Examples for higher dimensional data

[arkottke]

I am struggling with how to work with higher dimensional kernals. A little more information in the documentation or a tutorial would be very help. In my particular case, I am working with earthquake observations which include two positions (latitude and longitude of the epicenter, and latitude and longitude of the recording station).

Thanks.

p.s. George is pretty handy.

[tmcclintock]

This is probably late but here is a minimal example for multidimensional data

"""Minimal example showing multidimensional GP regression.
"""
import george
import numpy as np

x = np.linspace(0, 5*np.pi)
x = np.vstack((x,x)).T #2D data
#x.shape is now (50,2)

y = np.sin(x[:,0]) + np.cos(x[:,1])

kernel = george.kernels.ExpSquaredKernel(1, ndim=2)

gp = george.GP(kernel)
gp.compute(x)

#Note that multidimensional input requires a 2D array of data
#even for a single test point
x_test = np.array([[6, 6]])

print(gp.predict(y, x_test))
#Prints the mean and variance at the point `x_test`

The thing that people sometimes stumble over is the dimensionality of the x points that the GP is making a prediction at - the input must be a list of N-length arrays where each data point x is N dimensional.

[tmcclintock]

This is being addressed in #112 .