Turn a matrix into a product of two vectors with the smallest possible error

32 Views Asked by At

Suppose there is a matrix M. I would like to decompose it into a product of vectors a and b so that the actual difference of N = a x b and M is the least possible. I also want to do this iteratively which means after computing a and b and resulting N i would like to do the same for M(2) = M - N. Given the iteration nature of this there should be no direct matches (so any entry in N is not equal to the actual corresponding entry in M) - rather than this they all get closer to certain limit - the optimal solution. How to do this? Is there any literature about this? I could not state my thoughts - but probably there is a math term for that problem and all sorts of solutions. Thanks!

1

There are 1 best solutions below

0
On

Thanks to 5xum I've found really nice article covering this. Take a look here