将Python函数参数作为变量传递

Question

I'm trying to write a function that runs through various iterations of classification algorithm (k-Means).

In sklearn.neighbors.KNeighborsClassifier, there are a few parameters to adjust: n_neighbors and leaf_size. I'd like to know if there is a way to specify which parameter to adjustment during a particular iteration.

from sklearn.neighbors import KNeighborsClassifier
def useNeighbors(iterations, *args):
    print(iterations) #normal argument
    for arg in args:
        KNeighborsClassifier(arg=20)

useNeighbors(2, "n_neighbors", "leaf_size")

I want this to essentially instantiate a KNeighborsClassifer instance twice- the first time with the # of neighbors at 20, and then the second time with the leaf size at 20 (default values for # of neighbors is 5, and default leaf size is 30).

This, however, unsurprisingly yields

2
TypeError: _init_params() got an unexpected keyword argument 'arg'

It prints the iterations argument as expected, but then KNeighborsClassifer is not recognizing the string argument 'n_neighbors' as my attempt to specify which parameter to adjust.

How do I switch which parameter/argument I want to adjust across many different iterations?

Also, obviously this is a toy case- I'm asking because I'm hoping to integrate different ML classification algorithms into an ensemble package with hyperparameters tuned through a Markov Chain Monte Carlo iterative method. But in order to do that, I need to be able to specify which parameters in each algorithm take the "steps" found in the Markov Chain across each iteration.

Answer 1

If I understand what you want, you can use partials for this. Example

from functools import partial
from sklearn.neighbors import KNeighborsClassifier

    classifiers = [partial(KNeighborsClassifier, n_neighbors=20),
                   partial(KNeighborsClassifier, leaf_size=20)]
    for classifier in classifiers:
        classifier()

Here 'sa good explanation of using partials.

Answer 2

You just need to use a spread :

from sklearn.neighbors import KNeighborsClassifier
def useNeighbors(iterations, *args):
    print(iterations) #normal argument
    for arg in args:
        my_dict = {}
        my_dict[arg] = 20
        KNeighborsClassifier(**my_dict)

useNeighbors(2, "n_neighbors", "leaf_size")