In numerical QR decomposition, when we calculate the orthonormal factor Q of a matrix, what is the difference in results if we use Householder Reflections to normalize the matrix or use Gram-Schmidt orthonormalization? Which one of these methods is preferred?
If the results can be different, does it mean that QR decomposition is not unique?