On the bigℳ in the affine scaling algorithm