The CMF code requires inputs to be speficied in a specific sparse format. This function converts regular R matrices into that format.
A matrix of class
The element X[i,j] on the i:th row and j:t column is represented as a triple (i,j,X[i,k]). The input for CMF is then a matrix where each row specifies one element, and hence the representation is of size Nx3, where N is the total number of observed entries.
In the original input matrix the missing entries should be marked as NA. In the output they will be completely omitted.
Even though this format reminds the representation often used for representing sparse matrices, it is important to notice that observed zeroes are retained in the representation. The elements missing from this representation are considered unknown, not zero.
The input matrix in triplet/coordinate format.
Arto Klami and Lauri Väre
1 2 3