1: \begin{definition}[Input-tailored variational expansion] \label{def:input-tailor-var-exp}

2: Let the signal generator driven system $\statemapAut$ with enlarged state $\veccj{w}$ be as in Definition~\ref{def:sg-ode-casc}. Let

3: \begin{align*}

4: 	\veccj{w}(t;\alpha) = \sum_{i=1}^{N} \alpha^i \veccj{w}_i(t) + \text{O}(\alpha^{N+1}), \qquad t \in [0,T)

5: \end{align*} 

6: be the variational expansion of $\veccj{w}$ w.r.t.\ the initial conditions $\veccj{w}(0;\alpha) = \alpha \veccj{b}$. Let $\breve{\Transfer{W}}_i$ be the associated univariate frequency representations of $\veccj{w}_i$ as in Theorem~\ref{theor:Pxu_quadratic}.

7: 

8: Then the input-tailored variational expansion of $\vecbj{x}$ described by $\statemapAut$ (respectively by $\statemap$ and $\siggen$) is defined as

9: \begin{align*}

10: 	\vecbj{x}(t;\alpha) = \sum_{i=1}^{N} \alpha^i \vecbj{x}_i(t) + \text{O}(\alpha^{N+1}), \qquad \quad \vecbj{x}_i(t) =  \tensorcj{P}_x \, \veccj{w}_i(t)

11: \end{align*}

12: with $\tensorcj{P}_x =[\tensorbj{I}_N, \tensorbj{0}_{N,q}]$. The input-tailored frequency representations $\breve{\Transfer{X}}_i$ are given as

13: \begin{align*}

14: 	\breve{\Transfer{X}}_i(s) = \tensorcj{P}_x \breve{\Transfer{W}}_i(s), \qquad s \in \mathbb{C}.

15: \end{align*}

16: \end{definition}

17: