1: \begin{definition}[Dual Representer Theorem~\cite{scholkopf2001generalized}]

2: % \label{repre_thm}

3: % Suppose we are given samples $\phi(\mathbf{x}) \in \mathbb{R}^M $ in kernel-mapped feature space, then we can rewrite $\mathbf{w}^\prime$ as a linear combination of the mapped samples $\phi(\mathbf{x}^j_T)$ with the associated coefficient $\alpha_j$. That is, 

4: % \begin{align}

5: % \mathbf{w}^\prime = \sum_{j=1}^n \alpha_j \phi(\mathbf{x}^j_T) = \Phi(\mathbf{X})^T \bm{\alpha}.

6: % \end{align} 

7: % \end{definition}

8: