I have a matrix of 0's and 1's, and wish to compute its rank. One approach is to compute an approximate decomposition and return a count of the number of pivots that exceed some small threshold.
Though this will usually equal the rank for a reasonably chosen threshold, the two definitions are not mathematically equivalent. What is the best way to compute according to the mathematical definition of rank?