while (i < chars.length) {
if (!node.isLeaf()) {
if (chars[i] == '0') {
node = node.left;
} else if (chars[i] == '1'){
node = node.right;
}
i++;
}
if (node.isLeaf()) {
sb.append(node.ch);
node = root;
}
}
return sb.toString();
}
}
活动选择问题
要在一个会议室举办n个活动
- 每个活动有它们各自的起始和结束时间
- 找出时间上不冲突的活动组合,能够最充分利用会议室(举办的活动次数最多)
java
public class ActivitySelectionProblem{
static class Activity{
int index;
int start;
int end;
public Activity(int index, int start, int end) {
this.index = index;
this.start = start;
this.end = end;
}
public int getEnd() {
return end;
}
@Override
public void tostring() {
return "Activity(" + index + ")";
}
}
public static void main(String[] args) {
Activity[] activity = new Activity[]{
new Activity(0, 1, 3),
new Activity(1, 2, 4),
new Activity(2, 3, 5)
}
Arrays.sort(activity, Comparator.comparingInt(Activity::getEnd))
System.out.println(activity);
//
select(activity, activity.length);
}
public void select(Activity[] activity, int n) {
List<int[]> res = new ArrayList<>();
res.add(activity[0]);
Activity pre = res;
for(int i = 1; i < n; i++) {
Activity curr = activity[i];
if (curr.start >= pre.end) {
res.add(curr);
pre = curr;
}
}
for(Activity a : res) {
System.out.println(a);
}
}
}
Leetcode435
无重叠区间
java
public class Leetcode435{
public int eraseOverlapIntervals(int[][] intervals) {
// 根据数组中的第二个元素先升序排序
Arrays.sort(intervals, (a, b) -> a[1] - b[1]);
List<int[]> res = new ArrayList<>();
res.add(intervals[0]);
int[] pre = res.get(0);
int count = 0; // 减少个数
for(int i = 1; i < intervals.length; i++) {
int[] curr = intervals[i];
if (curr[0] < pre[1]) { // 当前的初始值小于前一个的结束值,有冲突
count++;
} else { // 只要当前的初始值大于等于前一个的结束值,则不冲突
res.add(curr);
pre = curr;
}
}
return count;
}
}
分数背包问题
- n个液体物品,有重量和价格属性
- 现取走10L物品
- 每次可以不拿,全拿,拿一部分,问最高价值是多少
java
public class FractionalKnapsackProblem{
static class Item{
int index;
int weight;
int value;
public Item(int index, int weight, int value) {
this.index = index;
this.weight = weight;
this.value = value;
}
public int perValue() {
return value / weight;
}
@Override
public void toString() {
return "Item(" + index + ")";
}
}
public static void main(String[] args) {
Item[] items = new Item[]{
new Item(0, 4, 24),
new Item(1, 8, 160),
new Item(2, 2, 4000),
new Item(3, 6, 108),
new Item(4, 1, 4000),
}
select(items, 10);
}
public int select(Item[] items, int n) {
Arrays.sort(items, Comparator.comparingInt(Item::preValue).reverse());
int sum = 0;
for(int i = 0; i < items.length; i++) {
Item curr = items[i];
if (n >= curr.weight) {
n -= curr.weight;
sum += curr.value;
} else {
sum += curr.perValue() * n;
break;
}
}
return sum;
}
}
0-1背包问题
- n个物体都是固体,有重量和价值
- 现取走不超过10克物品
- 每次可以不拿或者全拿,问最高价值是多少
java
public class FractionalKnapsackProblem{
static class Item{
int index;
int weight;
int value;
public Item(int index, int weight, int value) {
this.index = index;
this.weight = weight;
this.value = value;
}
public int perValue() {
return value / weight;
}
@Override
public void toString() {
return "Item(" + index + ")";
}
}
public static void main(String[] args) {
Item[] items = new Item[]{
new Item(0, 1, 1000000),
new Item(1, 4, 1600),
new Item(2, 8, 2400),
new Item(3, 5, 30),
}
select(items, 10);
}
public int select(Item[] items, int n) {
Arrays.sort(items, Comparator.comparingInt(Item::preValue).reverse());
int sum = 0;
for(int i = 0; i < items.length; i++) {
Item curr = items[i];
if (n >= curr.weight) {
n -= curr.weight;
sum += curr.value;
}
}
return sum;
}
}
得到的结果,最大价值结果是:1001630
,实际上选择钻石 和红宝石 得到的价值结果 1002400