Intelligent setting of algorithm parameters

In the use of various Algorithm Generally, you need to set some parameters. The more parameters that need to be set by the heuristic algorithm, the more parameters take GA for exmaple. (POP: Number of populations, maxiter: Maximum number of iterations, variation rate, exchange rate ). These parameters play a vital role in algorithm calculation. The parameters of these algorithms are only set by professionals. At this time, the problem arises, if your product (including some complex numerical algorithms) is for common users (or users do not know the algorithm and do not set parameters), then your product will have different problems, or the computing results for different situations of similar problems are unstable, that is, the solution is not very good (better than you expected ).
Parameter adjustment is a task in algorithm testing. It can be made interactive by professionals. I used to help others write some Program (For them, there is a good result to satisfy), but I can always change the parameters based on their actual problems, and I will get a better solution ....... If you are a product, you cannot help all users adjust parameters, or expect users to adjust parameters themselves. (I have never thought about this problem before. I always expect other people to be as professional as me..., or almost ). Therefore, if you want something better than others (easy to use and good performance), you need to introduce intelligent settings of algorithm parameters.
How can I only set it? Similarly, we construct an optimization problem with a single goal (simpler than the original problem, which can be understood by ordinary people). Here, the parameters of the original algorithm are new variables ........, only for different problems and single targets can only get good results with good parameters. Simply put, an evolutionary suanfa is introduced to optimize the parameters of the original problem, but the new algorithm may also need parameter settings ............., however, the solution to the original problem is less sensitive to parameters (new algorithms.
After talking about this, I don't know whether you understand it or not. The specific method involves many specific problems, which I will not talk about.
Copyright: ariszheng ariszheng@gmail.com