#include <bits/stdc++.h>
#define F first
#define S second
using namespace std;
int n;
typedef long long ll;
const ll inf=1e18;
template<int V, int E> struct MinCostFlow{
struct Edge{
int a, b;
ll ca, co;
} es[E*2];
int eu=0, nmz=0;
vector<int> g[V+1];
ll p[V+1], d[V+1];
int fr[V+1], u[V+1];
void addEdge(int a, int b, ll ca, ll co){
nmz=0;
es[eu++]={a, b, ca, co};
es[eu++]={b, a, 0, -co};
g[a].push_back(eu-2);
g[b].push_back(eu-1);
}
void normalize(int source){
if (nmz) return; nmz=1;
for (int i=1; i<=V; ++i){
p[i]=inf; u[i]=0;
}
p[source]=0;
queue<int> q; q.push(source);
while (!q.empty()){
int x=q.front();
u[x]=0;q.pop();
for (int e : g[x]){
if (es[e].ca>0 && p[x]+es[e].co<p[es[e].b]){
p[es[e].b]=p[x]+es[e].co;
if (!u[es[e].b]){
u[es[e].b]=1;
q.push(es[e].b);
}
}
}
}
}
ll augment(int x, ll fl){
if (fr[x]==-1) return fl;
ll r=augment(es[fr[x]].a, min(fl, es[fr[x]].ca));
es[fr[x]].ca-=r;
es[fr[x]^1].ca+=r;
return r;
}
pair<ll, ll> flow(int source, int sink, ll mf){
priority_queue<pair<ll, int> > dij;
for (int i=1; i<=V; ++i){
u[i]=0; fr[i]=-1; d[i]=inf;
}
d[source]=0;
dij.push({0, source});
while (!dij.empty()){
auto x=dij.top(); dij.pop();
if (u[x.S]) continue;
u[x.S]=1;
for (int e:g[x.S]){
ll nd=d[x.S]+es[e].co+p[x.S]-p[es[e].b];
if (es[e].ca>0 && nd<d[es[e].b]){
d[es[e].b]=nd;
fr[es[e].b]=e;
dij.push({-nd, es[e].b});
}
}
}
ll co=d[sink]+p[sink];
for (int i=1; i<=V; ++i){
if (fr[i]!=-1) p[i]+=d[i];
}
if (u[sink]){
ll fl=augment(sink, mf);
return{fl, fl*co};
}else return {0, 0};
}
pair<ll, ll> getKFlow(int source, int sink, ll k){
ll fl=0;
ll co=0;
normalize(source);
while (1){
pair<ll, ll> t=flow(source, sink, k);
fl+=t.F; k-=t.F; co+=t.S;
if (k==0||t.F==0) break;
}
return {fl, co};
}
};
int main(){
ios_base::sync_with_stdio(0); cin.tie(0);
cin >> n;
MinCostFlow<211, 15222> mcf;
for (int i=0; i<n; ++i){
mcf.addEdge(202, i+1, 1, 0);
mcf.addEdge(i+101, 203, 1, 0);
for (int j=0; j<n; ++j){
int c;
cin >> c;
mcf.addEdge(i+1, j+101, 1, c);
}
}
cout << (mcf.getKFlow(202, 203, inf).S) << "\n";
}