Inquire about the discrepancy between implementations of EoH

[taindp98]

Dear authors,
Thanks for organising the implementations of LLMs for algorithm design. I found an implementation of EoH in this repository, but the workflow of generating heuristics is very different from the original idea in EoH. More specifically,

EoH: "Five prompt strategies are designed to generate new heuristics. At each generation, each strategy is called N times to generate N heuristics. Each newly generated heuristic will be evaluated on problem instances and added to the current population if it is feasible. At most 5N new heuristics will be added to the current population at each generation. Then, N best individual solutions from the current population will be selected to form the population for the next generation.".

However, in this repository, LLM4AD: the implementation of EoH is quite different. There are only four prompt strategies, and at each generation, each strategy is called int(N/4). Finally, at most N new heuristics will be added to the current population at each generation.

Could you please help me understand this discrepancy and any reasons behind the changes in the settings?
Much looking forward to hearing from you.
Best.

[FeiLiu36]

The implementations in the EoH code and the LLM4AD platform are fundamentally the same, with two minor differences, as you noted:

1. Operators used: Our experiments show that the four operators (m1, m2, e1, and e2) are critical for performance, while the remaining one has negligible impact on the results.
2. Algorithm sampling method: Whether you sample N algorithms from one operator or iterate four times to sample N/4 algorithms per operator, the outcome is the same—as long as the total number of evaluations is fixed, all four operators are used with equal probability. The frequency of updating the population has limited influence on final performance once you have enough samples. The updated implementation in LLM4AD simplifies parallel sampling and evaluation.