Loguniform

[Kriti_Shahi]

Hi,

The questions are related to log transformations used in parameter_tuning_randomized_search notebook.

I know that we have specified the range while using the loguniform function but I want to print the values that lie in the range mentioned. How should I proceed ?

param_distributions = {
‘classifier__l2_regularization’: loguniform(1e-6, 1e3),
‘classifier__learning_rate’: loguniform(0.001, 10),
‘classifier__max_leaf_nodes’: loguniform_int(2, 256),
‘classifier__min_samples_leaf’: loguniform_int(1, 100),
‘classifier__max_bins’: loguniform_int(2, 255),
}

How do I know if I need to use log to the base 10 or log to the base 2 for different hyperparameters?

fig = px.parallel_coordinates(
cv_results.rename(shorten_param, axis=1).apply({
“learning_rate”: np.log10,
“max_leaf_nodes”: np.log2,
“max_bins”: np.log2,
“min_samples_leaf”: np.log10,
“l2_regularization”: np.log10,
“mean_test_score”: lambda x: x}),
color=“mean_test_score”,
color_continuous_scale=px.colors.sequential.Viridis,
)

Thanks in advance!

[glemaitre58]

I think that there is some confusion here. We used a logarithm distribution to sample parameter value. But the value of the parameter tested in the search CV is decimal and not on a log basis.

However, to make a better visualization and spread the axis, we scale the axis with log2 or log10 depending on the parameter range. In this case, the values shown on the axis of the parallel coordinate plot are in the log basis and need to be reverted with the basis used the latest cell that you shown.

[Kriti_Shahi]

The above answer makes sense. Thank you

[Marc_In_Singapore]

Can also be tested with: