what is the difference of GP and GEP in Engineering(running)?

[LJingnan]

It that running GEP require less RAM that GP

[ShuhuaGao]

I am not sure about the RAM, but GEP probably takes less memory thanks to its linear representation if you write the code well. Of course, geppy here does not focus on memory saving. Generally, unless your dataset is huge, the memory consumption should not be a problem.

For other cons and pros of GEP and GP, you'd better refer to technical papers. My feeling is that the search efficiency of GEP is higher than a naive tree-structure GP.

[LJingnan]

“My feeling is that the search efficiency of GEP is higher than a naive tree-structure GP.” Yes, I totally agree, but this is the case when every operation is legal. But in my field, some operation is illegal and I assign a score of big negative to these operation, which cause unefficiency, I am going to rewrite the mutate and initrepeat function so that when a new gene is generated, test its score first and if the score is nan, then repeat until not.

[ShuhuaGao]

test its score first

Another way is to test directly whether the gene is valid according to your predefined rule (if any). However, if it is not too difficult to get valid individuals via natural evolution, there seems to be no need to intervene manually.

An even more efficient way is to design custom mutation and crossover operators that can always produce valid individuals (if it is possible), i.e., preprocess vs postprocess.

[DanShai]

I did a similar attempt years ago I called it Genome Network . I didnt know that GEP and CGP names exist ! thanx for the explanation and the repos! I will take a look.