Setup of the ROI for the CMA-ES

The JAVA CMA-ES algorithm uses the parameters from Tab. 3. Note, that there are constraints on the ls and us parameters:

\begin{displaymath}
ls > 0 \textrm{ and } us > ls.
\end{displaymath} (2)

Therefore, we introduce a new variable, namely usfprop, which is defined as follows:
\begin{displaymath}
us = usprop \times (1.0 + ls).
\end{displaymath} (3)

These parameters are related to the algorithm design and will be specified in SPOT's ROI file.

Table 3: JAVA CMA-ES: Algorithm parameters as reported by the algorithm
\begin{table}
\begin{tabularx}{\textwidth}{lp{0.7\textwidth}l}
Name & Parameter ...
... standard deviations (step sizes) factor & USPROP \\
\end{tabularx}\end{table}


The corresponding ROI file for the CMA-ES looks as follows.

name low high type
S 2 50 INT
RESTARTS 1 10 INT
IPSF 1 5 FLOAT
ISD 0.1 3 FLOAT
LS 0 1 FLOAT
USPROP 1 2 FLOAT
SPOT provides an ROI template, which can be downloaded from the workshop's web site.



bartz 2010-07-08