TDL的YC牌

传说中的置换群?反正不懂。我的思路竟然是对的,可是为何只有20分?

(1)尼玛每行数据输出后回车不打!

(2)写gcd函数脑残把a mod b写成a-b,大大减慢速度…

(3)看标程才想到用快速幂,第一次知道置换也可以快速幂。

(4)大小姐啊麻烦你下次看数据范围别把0的个数给数错的……

果然什么题都要见识一下。跑数据的时候速度仍然堪忧(最后三个点),可能是电脑问题,也可能还能优化?

program yc;
const maxn=100000+10;
      q=1e-6;
var t1,t2,t3,jn,p:int64;
    tt3:real;
    a,b,c,d,base,baset,ans,anst:array[1..maxn] of longint;
    i,t,n,jt,x,k,j:longint;
function gcd(a,b:int64):integer;
var t:int64;
begin
  if b>a then
    begin
      t:=a;a:=b;b:=t;
    end;
  if (a=0) or (b=0) then exit(1);
  if a mod b=0 then gcd:=b else gcd:=gcd(b,a mod b);
end;

function eq(a:real;b:int64):boolean;
begin
  if abs(a-b)<=q then exit(true) else exit(false);
end;

begin
  assign(input,'yc10.in');reset(input);
  assign(output,'yc10.out');rewrite(output);
  readln(t);
  for i:=1 to t do
    begin
      readln(n,t1,t2);
      jn:=1;
      for j:=1 to n do
        read(b[j]);
      for j:=1 to n do
        begin
          jt:=1;x:=b[j];
          while x<>j do
            begin
              x:=b[x];
              inc(jt);
            end;
          jn:=jn div gcd(jn,jt) * jt;
        end;
      {for j:=1 to n do a[j]:=j;
      fillchar(jp,sizeof(jp),$7f);
      c:=a;
      for j:=1 to 30 do
        begin
          for k:=1 to n do
            d[k]:=c[b[k]];
          c:=d;
          for k:=1 to n do
            if (a[k]=d[k]) and (k<jp[k]) then jp[k]:=j;
        end;}
      {for j:=1 to n do
        jn:=jn*jt div gcd(jn,jt); }
      for k:=-t2 to t2 do
        begin
          tt3:=(t1+k*jn)/t2;
          if (eq(tt3,round(tt3))) and (tt3>0) then break;
        end;
      t3:=round(tt3);
      for j:=1 to n do
        a[j]:=j;
      for j:=1 to n do ans[j]:=j;
      for j:=1 to n do anst[j]:=j;
      for j:=1 to n do base[j]:=b[j];
      p:=t3;
      while p>0 do
      //for p:=1 to t3 do
        begin
          {for k:=1 to n do
            begin
              //c[k]:=b[a[k]];
              c[b[k]]:=a[k];
            end;
          a:=c;
          inc(p);}
          if p mod 2=1 then
            for j:=1 to n do anst[j]:=ans[base[j]];
          ans:=anst;
          for j:=1 to n do baset[j]:=base[base[j]];
          base:=baset;
          p:=p div 2;
        end;
      {for j:=1 to n do
        c[a[j]]:=j;}
      for j:=1 to n do
        write(ans[j],' ');
      writeln;
    end;
  close(input);close(output);
end.
yc

TDL的幼儿园

昨天又4个人AC,果然我是弱渣,根本看不出这是网络流模型- -!今天学习了Dinic算法,为何感觉比之前写的更简短且容易理解?(咳咳之前那次写的手工栈)

对题目进行网络流建模,源点(编号0)向每个小朋友连一条边,容量为Ci,所有糖向汇点(编号m+n+1)连一条边,容量为Vi,小朋友和它想要的糖之间连一条容量为inf的边。(要注意inf最好不要设置成maxlongint,否则万一它的flow为负,两者一加就超过范围了,结果坑爹的Lazarus还不报错,传进去一个接近-maxlongint的值,当然这有可能是初始化一开始出错造成的。最后我inf设置成了maxlongint div 2)

