1: \begin{definition} 

2: Suppose we have sequential time series snapshots $\bm{x}_1, \ldots, \bm{x}_{T+1}$ such that $\bm{x}_{j+1} = \bm{A} \bm{x}_j$. Let $\bm{A}$ have $r$ nonzero and distinct eigenvalues

3: $\lambda_1, \ldots, \lambda_r$

4: and corresponding eigenvectors 

5: $\bm{v}_1, \ldots, \bm{v}_r$.

6: We say that the {\bf DMD problem is well-posed}

7: if the conditions

8: \begin{enumerate}

9:  \item $\bm{x}_1$ is not orthogonal to any $\bm{v}_1, \ldots, \bm{v}_r$, and either

10:   \item $T \geq r$ and $\bm{X}_1$ and $\bm{X}_2$ share the same range, or

11:   \item $T \geq r+1$,

12: \end{enumerate}

13: %Theorem \ref{theorem:uniqueness} 

14: are satisfied. %\textcolor{blue}{too vague, put down all conditions. What is $r$?}

15: \end{definition}

16: