WIP: First draft of a random search in GridSearchCV #455
Conversation
| self.budget = budget | ||
| if self.budget: | ||
| self.rolled_out_grid = list(IterGrid(param_grid)) | ||
| random.shuffle(self.rolled_out_grid) |
There was a problem hiding this comment.
In order to have reproducible results, we should instead use a random_state (added and initialized in the constructor), and then call shuffle on it.
There was a problem hiding this comment.
Yes but in the following way:
from random import Random
from sklearn.utils import check_random_state
...
# in __init__
self.random_state = random_state
...
# then in fit
self.random_state = check_random_state(self.random_state)
if self.budget:
py_random_state = Random(self.random_random.rand())
self.rolled_out_grid = list(IterGrid(param_grid))
py_random_state.shuffle(self.rolled_out_grid)There was a problem hiding this comment.
What is the purpose of py_random_state? Not to change the estimator during fit? I would do
random_state = check_random_state(self.random_state) and then use random_state later. Or doesn't check_random_state make a copy? Maybe it should.
There was a problem hiding this comment.
because AFAIK the numpy rng cannot shuffle inplace a python list.
There was a problem hiding this comment.
Ok, that was the subtlety I knew I overlooked, thanks :)
|
Closing this one as I just merged #1194. |
Hi,
So far, no tests, so this is just to get feedback on the API and goals of an extension to sklearn's GridSearchCV that does random search when given a budget.