首先比 h 1 h_1 h1小的肯定没用，直接无视
然后考虑合并的顺序
①在无限制的情况下，合并多个不如一个一个合并
a < b < c a<b<c a<b<c时， a + b 2 + c 2 > a + b + c 3 {{a+b \over 2}+c\over 2}>{{a+b+c}\over 3} 22a+b+c>3a+b+c
②先合并小的比先合并大的优
整理一下，就是把 > h 1 > h_1 >h1的从小到大排序，然后每次把当前的和序列中靠前的若干个合并
显然是个斜率优化的模型
设 f k , n f_{k,n} fk,n表示前 n n n个合并 k k k次的最大值
f k , n = max 1 ≤ i < n f i + S n − S i n − i + 1 f_{k,n}=\max_{1\leq i<n}\frac{f_i+S_n-S_i}{n-i+1} fk,n=1≤i<nmaxn−i+1fi+Sn−Si
推一下式子发现很恶心
但注意到这个式子本身就是斜率的形式
可以看成所有 ( i − 1 , S i − f i ) (i-1,S_i-f_i) (i−1,Si−fi)到 ( n , S n ) (n,S_n) (n,Sn)的最大斜率
意识流一波，维护一个下凸壳，每次最优决策是过 ( n , S n ) (n,S_n) (n,Sn)的一条与凸壳切于端点的直线，可以三分一下……
注意到 h h h严格单调增，所以 ( n , S n ) (n,S_n) (n,Sn)本身就组成了下凸壳
对于两个决策 k 1 , k 2 k_1,k_2 k1,k2,设 i i i是 k 2 k_2 k2优于 k 1 k_1 k1的 ( i , S i ) (i,S_i) (i,Si),即 ( k 1 − 1 , S k 1 − f k 1 ) , ( k 2 − 1 , S k 2 − f k 2 ) , ( i , S i ) (k_1-1,S_{k_1}-f_{k_1}),(k_2-1,S_{k_2}-f_{k_2}),(i,S_i) (k1−1,Sk1−fk1),(k2−1,Sk2−fk2),(i,Si)往上翘
如图，
k
i
−
1
,
i
>
k
k
2
,
i
k_{i-1,i}>k_{k_2,i}
ki−1,i>kk2,i,又因为
k
i
,
i
+
1
>
k
i
−
1
,
i
k_{i,i+1}>k_{i-1,i}
ki,i+1>ki−1,i,所以对于
i
+
1
i+1
i+1
k
2
k_2
k2仍然更优
所以可以把 k 1 k_1 k1弹掉
这样可以做到 O ( n k p ) O(nkp) O(nkp) 这里因为最多只能取 n − 1 n-1 n−1次合并，所以 n , k n,k n,k是同阶的
然后可以用低精度算 d p dp dp并记录决策，最后用高精度算一次 复杂度 O ( n k + n p ) O(nk+np) O(nk+np)
// This is an empty program with decimal lib
#include <cstdlib>
#include <cstring>
#include <string>
#include <cctype>
#include <cstdio>
// ---------- decimal lib start ----------
const int PREC = 3500;
class Decimal {
public:
Decimal();
Decimal(const std::string &s);
Decimal(const char *s);
Decimal(int x);
Decimal(long long x);
Decimal(long double x);
bool is_zero() const;
// p (p > 0) is the number of digits after the decimal point
std::string to_string(int p) const;
long double to_double() const;
friend Decimal operator + (const Decimal &a, const Decimal &b);
friend Decimal operator + (const Decimal &a, int x);
friend Decimal operator + (int x, const Decimal &a);
friend Decimal operator + (const Decimal &a, long long x);
friend Decimal operator + (long long x, const Decimal &a);
friend Decimal operator + (const Decimal &a, long double x);
friend Decimal operator + (long double x, const Decimal &a);
friend Decimal operator - (const Decimal &a, const Decimal &b);
friend Decimal operator - (const Decimal &a, int x);
friend Decimal operator - (int x, const Decimal &a);
friend Decimal operator - (const Decimal &a, long long x);
friend Decimal operator - (long long x, const Decimal &a);
friend Decimal operator - (const Decimal &a, long double x);
friend Decimal operator - (long double x, const Decimal &a);
friend Decimal operator * (const Decimal &a, int x);
friend Decimal operator * (int x, const Decimal &a);
friend Decimal operator / (const Decimal &a, int x);
friend bool operator < (const Decimal &a, const Decimal &b);
friend bool operator > (const Decimal &a, const Decimal &b);
friend bool operator <= (const Decimal &a, const Decimal &b);
friend bool operator >= (const Decimal &a, const Decimal &b);
friend bool operator == (const Decimal &a, const Decimal &b);
friend bool operator != (const Decimal &a, const Decimal &b);
Decimal & operator += (int x);
Decimal & operator += (long long x);
Decimal & operator += (long double x);
Decimal & operator += (const Decimal &b);
Decimal & operator -= (int x);
Decimal & operator -= (long long x);
Decimal & operator -= (long double x);
Decimal & operator -= (const Decimal &b);
Decimal & operator *= (int x);
Decimal & operator /= (int x);
friend Decimal operator - (const Decimal &a);
// These can't be called
friend Decimal operator * (const Decimal &a, long double x);
