-1

Code to split data into a training / test subsets is below. Note that data_points is one long vector of size items*attr, and data_labels is a vector of size items.

int split_data(int items, int attr, double *data_points, int *data_labels, double **split_train_points, int **split_train_labels, double **split_test_points, int **split_test_labels)
{
  srand(time(NULL));
  int i, j;
  double temp0, temp1;
  double sorter[items][2];

  *split_train_points = malloc(floor(SPLIT_PROP*items * attr) * sizeof(double));
  *split_train_labels = malloc(floor(SPLIT_PROP*items       ) * sizeof(int));

  *split_test_points  = malloc(ceil((1-SPLIT_PROP)*items * attr) * sizeof(double));
  *split_test_labels  = malloc(ceil((1-SPLIT_PROP)*items       ) * sizeof(int));

  // create a 2d array with element number in one column and a random number in the other
  for (i = 0; i < items; i++) {
      sorter[i][0] = i;
      sorter[i][1] = rand() / (double)RAND_MAX;
  }

  // sort by the random number column
  for (i = items-1; i > 0; i--) {
    for (j = 1; j <= i; j++) {
      if (sorter[j-1][1] > sorter[j][1]) {
        temp0 = sorter[j-1][0];
        temp1 = sorter[j-1][1];

        sorter[j-1][0] = sorter[j][0];
        sorter[j-1][1] = sorter[j][1];

        sorter[j][0] = temp0;
        sorter[j][1] = temp1;
      }
    }
  }

  int cutoff = floor(SPLIT_PROP*items);
  int element = 0;
  // now we have a bunch of indices in a random order.  we select the first 70% to store into our split_train datasets
  for (i = 0; i < cutoff; i++) {
    element = (int)sorter[i][0];
    *split_train_labels[i] = data_labels[element];
    printf("success!\n");
    for (j = 0; j < attr; j++) {
      printf("j: %d, data_points_element: %d\n",j,attr*element+j);

      //SEGFAULT OCCURS HERE WHEN J=4 EVERY TIME EVEN AS ELEMENT VALUE CHANGES DUE TO RANDOMNESS
      *split_train_points[attr*i+j] = data_points[attr*element+j];
      printf("j out! %d\n",j);
    }
  }


  for (i = cutoff; i < items; i++) {
    *split_train_labels[i - cutoff] = data_labels[(int)sorter[i][0]];

    for (j = 0; j < attr; j++) {
      *split_train_points[attr*(i-cutoff)+j] = data_points[attr*(int)sorter[i][0]+j];
    }
  }  

  return 0;
}

As noted in the code, SEGFAULT occurs at the same line, at j=4, even though "element" is a random number.

4

1 回答 1

2

我猜这是因为这个表达*split_train_labels[i]并不意味着你认为它的意思。对于编译器来说是一样的*(split_train_labels[i])，但是你可能的意思是(*split_train_labels)[i]。您在多个地方都有这个问题。

数组索引的优先级高于指针取消引用。

于 2013-04-25T08:41:38.763 回答