Origin of param_grid ranges?

M6. Ensemble of models M6. Exercise M6.01
#1

Is there a good resource for understanding the range of values that were chosen in the solution for param_grid? Why these values? Is there a proper methodology for determining these value ranges in general?

param_grid = {
“n_estimators”: randint(10, 30),
“max_samples”: [0.5, 0.8, 1.0],
“max_features”: [0.5, 0.8, 1.0],
“base_estimator__max_depth”: randint(3, 10),
}

#2

There is no magical methodology to set the hyperparameter range. It comes back to:

  • knowing how a specific model work and what is the impact of hyperparameters
  • knowing the effect of a hyperparameter on other hyperparameters
  • knowing the computational limitation in which you want to train/deploy the predictive model
  • identifying the complexity of the problem to be solved
  • and finally, some “try and see” mechanism

When we presented the linear models and tree-based models, we try to provide insights into the first two points. We did not go in the remaining 3 different aspects since it is more difficult to tackle inside a course but they would be encountered during the daily job of a data scientist.