friend Decimal operator * (long double x, const Decimal &a);
friend Decimal operator / (const Decimal &a, long double x);
Decimal & operator *= (long double x);
Decimal & operator /= (long double x);
private:
static const int len = PREC / 9 + 1;
static const int mo = 1000000000;
static void append_to_string(std::string &s, long long x);
bool is_neg;
long long integer;
int data[len];
void init_zero();
void init(const char *s);
};
Decimal::Decimal() {
this->init_zero();
}
Decimal::Decimal(const char *s) {
this->init(s);
}
Decimal::Decimal(const std::string &s) {
this->init(s.c_str());
}
Decimal::Decimal(int x) {
this->init_zero();
if (x < 0) {
is_neg = true;
x = -x;
}
integer = x;
}
Decimal::Decimal(long long x) {
this->init_zero();
if (x < 0) {
is_neg = true;
x = -x;
}
integer = x;
}
Decimal::Decimal(long double x) {
this->init_zero();
if (x < 0) {
is_neg = true;
x = -x;
}
integer = (long long)x;
x -= integer;
for (int i = 0; i < len; i++) {
x *= mo;
if (x < 0) x = 0;
data[i] = (int)x;
x -= data[i];
}
}
void Decimal::init_zero() {
is_neg = false;
integer = 0;
memset(data, 0, len * sizeof(int));
}
bool Decimal::is_zero() const {
if (integer) return false;
for (int i = 0; i < len; i++) {
if (data[i]) return false;
}
return true;
}
void Decimal::init(const char *s) {
this->init_zero();
is_neg = false;
integer = 0;
// find the first digit or the negative sign
while (*s != 0) {
if (*s == '-') {
is_neg = true;
++s;
break;
} else if (*s >= 48 && *s <= 57) {
break;
}
++s;
}
// read the integer part
while (*s >= 48 && *s <= 57) {
integer = integer * 10 + *s - 48;
++s;
}
// read the decimal part
if (*s == '.') {
int pos = 0;
int x = mo / 10;
++s;
while (pos < len && *s >= 48 && *s <= 57) {
data[pos] += (*s - 48) * x;
++s;
x /= 10;
if (x == 0) {
++pos;
x = mo / 10;
}
}
}
}
void Decimal::append_to_string(std::string &s, long long x) {
if (x == 0) {
s.append(1, 48);
return;
}
char _[30];
int cnt = 0;
while (x) {
_[cnt++] = x % 10;
x /= 10;
}
while (cnt--) {
s.append(1, _[cnt] + 48);
}
}
std::string Decimal::to_string(int p) const {
std::string ret;
if (is_neg && !this->is_zero()) {
ret = "-";
}
append_to_string(ret, this->integer);
ret.append(1, '.');
for (int i = 0; i < len; i++) {
// append data[i] as "%09d"
int x = mo / 10;
int tmp = data[i];
while (x) {
ret.append(1, 48 + tmp / x);
tmp %= x;
x /= 10;
if (--p == 0) {
break;
}
}
if (p == 0) break;
}
if (p > 0) {
ret.append(p, '0');
}
return ret;
}
long double Decimal::to_double() const {
long double ret = integer;
long double k = 1.0;
for (int i = 0; i < len; i++) {
k /= mo;
ret += k * data[i];
}
if (is_neg) {
ret = -ret;
}
return ret;
}
bool operator < (const Decimal &a, const Decimal &b) {
if (a.is_neg != b.is_neg) {
return a.is_neg && (!a.is_zero() || !b.is_zero());
} else if (!a.is_neg) {
// a, b >= 0
if (a.integer != b.integer) {
return a.integer < b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] < b.data[i];
}
}
return false;
} else {
// a, b <= 0
if (a.integer != b.integer) {
return a.integer > b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] > b.data[i];
}
}
return false;
}
}
bool operator > (const Decimal &a, const Decimal &b) {
if (a.is_neg != b.is_neg) {
return !a.is_neg && (!a.is_zero() || !b.is_zero());
} else if (!a.is_neg) {
// a, b >= 0
if (a.integer != b.integer) {
return a.integer > b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] > b.data[i];
}
}
return false;
} else {
// a, b <= 0
if (a.integer != b.integer) {
return a.integer < b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] < b.data[i];
}
}
return false;
}
}
bool operator <= (const Decimal &a, const Decimal &b) {
