Hi,
I just loaded a new example in the
repository.
Example "10. Symbolic regression - faster version"
It is very similar with example "01. Symbolic regression", but during the "compute_eval_matrix" function the values for variables are not copied, but instead are taken directly from the "training_data" matrix.
One more change was needed for this feature: "training_data" matrix is now stored as an array of columns instead of an array of rows.
On a simple test (see "building1.txt" data) we have reduced the running time from 25 seconds down to 17 seconds.
best regards,
Mihai