Imagine that I have a group of pipelines composed of diferent sequences of "imputation -> classification algorithm -> hyperparameters of this algorithm". So, TPE will creat those two distribuitions for each hyperparameter of each algorithm in each imputation method? Obviusly its not what is haṕpening, but, how dos the TPE deal with this kind of evaluation? how does it choses when to try another algorithm that have a completly diferent set of hyperparameters?
Imagine that I have a group of pipelines composed of diferent sequences of "imputation -> classification algorithm -> hyperparameters of this algorithm". So, TPE will creat those two distribuitions for each hyperparameter of each algorithm in each imputation method? Obviusly its not what is haṕpening, but, how dos the TPE deal with this kind of evaluation? how does it choses when to try another algorithm that have a completly diferent set of hyperparameters?