Fast way to calculate kernel matrix, python

Question

let K(x, z) be (x_transpose*z + p_constant)**2.

I want to compute the n*n matrix K, where K_ij = k(X_i, X_j)

X is a n by d matrix, and X_i is the transpose of the ith row of X.

Does anyone know of a quick way to compute this? I'm using python.

Wait a second, is K just XX^T?

check this page: http://stackoverflow.com/questions/5889142/python-numpy-scipy-finding-the-null-space-of-a-matrix — roadrunner66, Mar 31 '16 at 05:13

score 0 · Answer 1 · edited May 23 '17 at 12:24

0

import numpy as np
def K(x,z, p_constant=1.0):
  return (np.dot(x.T,z)+p_constant)**2
#...
x=np.arange(100).reshape((10,10))
np.fromfunction(np.vectorize(lambda i,j: K(x[i],x[:,j])), x.shape, dtype=x.dtype)

I had some trouble with np.fromfunction's misleading documentation.

edited May 23 '17 at 12:24

Community

1
1

answered Mar 31 '16 at 13:32

Felipe Lema

2,700
12
19

score 0 · Answer 2 · edited Jan 31 '21 at 23:34

0

Yes, the answer on question "How to compute linear kernel matrix" is

np.dot(X , np.transpose(X).

edited Jan 31 '21 at 23:34

Trenton McKinney

56,955
33
144
158

answered Jan 31 '21 at 21:43

Hodlevsky abcdefg

1
1

Fast way to calculate kernel matrix, python

2 Answers2