Question

我正在对张量进行一些迭代计算。我必须将张量映射到计算中的矩阵。我用于矩阵运算的库是“Eigen”。

类Tensor的描述：

*运算符被重载以执行张量乘法。（共同指数总结）

类SVD的描述：

用于执行奇异值分解（SVD）。
分解由Eigen :: JacobiSVD完成。
它用于在程序中输出ltensor和rtensor（均为3级）。
示例：如果输入矩阵为D ^ 2xD ^ 2，则ltensor为DxDxD ^ 2，rtensor为D ^ 2xDxD。
ltensor和rtensor的乘法将给出原始张量。

该计划的描述：

初始张量为等级4，每个索引的维度为2（2x2x2x2）。然后通过函数“transform_to_matrix（）”将其映射到4x4矩阵。
经过一些计算后，张量T的每个指数的维数被放大到4（4x4x4x4）。
在下一次迭代中，我希望新的张量映射到16x16矩阵。

错误：
错误“free（）：invalid pointer”出现在第二次迭代中标有（*）的行。

void SOLVE::solve(){
  Tensor S=this->tensor_; // initial tensor with index {"u","d","l","r"}
  std::vector<std::string> row1{"u","l"};
  std::vector<std::string> col1{"d","r"};
  std::vector<std::string> row2{"u","r"};
  std::vector<std::string> col2{"d","l"};
  Tensor S1;
  Tensor S2;
  Tensor S3;
  Tensor S4;
  SVD svd;
  for(int i=0;i<2;i++){
    std::cout << i<< "check" << std::endl;
    Eigen::Matrix<double,Eigen::Dynamic,Eigen::Dynamic> matrix1=
    transform_to_matrix(S,row1,col1); // (*) map to matrix by the sets of 
                                      // indices
    std::cout << i<< "check" << std::endl;
    svd.set_matrix(matrix1); // input matrix to do SVD
    svd.solve(); // solve the SVD
    S1=svd.get_ltensor("u","l","e"); // "u", "l", "e" are for index 
                                     // labeling
    S3=svd.get_rtensor("f","r","d"); // "f", "r", "d" are for index
                                     // labeling
    Eigen::Matrix<double,Eigen::Dynamic,Eigen::Dynamic>matrix2=
    transform_to_matrix(S,row2,col2); // map to matrix by another sets of 
                                      // indices
    svd.set_matrix(matrix2); // update to new matrix
    svd.solve(); // solve the SVD problem again
    S4=svd.get_ltensor("u","r","g"); 
    S2=svd.get_rtensor("h","d","l"); 
    S1.set_index("u","a"); // rename the index "u" to "a"
    S1.set_index("l","n");
    S1.set_index("e","r");
    S2.set_index("l","b");
    S2.set_index("d","a");
    S2.set_index("h","u");
    S3.set_index("r","b");
    S3.set_index("d","m");
    S3.set_index("f","l");
    S4.set_index("u","m");
    S4.set_index("r","n");
    S4.set_index("g","d");
    // Indice of
    // S1: {"a","n","r"}
    // S2: {"b","a","u"}
    // S3: {"b","m","l"}
    // S4: {"m","n","d"}
    S=S1*S2*S3*S4; // indices of S: {"u","d","l","r"}
  }      
}

Eigen::Matrix<double,Eigen::Dynamic,Eigen::Dynamic>       
transform_to_matrix
(const Tensor &S,
std::vector<std::string> row,
std::vector<std::string> col)
{
  // This function will map a DxDxDxD tensor to a D^2xD^2 matrix,
  // according to the row indices and col indices required.
  // Example:
  // row={"u","d"}, col={"l","r"}
  // The indices "u" and "d" are grouped into the row
  // The indices "l" and "r" are grouped into the col
  Eigen::Matrix<double,Eigen::Dynamic,Eigen::Dynamic> matrix;
  ... // some mapping procedures.
  std::cout << matrix << std::endl;
  return matrix;
}

