[算法]01分数规划

最新推荐文章于 2022-06-07 15:30:56 发布

Aurora141592

最新推荐文章于 2022-06-07 15:30:56 发布

阅读量66

点赞数

分类专栏：优化数论

本文链接：https://blog.csdn.net/wulalalawu/article/details/111858293

版权

数论同时被 2 个专栏收录

37 篇文章 1 订阅

订阅专栏

优化

21 篇文章 0 订阅

订阅专栏

https://blog.csdn.net/hzoi_ztx/article/details/54898323
http://www.51nod.com/Challenge/Problem.html#!#problemId=1257
给定n个二元组((value_i, cost_i))，(value_i)是选择此二元组获得的价值（非负），(cost_i)是选择此二元组付出的代价（非负），设(x_i (x_i \in{{0,1}} ))代表第i个二元组的选与不选，最大(小)化下式
$\frac {\sum {value_i · x_i} } {\sum {cost_i · x_i} }$
如何解决这个问题呢？化简可以得到
$\sum {cost_i · x_i} = \sum {value_i · x_i}$
二分r，选前k个求(value_i – r · cost_i)的和，大于eps说明r还能再大。

struct e{
    int v, c;
    double r;
}a[maxn];
bool cmp(e a, e b){return a.r > b.r;}
int main()
{
    int n, k;
    n = in(); k = in();
    for (int i = 0; i < n; ++i){
        a[i].c = in(); a[i].v = in();
    }
    double l = 0, r = 50001, ans;
    for (int i = 0; i < 32; ++i){
        double m = (l + r) / 2;
        for (int j = 0; j < n; ++j) a[j].r = a[j].v - m * a[j].c;
        sort(a, a + n, cmp);
        double sum = 0;
        for (int j = 0; j < k; ++j) sum += a[j].r;
        if (sum <= eps){
            ans = l;
            r = m;
        }else{
            l = m;
        }
    }
    for (int i = 0; i < n; ++i) a[i].r = a[i].v - ans * a[i].c;
    sort(a, a + n, cmp);
    ll sa = 0, sb = 0;
    for (int i = 0; i < k; ++i){
        sa += a[i].v, sb += a[i].c;
    }
    ll c = gcd(sa, sb);
    sa /= c, sb /= c;
    out(sa); putchar('/'); out(sb);
    return 0;
}