How is jgap able to quicky evaluate many expressions? How does this compare to writing your own compiler?
When solving a regression problem, with RMSE as a fitness function, a single expression needs to evaluated on all data rows.
The fastest way to do this, would be to compile a single expression and run the assembler code?
As far as I understand jgap represents expressions using a tree. This is a convenient data structure for applying cross over and mutation.
But how is a expression converted to a format that allows for fast execution?