Consider the following situation: The user has created a working directory for running the experiments, say MyJavaExperiments. This directory contains the following files.
R is started in the working directory. The following command starts SPOT's automatic tuning procedure.
> library(SPOT) > spot("java0.conf")Sometimes is it required to start a clean R session, because data from previous runs are in the workspace. Execute
rm(list=ls());to perform a cleanup before SPOT is loaded and run.
The output from Fig. 1 is shown by default during the tuning process.
The tuning process terminates with the following output:
Best solution found with 507 evaluations: Y SIGMANULL VARA VARG COUNT CONFIG 0.1029542 0.445661 1.050709 72 5 110
A short summary is also shown:
y SIGMANULL VARA VARG Min. :0.01184 Min. :0.1035 Min. :1.023 Min. : 3.0 1st Qu.:0.11606 1st Qu.:0.2885 1st Qu.:1.064 1st Qu.:18.0 Median :0.19034 Median :0.4112 Median :1.096 Median :50.0 Mean :0.25662 Mean :0.6167 Mean :1.118 Mean :43.7 3rd Qu.:0.31951 3rd Qu.:0.7337 3rd Qu.:1.145 3rd Qu.:64.0 Max. :1.68859 Max. :4.5725 Max. :1.984 Max. :98.0
A result file (RES), which contains important information from the tuning process, has been written to the working directory. It can be downloaded as java0.res from the workshop's web page.
bartz 2010-07-08