I understand that a linear transformation from a vector space $V$ to a vector space $W$ is a rank-$2$ tensor. What I would like some help with is how exactly to represent specific linear transformations in tensor notation. I will write below my attempt to link the two; please let me know if it makes sense, and provide whatever suggestions you think might help me.

So, how much, if any, of the following is correct?

Let $(v_1,\ldots,v_n)$ and $(w_1,\ldots,w_m)$ be bases of $V$ and $W$, respectively. Then $v_i \otimes w_j$ is the linear transformation that maps $v_i$ to $w_j$, and $\{v_i \otimes w_j:{i \in \{1,\ldots,n\}, j \in \{1,\ldots,m\}}\}$ is a basis of $V \otimes W$.

If some linear transformation $T$ has the matrix \begin{bmatrix} t_{1,1} & t_{1,2} & \ldots & t_{1,n}\\ t_{2,1} & t_{2,2} & \ldots & t_{2,n}\\ \vdots & \vdots & \ddots & \vdots \\ t_{m,1} & t_{m,2} & \ldots & t_{m,n} \end{bmatrix}

then $$T= \sum_{i \in \{1,\ldots,n\},\\ j \in \{1,\ldots,m\}}(t_{j,i} [v_i \otimes w_j])$$ is the tensor representation of $T$.


how exactly to represent specific linear transformations in tensor notation.

Indeed every linear map $A:V\rightarrow W$ has a representation as a tensor $T\in V^{*}\otimes W$ such that $\forall x\in V:Ax=T(x)\in W$. In chosen bases,

$$ T=\sum a_{ij}\;v_j^*\otimes w_{i} $$

where $a_{ij}$ is the matrix of the linear map and $\{v_j^*\}$ is the basis in $V^*$ dual to $\{v_i\}$ so that each term is a map reading off the j-th coordinate of the $x$, $x\mapsto a_{ij}x^j\:w_i$.


Let $E:=\mathbb{R}^n$ and $E^*$ its dual.

Here is a "concrete" way of seeing the equivalence of a linear transform $L:E \rightarrow E$ with a linear application from $L:E \otimes E^* \rightarrow \mathbb{R}$:

It is nothing more that the decomposition of matrix $M=[m_{i,j}]$ of linear mapping $L$ (with respect to canonical bases of $E$ and $E^*$) under the classical form:

$$L=\sum_{i=1}^n\sum_{j=1}^n m_{i,j}E_{i,j}=\sum_{i=1}^n\sum_{j=1}^n m_{i,j}U_i \otimes V_j^* \ \ \ \ (1)$$

where $E_{i,j}$ is the matrix with all its entries equal to $0$, but its $i,j$th entry equal to one, and

$$U_i=\begin{bmatrix}0\\\cdots\\1\\\cdots\\0\end{bmatrix} \in E \ \ and \ \ V_j=\begin{bmatrix}0&\cdots&1&\cdots&0\end{bmatrix} \in E^*$$

(for $U_i$, $i$th coefficient $\neq 0$, for $V_j$, $j$th coefficient $\neq 0$)

Remark: the tensor product could be interpreted either as a Kronecker product, or as an ordinary product by a column vector with a row vector (element of $E^*$)


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.