How to obtain optimized length scale assigned to each dimension

[Swarnendu-sekhar-ghosh]

If I am using automatic relevance determination inside RBF, when I print the model how can I get the optimized values of length scales assigned to each dimension? Also, if I use a combination of kernels then how to specifically obtain the length scales of the RBF kernel?

[lawrennd]

Hi Swarnendu, From memory it doesn't display because it's variable length and can be long if input dimension is high.

Here's some code for displaying and changing values.

import GPy
import numpy as np
from IPython.display import display

kern = GPy.kern.RBF(4, ARD=True)
display(kern.lengthscale)

kern.lengthscale = [0.3, 1.2, 1.4, 1.2]
display(kern.lengthscale)

[Swarnendu-sekhar-ghosh]

Thank you so much, sir. Appreciate your quick response.

[shadzzz90]

Hi @lawrennd, what if we are using different combination of kernels?