if (a.is_neg != b.is_neg) {
return a.is_neg || (a.is_zero() && b.is_zero());
} else if (!a.is_neg) {
// a, b >= 0
if (a.integer != b.integer) {
return a.integer < b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] < b.data[i];
}
}
return true;
} else {
// a, b <= 0
if (a.integer != b.integer) {
return a.integer > b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] > b.data[i];
}
}
return true;
}
}
bool operator >= (const Decimal &a, const Decimal &b) {
if (a.is_neg != b.is_neg) {
return !a.is_neg || (a.is_zero() && b.is_zero());
} else if (!a.is_neg) {
// a, b >= 0
if (a.integer != b.integer) {
return a.integer > b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] > b.data[i];
}
}
return true;
} else {
// a, b <= 0
if (a.integer != b.integer) {
return a.integer < b.integer;
}
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) {
return a.data[i] < b.data[i];
}
}
return true;
}
}
bool operator == (const Decimal &a, const Decimal &b) {
if (a.is_zero() && b.is_zero()) return true;
if (a.is_neg != b.is_neg) return false;
if (a.integer != b.integer) return false;
for (int i = 0; i < Decimal::len; i++) {
if (a.data[i] != b.data[i]) return false;
}
return true;
}
bool operator != (const Decimal &a, const Decimal &b) {
return !(a == b);
}
Decimal & Decimal::operator += (long long x) {
if (!is_neg) {
if (integer + x >= 0) {
integer += x;
} else {
bool last = false;
for (int i = len - 1; i >= 0; i--) {
if (last || data[i]) {
data[i] = mo - data[i] - last;
last = true;
} else {
last = false;
}
}
integer = -x - integer - last;
is_neg = true;
}
} else {
if (integer - x >= 0) {
integer -= x;
} else {
bool last = false;
for (int i = len - 1; i >= 0; i--) {
if (last || data[i]) {
data[i] = mo - data[i] - last;
last = true;
} else {
last = false;
}
}
integer = x - integer - last;
is_neg = false;
}
}
return *this;
}
Decimal & Decimal::operator += (int x) {
return *this += (long long)x;
}
Decimal & Decimal::operator -= (int x) {
return *this += (long long)-x;
}
Decimal & Decimal::operator -= (long long x) {
return *this += -x;
}
Decimal & Decimal::operator /= (int x) {
if (x < 0) {
is_neg ^= 1;
x = -x;
}
int last = integer % x;
integer /= x;
for (int i = 0; i < len; i++) {
long long tmp = 1LL * last * mo + data[i];
data[i] = tmp / x;
last = tmp - 1LL * data[i] * x;
}
if (is_neg && integer == 0) {
int i;
for (i = 0; i < len; i++) {
if (data[i] != 0) {
break;
}
}
if (i == len) {
is_neg = false;
}
}
return *this;
}
Decimal & Decimal::operator *= (int x) {
if (x < 0) {
is_neg ^= 1;
x = -x;
} else if (x == 0) {
init_zero();
return *this;
}
int last = 0;
for (int i = len - 1; i >= 0; i--) {
long long tmp = 1LL * data[i] * x + last;
last = tmp / mo;
data[i] = tmp - 1LL * last * mo;
}
integer = integer * x + last;
return *this;
}
Decimal operator - (const Decimal &a) {
Decimal ret = a;
// -0 = 0
if (!ret.is_neg && ret.integer == 0) {
int i;
for (i = 0; i < Decimal::len; i++) {
if (ret.data[i] != 0) break;
}
if (i < Decimal::len) {
ret.is_neg = true;
}
} else {
ret.is_neg ^= 1;
}
return ret;
}
Decimal operator + (const Decimal &a, int x) {
Decimal ret = a;
return ret += x;
}
Decimal operator + (int x, const Decimal &a) {
Decimal ret = a;
return ret += x;
}
Decimal operator + (const Decimal &a, long long x) {
Decimal ret = a;
return ret += x;
}
Decimal operator + (long long x, const Decimal &a) {
Decimal ret = a;
return ret += x;
}
Decimal operator - (const Decimal &a, int x) {
Decimal ret = a;
return ret -= x;
}
Decimal operator - (int x, const Decimal &a) {
return -(a - x);
}
Decimal operator - (const Decimal &a, long long x) {
