Let $H$ be an infinite dimensional separable Hilbert space. Let $B(H)$ be the algebra of bounded operators on $H$. **Definition** : $T \in B(H)$ is **quasi-diagonalizable** if $\exists r \in \mathbb{N}$ and $(e_{n})_{n \in \mathbb{N}}$ orthonormal basis : $$ \forall n , m \in \mathbb{N} \ , \ (T.e_{n}, e_{m})\ne 0 \Rightarrow \vert n-m \vert \leq r $$ **[Invariant subspace problem (ISP)][1]**: Is it true that for all $T \in B(H)$ there exists a proper, closed and $T$-invariant subspace $K \subset H$ (i.e. $T(K) \subset K = \overline{K} \subsetneq H$)? **Remark**: This problem is known for the Banach spaces in general, but still open for an Hilbert space. > Is it true that the quasi-diagonalizable operators check the ISP? **Remark**: I post the following question on MSE : [Are all the operators quasi-diagonalizable?][2] It's probably false, but of course, if it's true, the question here would be equivalent to the ISP, and I would be interested to know if this way is (in your opinion) promising, and ever known (reference). [1]: http://en.wikipedia.org/wiki/Invariant_subspace_problem [2]: http://math.stackexchange.com/questions/452908/are-all-the-operators-quasi-diagonalizable