节目输出：

0check
... (a 4*4 matrix is printed here)
0check
1check
... (a 16*16 matrix is printed here)
Error message is printed here: "free(): invalid pointer:"

打印出16x16矩阵，所以一切都很好，直到函数transform_to_matrix（）结束。所以问题是第二次迭代中的“矩阵”不能接受来自transform_to_matrix（）的返回。为什么会这样？我该如何解决？

更新：

错误有时是“free（）：next next（fast）：”有时“free（）：无效指针”。

使用gdb进行调试

输出：

0check
... (a 4*4 matrix is printed here)
0check
1check
... (a 16*16 matrix is printed here)
Error message is printed here: "free(): invalid next size (fast): 0x00000000006c5570" 
...
Program received signal SIGABRT, Aborted.
0x00007ffff71a1428 in __GI_raise (sig=sig@entry=6) at 
../sysdeps/unix/sysv/linux/raise.c:54
54  ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.

键入bt：

#0  0x00007ffff71a1428 in __GI_raise (sig=sig@entry=6) at../sysdeps/unix/sysv/linux/raise.c:54
#1  0x00007ffff71a302a in __GI_abort () at abort.c:89
#2  0x00007ffff71e37ea in __libc_message (do_abort=do_abort@entry=2, 
fmt=fmt@entry=0x7ffff72fce98 "*** Error in `%s': %s: 0x%s ***\n") at 
../sysdeps/posix/libc_fatal.c:175
#3  0x00007ffff71ec37a in malloc_printerr (ar_ptr=<optimized out>, 
ptr=<optimized out>, str=0x7ffff72fcf10 "free(): invalid next size 
(fast)", action=3) at malloc.c:5006
#4  _int_free (av=<optimized out>, p=<optimized out>, have_lock=0) at 
malloc.c:3867
#5  0x00007ffff71f053c in __GI___libc_free (mem=<optimized out>) at 
malloc.c:2968
#6  0x0000000000404cbe in __gnu_cxx::new_allocator<int>::deallocate 
(this=0x7fffffffcc90, __p=0x6c5570) at 
/usr/include/c++/5/ext/new_allocator.h:110
#7  0x00000000004048b0 in std::allocator_traits<std::allocator<int> 
>::deallocate (__a=..., __p=0x6c5570, __n=2) at 
/usr/include/c++/5/bits/alloc_traits.h:517
#8  0x0000000000404268 in std::_Vector_base<int, std::allocator<int> 
>::_M_deallocate (this=0x7fffffffcc90, __p=0x6c5570, __n=2) at 
/usr/include/c++/5/bits/stl_vector.h:178
#9  0x0000000000403813 in std::_Vector_base<int, std::allocator<int> 
>::~_Vector_base (this=0x7fffffffcc90, __in_chrg=<optimized out>) at 
/usr/include/c++/5/bits/stl_vector.h:160
#10 0x0000000000407fdd in std::vector<int, std::allocator<int> 
>::~vector (this=0x7fffffffcc90, __in_chrg=<optimized out>) at 
/usr/include/c++/5/bits/stl_vector.h:425
#11 0x0000000000467795 in transform_to_matrix (t=..., row=std::vector 
of length 2, capacity 2 = {...}, 
col=std::vector of length 2, capacity 2 = {...}) at helper.cpp:189
#12 0x000000000040bcc2 in SOLVE::solve (this=0x7fffffffe1e0) at 
solve.cpp:22
#13 0x0000000000402279 in main () at main.cpp:43

＃13：

t.solve(); // call solve(); main.cpp:43

＃12：

Eigen::Matrix<double,Eigen::Dynamic,Eigen::Dynamic> matrix1=transform_to_matrix(S,row1,col1); // give error in the second iteration; solve.cpp:22

＃11：

std::vector<int> col_arr(col_size,0); // col_size=2 ; helper.cpp:189

返回特征矩阵时出现“free（）：无效指针”错误

0 个答案: