Merge pull request #13 from automl-edu/w08_update

keep the description consistent with the formulation

w08_hpo_multicrit/t03_bo.tex

@@ -152,18 +152,18 @@
     $$
     c = \max_{i=1,\dots,m}\left(w_i \cost_i(\conf)\right) + \rho \sum_{i=1}^m w_i\cost_i(\conf),
     $$
-    % \begin{scriptsize}
+    \begin{scriptsize}
     \begin{itemize}
         \item The norm consists of two components:
             \begin{itemize}
-                    \item $\max_{i=1,\dots,m}\left(w_i \cost_i(\conf)\right)$ takes only the cost function with maximum weight into account.
+                    \item $\max_{i=1,\dots,m}\left(w_i \cost_i(\conf)\right)$ takes only the maximum weighted cost into account.
                     \item $\sum_{i=1}^m w_i\cost_i(\conf)$ is the weighted sum of all cost functions.
             \end{itemize}
         \item $\rho$ describes the trade-off between these components.
         \item By the randomized weights in each iteration and the usually small value of $\rho = 0.05$, this allows exploration of extreme points of single cost functions.
         \item One can prove: \textbf{Every solution of the scalarized problem is pareto-optimal!}
     \end{itemize}
-    % \end{scriptsize}
+    \end{scriptsize}
 \end{frame}
 
 \begin{frame}{ParEGO Algorithm}