Gabor kernel

[changlan]

Gabor kernel is not implemented and it would be great to have it. The covariance function is parameterized as: k(x, x') = exp(-||x - x'||^2 / (2*l^2)) cos(2*pi*||x - x'|| / p) in which the hyperparameters are:

• l: length scale
• p: period

[mzwiessele]

We are always happy to see people engage with the framework and look forward to any contribution.

If you would like to implement it, here's the link to how to implement a new kernel: http://pythonhosted.org/GPy/tuto_creating_new_kernels.html

and the contribution guidlines: https://github.com/SheffieldML/GPy#contributing-to-gpy

Thanks for using the framework!

[Bonnevie]

For the record, there is already a cosine kernel so this is just a product of existing kernels.

[mzwiessele]

Is there a reason to implement a new kernel? Or could you just use the product of the RBF and Cosine kernel as the Gabor kernel?