Decimal ret = a;
return ret -= x;
}
Decimal operator - (long long x, const Decimal &a) {
return -(a - x);
}
Decimal operator * (const Decimal &a, int x) {
Decimal ret = a;
return ret *= x;
}
Decimal operator * (int x, const Decimal &a) {
Decimal ret = a;
return ret *= x;
}
Decimal operator / (const Decimal &a, int x) {
Decimal ret = a;
return ret /= x;
}
Decimal operator + (const Decimal &a, const Decimal &b) {
if (a.is_neg == b.is_neg) {
Decimal ret = a;
bool last = false;
for (int i = Decimal::len - 1; i >= 0; i--) {
ret.data[i] += b.data[i] + last;
if (ret.data[i] >= Decimal::mo) {
ret.data[i] -= Decimal::mo;
last = true;
} else {
last = false;
}
}
ret.integer += b.integer + last;
return ret;
} else if (!a.is_neg) {
// a - |b|
return a - -b;
} else {
// b - |a|
return b - -a;
}
}
Decimal operator - (const Decimal &a, const Decimal &b) {
if (!a.is_neg && !b.is_neg) {
if (a >= b) {
Decimal ret = a;
bool last = false;
for (int i = Decimal::len - 1; i >= 0; i--) {
ret.data[i] -= b.data[i] + last;
if (ret.data[i] < 0) {
ret.data[i] += Decimal::mo;
last = true;
} else {
last = false;
}
}
ret.integer -= b.integer + last;
return ret;
} else {
Decimal ret = b;
bool last = false;
for (int i = Decimal::len - 1; i >= 0; i--) {
ret.data[i] -= a.data[i] + last;
if (ret.data[i] < 0) {
ret.data[i] += Decimal::mo;
last = true;
} else {
last = false;
}
}
ret.integer -= a.integer + last;
ret.is_neg = true;
return ret;
}
} else if (a.is_neg && b.is_neg) {
// a - b = (-b) - (-a)
return -b - -a;
} else if (a.is_neg) {
// -|a| - b
return -(-a + b);
} else {
// a - -|b|
return a + -b;
}
}
Decimal operator + (const Decimal &a, long double x) {
return a + Decimal(x);
}
Decimal operator + (long double x, const Decimal &a) {
return Decimal(x) + a;
}
Decimal operator - (const Decimal &a, long double x) {
return a - Decimal(x);
}
Decimal operator - (long double x, const Decimal &a) {
return Decimal(x) - a;
}
Decimal & Decimal::operator += (long double x) {
*this = *this + Decimal(x);
return *this;
}
Decimal & Decimal::operator -= (long double x) {
*this = *this - Decimal(x);
return *this;
}
Decimal & Decimal::operator += (const Decimal &b) {
*this = *this + b;
return *this;
}
Decimal & Decimal::operator -= (const Decimal &b) {
*this = *this - b;
return *this;
}
// ---------- decimal lib end ----------
inline int read()
{
int ans=0;
char c=getchar();
while (!isdigit(c)) c=getchar();
while (isdigit(c)) ans=(ans<<3)+(ans<<1)+(c^48),c=getchar();
return ans;
}
#include <algorithm>
#include <iostream>
#define MAXN 8005
using namespace std;
int s[MAXN],tot;
long double f[2][MAXN];
short key[MAXN][MAXN];
int q[MAXN],head,tail;
inline long double calc(int n,int i,int t){return (s[n]-s[i]+f[t^1][i])/(n-i+1);}
inline bool cmp(int i,int j,int k,int t){return (s[i]-f[t^1][i]-s[j]+f[t^1][j])*(i-k)<(s[i]-f[t^1][i]-s[k]+f[t^1][k])*(i-j);}
Decimal getans(int k,int n)
{
if (!k) return s[1];
return (s[n]-s[key[k][n]]+getans(k-1,key[k][n]))/(n-key[k][n]+1);
}
long double ans=0;
int pos;
int main()
{
int n,k,p;
n=read(),k=read(),p=read();
s[tot=1]=read();
for (int i=1;i<n;i++)
{
int t=read();
if (t>s[1]) s[++tot]=t;
}
n=tot;
sort(s+1,s+n+1);
if (k>=n) k=n-1;
for (int i=1;i<=n;i++) s[i]+=s[i-1],f[0][i]=s[1];
f[1][1]=s[1];
for (int i=1,t=1;i<=k;i++,t^=1)
{
head=1,tail=1;
q[1]=1;
for (int j=2;j<=n;j++)
{
while (head<tail&&calc(j,q[head],t)<calc(j,q[head+1],t)) ++head;
f[t][j]=calc(j,key[i][j]=q[head],t);
while (head<tail&&cmp(j,q[tail],q[tail-1],t)) --tail;
q[++tail]=j;
}
}
cout<<getans(k,n).to_string(2*p);
return 0;
}
423
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
