I have come across a definition of rank in some lecture notes I am using to prepare for a linear algebra qualifying exam and the notes define the rank of a linear map in a somewhat different manner than I am used to.

let $V$ be a finite dimensional vector space and $f \in End(V)$. Since $End(V) \cong V^* \otimes V$ we can provide a new definition of the rank of $f$.

How do we show $dim(Im(f)) = min \{ t : f = \sum_{i=1}^{t} v_{i}^{*} \otimes v_i \}$ where $v_{i}^{*} \in V^*$ and $ v_i \in V$?


It should be immediate that $\dim(\mathrm{Im}\;f))$ (let's call that the "geometrical rank") cannot be more than the tensor rank, because $v_i^*\otimes v_i$ represents an endomorphism that always produces a multiple of $v_i$.

To show that the tensor rank cannot be more than the geometrical rank, choose a basis $(v_1,\ldots,v_t)$ for the image of $f$. Then define $v_i^*$ as the linear map that sends $w\in V$ to the coefficient of $v_i$ of $f(w)$. Then $f$ is represented by $\sum_{i=1}^t v_i^* \otimes v_i$, so the tensor rank is at most $t$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.