Good reduction algorithms, computer programs usually choose a column as absolute value of the largest element in yuan, this approach is often referred to as part of the principal component method, why some principal component method can reduce the rounding error in the calculation?