Open
Description
Describe the workflow you want to enable
sklearn.preprocessing.OneHotEncoder should use as dtype the np.uint8 by default instead of np.float64 as it is currently. I don't see the reasoning behind using np.float64, this if anything only causes memory explosions and potentially (?) even slows down the computation.
Describe your proposed solution
Make np.uint8 as the default dtype parameter of sklearn.preprocessing.OneHotEncoder
Describe alternatives you've considered, if relevant
No response
Additional context
No response