• bzoj3698


    显然是有源有汇有下界最大流,不刷不知道,一刷吓一跳
    发现了我之前求有源有汇有下界最大流的错误,具体见我那篇介绍有下界的网络流的解题报告(bzoj2502),已经更正

      1 const inf=10000007;
      2 type node=record
      3        po,next,flow:longint;
      4      end;
      5 
      6 var e:array[0..40010] of node;
      7     pre,p,numh,h,cur,d:array[0..220] of longint;
      8     a:array[0..110,0..110] of double;
      9     len,tot,i,j,n,s,t,ss,tt:longint;
     10 
     11 function min(a,b:longint):longint;
     12   begin
     13     if a>b then exit(b) else exit(a);
     14   end;
     15 
     16 procedure add(x,y,f:longint);
     17   begin
     18     inc(len);
     19     e[len].po:=y;
     20     e[len].flow:=f;
     21     e[len].next:=p[x];
     22     p[x]:=len;
     23   end;
     24 
     25 procedure build(x,y,f:longint);
     26   begin
     27     add(x,y,f);
     28     add(y,x,0);
     29   end;
     30 
     31 function sap(s,t:longint):longint;
     32   var u,neck,tmp,i,j,q:longint;
     33   begin
     34     fillchar(numh,sizeof(numh),0);
     35     fillchar(h,sizeof(h),0);
     36     for i:=0 to tt do
     37       cur[i]:=p[i];
     38     u:=s;
     39     sap:=0;
     40     numh[0]:=tt+1;
     41     neck:=inf;
     42     while h[s]<tt+1 do
     43     begin
     44       d[u]:=neck;
     45       i:=cur[u];
     46       while i<>-1 do
     47       begin
     48         j:=e[i].po;
     49         if (e[i].flow>0) and (h[u]=h[j]+1) then
     50         begin
     51           neck:=min(neck,e[i].flow);
     52           cur[u]:=i;
     53           pre[j]:=u;
     54           u:=j;
     55           if u=t then
     56           begin
     57             sap:=sap+neck;
     58             while u<>s do
     59             begin
     60               u:=pre[u];
     61               j:=cur[u];
     62               dec(e[j].flow,neck);
     63               inc(e[j xor 1].flow,neck);
     64             end;
     65             neck:=inf;
     66           end;
     67           break;
     68         end;
     69         i:=e[i].next;
     70       end;
     71       if i=-1 then
     72       begin
     73         dec(numh[h[u]]);
     74         if numh[h[u]]=0 then exit;
     75         q:=-1;
     76         tmp:=tt;
     77         i:=p[u];
     78         while i<>-1 do
     79         begin
     80           j:=e[i].po;
     81           if e[i].flow>0 then
     82             if h[j]<tmp then
     83             begin
     84               q:=i;
     85               tmp:=h[j];
     86             end;
     87           i:=e[i].next;
     88         end;
     89         cur[u]:=q;
     90         h[u]:=tmp+1;
     91         inc(numh[h[u]]);
     92         if u<>s then
     93         begin
     94           u:=pre[u];
     95           neck:=d[u];
     96         end;
     97       end;
     98     end;
     99   end;
    100 
    101 begin
    102   len:=-1;
    103   fillchar(p,sizeof(p),255);
    104   readln(n);
    105   dec(n);
    106   s:=0; t:=2*n+1; ss:=2*n+2; tt:=2*n+3;
    107   for i:=1 to n+1 do
    108     for j:=1 to n+1 do
    109       read(a[i,j]);
    110   for i:=1 to n do
    111   begin
    112     if a[i,n+1]<>trunc(a[i,n+1]) then
    113       build(s,i,1);
    114     d[i]:=d[i]+trunc(a[i,n+1]);
    115     d[s]:=d[s]-trunc(a[i,n+1]);
    116     if a[n+1,i]<>trunc(a[n+1,i]) then
    117       build(i+n,t,1);
    118     d[i+n]:=d[i+n]-trunc(a[n+1,i]);
    119     d[t]:=d[t]+trunc(a[n+1,i]);
    120   end;
    121   for i:=1 to n do
    122     for j:=1 to n do
    123     begin
    124       if trunc(a[i,j])<>a[i,j] then build(i,j+n,1);
    125       d[i]:=d[i]-trunc(a[i,j]);
    126       d[j+n]:=d[j+n]+trunc(a[i,j]);
    127     end;
    128   for i:=s to t do
    129     if d[i]>0 then
    130     begin
    131       build(ss,i,d[i]);
    132       tot:=tot+d[i];
    133     end
    134     else if d[i]<0 then build(i,tt,-d[i]);
    135   build(t,s,inf);
    136   if sap(ss,tt)<>tot then writeln('No')
    137   else writeln(sap(s,t)*3);
    138 end.
    View Code
  • 相关阅读:
    在pytorch中使用可视化工具visdom
    3"GAN-GAN实战:GD的实现
    31:GAN-js散度的缺陷以及解决办法:WGAN的提出
    31:GAN-DCGAN(deconvolution GAN逆卷积操作)以及JS散度的缺陷
    31:GAN-纳什均衡
    31:GAN-GAN原理
    30:无监督学习-自编码网络的变种
    30:无监督学习-自编码(Auto-Encoder简称AE)原理
    如何打印网络的参数量
    29:迁移学习(当数据集比较小时,可以使用迁移学习达到一个更好的效果)
  • 原文地址:https://www.cnblogs.com/phile/p/4472957.html
Copyright © 2020-2023  润新知