Hi,
Is the exercice somewhat unfinished? I have no convergence (I think my model is Ok).
Not sure what are we supposed to do to “conclude”?
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.
Hi,
Is the exercice somewhat unfinished? I have no convergence (I think my model is Ok).
Not sure what are we supposed to do to “conclude”?
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.
I just saw the solution. I think I may have my answer on “unfinished business” – need more than 20 iterations?
However, I don’t understand this concluding sentence in the solution:
So the best hyperparameters give a model where the features are scaled but not centered and the final model is regularized.
Which features? The numerical attributes? StandardScaler does not center à la z-score?
Regularized?
The warning is raised by the LogisticRegression
in your model and not the randomized search itself. The number of iteration mentioned here would be controlled by the max_iter
parameter of the LogisiticRegression
and not n_iter
from the cross_validate
.
I think that in the first module we mentioned that not scaling features can lead to LogisticRegression
to not converge fastly. Having the combination with_mean=False
, with_std=False
will lead to not scaling features and thus is probably raising the warning.
True we should be more precise and mention the numerical features.
They are controlled by with_mean
and with_std
. Thus, some of the combinations will not be centered and only scaled or only centered and not scale, etc.
Regularized is controlled by the parameter C
but we should also bring more information here.