P.S. 题解要不要写的那么暴力啊,一会儿割小朋友,一会儿把它的所有的糖给割掉…

跟踪了一些变量来搞懂,没有时间把它搞精通了,所以默默地背代码去了。

话说,我100年没有写链表了!…貌似,这次写的还比较顺…

程序很大程度参考了http://www.cnblogs.com/htfy/archive/2012/02/15/2353147.html,我没有用静态链表,所以没写xor 1这种办法,用了dec_flow这样一个查找+修改flow的过程。

一开始跑数据死循环啊!尼玛建边的时候下次注意一点,这次建双向边但是回边的flow是0,不能和去边一样啊 =v=

program kd2;
type ptype=^node;
     node=record
             w,v,flow:longint;
             next:ptype;
           end;
const inf=maxlongint div 2;
      maxn=1000;
var n,m,tar,sta,v,c,x,cn,i,j:longint;
    head:array[0..2*maxn+10] of ptype;
    visit:array[0..2*maxn+10] of boolean;
    d,q:array[0..2*maxn+10] of longint;// distance to st;

function min(a,b:longint):longint;
begin
  if a<b then exit(a) else exit(b);
end;

procedure add_edge(st,ed,w:longint);
var p,q,pre:ptype;
begin
  new(q);
  q:=head[st];
  new(p);
  p^.w:=w;p^.v:=ed;p^.flow:=0;p^.next:=nil;
  if q=nil then
    begin
      new(head[st]);
      head[st]^:=p^;
    end
  else
    begin
      while q<>nil do
        begin
          pre:=q;
          q:=q^.next;
        end;
      new(q);
      q^:=p^;
      pre^.next:=q;
    end;
end;

function bfs:boolean;
var star,rear,x,u:longint;
    y:ptype;
begin
  fillchar(visit,sizeof(visit),false);
  fillchar(q,sizeof(q),0);
  star:=1;rear:=1;d[sta]:=1;visit[sta]:=true;q[star]:=sta;
  while star<=rear do
    begin
      x:=q[star];
      y:=head[x];
      while y<>nil do
        begin
          if (not visit[y^.v]) and (y^.flow<y^.w) then
            begin
              visit[y^.v]:=true;
              d[y^.v]:=d[x]+1;
              inc(rear);
              q[rear]:=y^.v;
            end;
          y:=y^.next;
        end;
      inc(star);
    end;
  exit(visit[tar]);
end;

procedure dec_flow(st,ed,delta:longint);
var x:ptype;
begin
  x:=head[st];
  while x^.v<>ed do x:=x^.next;
  x^.flow:=x^.flow-delta;
end;

function add_flow(p,maxflow:longint):longint;
var u,o:longint;
    y:ptype;
begin
  if (p=tar) or (maxflow=0) then exit(maxflow);
  add_flow:=0;
  y:=head[p];
  while y<>nil do
    begin
      if (d[y^.v]=d[p]+1) and (y^.flow<y^.w) then
        begin
          o:=add_flow(y^.v,min(maxflow,y^.w-y^.flow));
          if o>0 then
            begin
              inc(y^.flow,o);
              //dec(op_y.flow,o);
              dec_flow(y^.v,p,o);
              dec(maxflow,o);inc(add_flow,o);
              if maxflow=0 then break;
            end;
        end;
      y:=y^.next;
    end;
end;

function network:longint;
begin
  network:=0;
  while bfs do
    begin
      inc(network,add_flow(sta,inf));
    end;
end;

begin
  assign(input,'kd.in');reset(input);
  assign(output,'kd.out');rewrite(output);
  readln(n,m);sta:=0;tar:=m+n+1;
  for i:=1 to m do
    begin
      read(v);
      add_edge(n+i,tar,v);
      add_edge(tar,n+i,0);
    end;
  for i:=1 to n do
    begin
      read(c);
      cn:=cn+c;
      add_edge(sta,i,c);
      add_edge(i,sta,0);
      read(x);
      for j:=1 to x do
        begin
          read(c);
          add_edge(i,n+c,inf);
          add_edge(n+c,i,0);
        end;
      readln;
    end;
  writeln(cn-network);
end.
kd

 

 posted on 2014-07-10 16:05  Sky-Grey  阅读(183)  评论(0编辑  收藏  举报