我试图计算一个有限元程序,在这里我需要计算d维的偏导数。在有限元中,基函数N(x,y,z)=N(x)N(y)N(z),因此一阶导数是:
N(x)'N(y)N(z) N(x)N(y)'N(z) N(x)N(y)N(z)'
二阶导数是
N(x)''N(y)N(z) N(x)'N(y)'N(z) N(x)'N(y)N(z)' N(x)N(y)''N(z) N(x)N(y)N(z)' N(x)N(y)N(z)''
我希望有一个输入(i,d)的函数,以便在下表中告诉我这些模式:
我认为必须有一个简单的算法来实现这一目标。有人能帮我吗?THx
发布于 2016-12-27 22:09:27
这可以通过嵌套循环来解决:
int * part_deriv(int i, int d){
int *res;
int *curr;
int num_el = fact(i+d-1) / ( fact(i) * fact(d-1) ); //fact() is the factorial function
int el_size = d*sizeof(int);
int el;
res = calloc(num_el,el_size);
curr = calloc(d,sizeof(int));
*curr = i;
memcpy(res,curr,el_size); //put the first element in the array
el = 0;
while(el<num_el){
if(*curr != 0){
for( d_idx = 1 ; d_idx<d ; d_idx++, *cur++){
*curr--; // "move" one derivative from the first variable to 'd_idx' variable
*(curr+d_idx)++;
el++;
memcpy(res+(el*el_size),curr,el_size); //put the element in the array
}
*curr--;
} else{
break; //shouldn't be reached, but added to be sure
}
}
return res;
}
我还没有完全理解如何输出结果,所以可以用d
块解析输出的数组。
发布于 2016-12-28 02:00:15
我通过调用一个函数来实现这一点。
#include <vector>
#include <iostream>
using namespace std;
void func (int d, int i, vector<int> &k, int n, int start, vector<vector<int>> &a){
if (n==i)
{
vector<int> m;
int it=0;
for(int it1=0;it1<d;++it1){
int amount=0;
while(find(k.begin(),k.end(),it)!= k.end()){
amount++;
it++;
}
m.push_back(amount);
it++;
}
a.push_back(m);
}
else{
for(int jj=start;jj<d+i-(i-n);++jj){
k[n]=jj;
func(d,i,k,n+1,jj+1, a
);
}
}
}
vector<vector<int>> test(int d, int i){
vector<int> kk(i);
vector<vector<int>> a;
func(d,i,kk,0,0, a);
return a;
}
int main(){
auto f = test(4,2);
for(auto it1=f.begin();it1!=f.end();++it1){
for(auto it2= it1->begin();it2!=it1->end();++it2)
cout<<*it2<<" ";
cout<<endl;
}
}
以下是我对i=2,d=4的结果:
2 0 0 0
1 1 0 0
1 0 1 0
1 0 0 1
0 2 0 0
0 1 1 0
0 1 0 1
0 0 2 0
0 0 1 1
0 0 0 2
https://stackoverflow.com/questions/41348642
复制相似问题