Bounding the distance between projection onto random subspace and projection onto the average subspace

14 Views Asked by At

Using results like the Davis-Kahan theorem and Wedin's $\sin \theta$ theorem, we can bound quantities like $\|\widehat{V}\widehat{V}^{\top} - VV^{\top}\|_F$ where $X = V\Sigma V^{\top}$ and $\widehat{X} = \widehat{V} \widehat{\Sigma} \widehat{V}^{\top}$ are singular value decompositions of $X$ and $\widehat{X}$ respectively. If $\widehat{X}$ is random and $\mathbb{E}(\widehat{X}) = X$, what references are there for bounding $\|\mathbb{E}(\widehat{V}\widehat{V}^{\top}) - VV^{\top}\|_F$?