Open bjschoenfeld opened 5 years ago
We could build a simple problem to classify whether pipelines are good (i.e. within 5% of the best) or not. This may be easier to learn and more helpful to an AutoML system than predicting the precise pipeline score.
We could build a simple problem to classify whether pipelines are good (i.e. within 5% of the best) or not. This may be easier to learn and more helpful to an AutoML system than predicting the precise pipeline score.