At this point you can design the desired experiments. First of all, you must select one of the following options:
- k-fold cross validation
- 5x2 cross validation
- without validation

Once the previous step is done, you must push one of the following buttons, depending on if you want to do a classification experiment, a regression one or an unsupervised learning one:

In the Execution Options menu you can find some performance options to apply to the experiment. The menu is located in the Tools Menu:

The new windows offer two options which will be applied to each execution of your experiments. They are the following:
- Java Heap Size: Indicate the number of MB that will be allocated in each execution of the algorithm. Default value is 512MB. Please do not set a higher value than your actual amount of RAM. The MINIMUM accepted value has been set to 32MB.
- Java HotSpot Server VM “-server” option: If selected, the Java HotSpot VM will be invoked in each execution. This option is intended for long run experiments, in which the Java HotSpot VM will be able to detect the bottlenecks of your algorithm and optimize them. In order to use this option you MUST have the “server” subdirectory in your JRE “bin” directory. This “server” directory is offered in the JDK version of Java. You can copy the “server” directory from the path “<java installation directory>\jdk1.6.0\jre\bin”, and paste in “<java installation directory>\jre1.6.0\bin” and this option should works fine for you, from a clean JDK 6 installation.
