Approximation in Frobenius norm by low rank matrices

257 Views Asked by At

I am trying to optimize the following matrix function

$\min_{H,V}$ $\|A-HV\|_F^2$

under the following constraints (1) $H$ is low rank (2) $V$ is positive (all entries of $V$ are non-negative)

Any pointers will be welcome ??

  • can I alternate between solving for $H$ and $V$
  • fix $H$ and solve for $V$ and then vice-versa