Potential bug in the code

[won-bae]

Hi authors,

Thank you for sharing the code for the great work. While playing with the code, I noticed that it seems like the score is not weighted by the kernels, which is quite different from eq.(4) in the paper. Is there a particular reason for this?
When I added the weights for the score, it does not seem to work with the default and some other hyperparameters.