CODEWITHSUNDEEP
pythonarraysnumpyfunctionvectorization
johnny rotten
johnny rotten

Reputation: 23

Misunderstanding numpy.vectorize

I want to apply a function to each row in a vector. Even on a simple example like the one below I cant get it to work.

I make a function that takes two vectors and applies the dot product to them.

import numpy as np


def func(x,y):
    return np.dot(x,y)

y=np.array([0, 1, 2])
x=np.array([0, 1, 2])

print(func(x,y))

Of course the output is 5. Now I want to plug in multiple vectors x, and get a solution back for each one. I dont want to use a for loop, so I tried using the vectorize function. For instance below I define X=(x1,x2,x3) and I want the output func(X,y)=(func(x1,y), func(x2,y), func(x3,y)). Why doesn't the following code do that:

import numpy as np


def func(x,y):
    return np.dot(x,y)

y=np.array([0, 1 , 2])
X=np.array([[0,0,0], [1,1,1], [2,2,2]])


vfunc=np.vectorize(func)
print(vfunc(X,y))

Upvotes: 0

Views: 440

Answers (1)

Valdi_Bo
Valdi_Bo

Reputation: 30991

The first trick to do is to exclude y argument (it is a fixed value for all rows from x).

The second trick is to pass the signature: Both arguments are arrays and the result is a scalar.

So, to vectorize your function, run:

vfunc = np.vectorize(func, excluded=['y'], signature='(n),(n)->()')

Then, when you call vfunc(X,y), you will get:

array([0, 3, 6])

Upvotes: 1

Related Questions