問(wèn)題描述
給定n個(gè)設(shè)備和m個(gè)顧客
- 每個(gè)顧客有一定的需求(
demand)需要被設(shè)備滿足,記為d[i] - 每個(gè)設(shè)備有不同的滿足顧客需求的容量(
capacity),記為c[j] - 每個(gè)設(shè)備有不同的開(kāi)啟費(fèi)用(
opening cost),記為o[j] - 不同顧客分配到不同設(shè)備會(huì)有不同的分配費(fèi)用(
assignment cost),記為a[i][j]
現(xiàn)在所要達(dá)成的目的是要滿足所有顧客的要求,同時(shí)使opening cost和assignment cost的總和(total cost)最小。問(wèn):
- 最終哪些設(shè)備要開(kāi)啟?
- 對(duì)所有顧客的最終分配方案是怎么樣的?
分析
m個(gè)顧客會(huì)有n中不同的選擇,如果我們要窮盡所有的可能,時(shí)間復(fù)雜度將會(huì)是 O(n^m),顯然這不是一個(gè)很好的選擇;所以我們應(yīng)該采取更優(yōu)的做法。
首先先寫(xiě)出主程序框架:
int main(){
for(int testCase = 1; testCase <= 71; testCase ++){
// Data declaration
int n, m; // n facilities, m customers
int* d; // demands of customers
int* c; // capacities of facilities
int* o; // opening cost of facilities
int** a; // assignment cost for each (cus, fac)
bool* open_state; // result: open state of facilities
int* assign_state; // result: assignment of cus->fac
// Load files
ifstream infile;
char filename[32];
sprintf_s(filename, "Instances\\p%d", testCase);
infile.open(filename, ios::in);
if(!infile.is_open ()){
cout << "Open file failure: " << filename << endl;
continue;
}
infile >> n >> m;
c = new int[n];
o = new int[n];
for(int i = 0; i < n; ++i){
infile >> c[i] >> o[i];
}
d = new int[m];
a = new int*[m];
for(int i = 0; i < m; ++i){
a[i] = new int[n];
infile >> d[i];
}
for(int i = 0; i < m; ++i){
for(int e = 0; e < n; ++e){
infile >> a[i][e];
}
}
infile.close();
open_state = new bool[n];
for(int i = 0; i < n; ++i) open_state[i] = false;
assign_state = new int[m];
// Calculate
int result = Algorithm(n,m,d,c,o,a,open_state,assign_state);
// Output Results
sprintf_s(filename, "results\\greedy\\p%d", testCase);
ofstream outfile(filename);
if(result == -1){
outfile << "No result" << endl;
continue;
}
outfile << result << endl;
for(int i = 0; i < n; ++i)
outfile << open_state[i] << ' ';
outfile << endl;
for(int i = 0; i < m; ++i)
outfile << assign_state[i] << ' ';
outfile << endl;
// Release space
for(int i = 0; i < m; ++i){
delete[] a[i];
}
delete[] c, o, open_state, assign_state;
}
}
貪心算法
首先比較容易想到的是貪心算法。按照逐個(gè)顧客分配的辦法,每次為顧客進(jìn)行分配的時(shí)候,需要找到為他分配設(shè)備的最小花費(fèi),這有可能是:
- 無(wú)需重新分配新的設(shè)備,花費(fèi)總額為
a[i][j] - 分配新的設(shè)備,花費(fèi)總額為
c[j]+a[i][j]
值得一提的是,貪心算法大概率不是最優(yōu)算法,甚至有可能會(huì)有在原本有解的情況下算出無(wú)解的答案。但其優(yōu)點(diǎn)是會(huì)比較快。
#include <iostream>
#include "greedy.hpp"
using namespace std;
