I am trying to implement kernelized ridge regression. There are 20000 data rows approximately and about 150 features.
This is the model being fit:
KernelRidge(alpha, kernelType, gamma=0.005, degree=3, coef0, kernel_params=None)
where kerneltype has been set to 'rbf' and 'linear', both times slowing down and eventually eating up a lot of memory
I dont think degree is causing the problem. I tried the same thing with degree=1 as well.
I also tried the same thing with gamma = 1 but faced the same issue.
Regular ridge, lasso and linear regression take not more than a second to complete (for the same data).
Where could I be going wrong?