Description |
1) How to compute a Pearson correlation coefficient in MATLAB? The idea is a simple fact in math: for two standardized variables (0 mean and 1 std), the correlation is equivalent to their inner product. Thus, for two large sample matrices X and Y, the sample correlation matrix is R=X'Y/(n-1), where n is the number of samples.
2) How to build a sparse matrix for a huge graph (e.g., 100,000 nodes) based on massive samples? Using the above code of computing fast correlation, together with an idea of dividing the sample matrix into several blocks, the task becomes pretty easy and efficient in MATLAB. |