int Greedy(int n, int m, int *d, int *c, int *o, int **a, bool *os, int* as){
int tc = 0;
for(int curCus = 0; curCus != m; ++curCus){
int minCost = INT32_MAX;
int curChoice = -1;
for(int curFac = 0; curFac != n; curFac++){
if(d[curCus] <= c[curFac]){
int tempCost = a[curCus][curFac];
tempCost += os[curFac] ? 0 : o[curFac];
if(tempCost < minCost){
minCost = tempCost;
curChoice = curFac;
}
}
}
if(curChoice == -1)
return -1;
os[curChoice] = true;
as[curCus] = curChoice;
c[curChoice] -= d[curCus];
tc += minCost;
}
return tc;
}
運(yùn)算結(jié)果:
| problem | totalCost | time |
|---|---|---|
| p1 | 18076 | 0 |
| p2 | 12899 | 0 |
| p3 | 18441 | 0 |
| p4 | 22698 | 0 |
| p5 | 17665 | 0 |
| p6 | 12930 | 0 |
| p7 | 20782 | 0 |
| p8 | 23191 | 0 |
| p9 | 15685 | 0 |
| p10 | 11172 | 0 |
| p11 | 20501 | 0 |
| p12 | 21717 | 0 |
| p13 | 15862 | 0 |
| p14 | 12419 | 0 |
| p15 | 16778 | 0 |
| p16 | 23752 | 0 |
| p17 | 15009 | 0 |
| p18 | 12005 | 0 |
| p19 | 18851 | 0 |
| p20 | 22479 | 0 |
| p21 | 15252 | 0 |
| p22 | 11524 | 0 |
| p23 | 17018 | 0 |
| p24 | 22259 | 0 |
| p25 | 16138 | 0 |
| p26 | 17971 | 0 |
| p27 | 18773 | 0 |
| p28 | 25728 | 0 |
| p29 | 19800 | 0 |
| p30 | 17967 | 0 |
具體的分配結(jié)果可見(jiàn)github倉(cāng)庫(kù)
貪心算法V1
在上面的算法中,我們?nèi)?code>p1的結(jié)果為例:
18076
1 1 1 1 0 1 1 0 1 1
1 1 1 8 1 9 8 9 8 9 9 9 9 9 9 9 5 9 9 5 5 5 5 5 5 5 5 5 5 0 3 0 5 0 3 0 0 3 0 0 0 3 3 3 8 3 3 2 2 6
我們可以發(fā)現(xiàn)使用貪心算法,顧客傾向于不開(kāi)新的設(shè)備,因?yàn)樾麻_(kāi)設(shè)備要加上開(kāi)設(shè)備的費(fèi)用。可以理解到,每個(gè)顧客都把開(kāi)某個(gè)設(shè)備的費(fèi)用算到了自己頭上,就像滾燙的山芋不愿意握在手上;但實(shí)際上開(kāi)設(shè)備的費(fèi)用最終是落到總費(fèi)用上,所以我們可以試一下顧客不去考慮開(kāi)工廠的費(fèi)用,到最后再去算。
代碼和上面類似V1版本的類似,只有一些細(xì)微的修改
// 刪去這行
tempCost += os[curFac] ? 0 : o[curFac];
// 增加這個(gè)
if(!os[curChoice]){
minCost += o[curChoice];
os[curChoice] = true;
}
結(jié)果:
| problem | totalCost | time |
|---|---|---|
| p1 | 10355 | 0 |
| p2 | 9041 | 0 |
| p3 | 11041 | 0 |
| p4 | 13041 | 0 |
| p5 | 10827 | 0 |
| p6 | 9513 | 0 |
| p7 | 11513 | 0 |
| p8 | 13513 | 0 |
| p9 | 10355 | 0 |
| p10 | 9041 | 0 |
| p11 | 11041 | 0 |
| p12 | 13041 | 0 |
| p13 | 11915 | 0 |
| p14 | 9426 | 0 |
| p15 | 13026 | 0 |
| p16 | 16626 | 0 |
| p17 | 11915 | 0 |
| p18 | 9426 | 0 |
| p19 | 13026 | 0 |
| p20 | 16626 | 0 |
| p21 | 11915 | 0 |
| p22 | 9426 | 0 |
| p23 | 13026 | 0 |
| p24 | 16626 | 0 |
| p25 | 16655 | 0 |
| p26 | 14125 | 0 |
| p27 | 18925 | 0 |
| p28 | 23725 | 0 |
| p29 | 17487 | 0 |
| p30 | 14750 | 0 |