Trace minimization when some matrix is unknown

234 Views Asked by At

The problem is as follows: $\displaystyle\min_{V}$ trace($V^TH^T\Phi HV$)$\\$ s.t. $V^TV=I_d$ in the case when $H$ is not known.

When $H$ is known, the solution is given by the eigenvectors corresponding to $d$ minimum eigenvalues of $ H^T\Phi H$.

Is there anyway to solve the problem in the absence of $H$? $H$ is a full column rank matrix. $\Phi$ is symmetric and positive semi-definite.

I hope to receive the suggestions if there are any. Or is there any technique that I can rephrase the problem.

Thank you for your help.