Questions about the multi-variate kernel density estimation

[songlin]

Hello,

Thanks for sharing the work. We are trying to follow your work in our paper. However, I am a little puzzled by the calculation here $\sigma_{rgb} * (0.8/n)^{-1/7}$

The expression is corresponding to the following code. https://github.com/poetrywanderer/CF-NeRF/blob/5e5a58304df1f1817c821466be4992e2cd29cde6/run_nerf_uncertainty_NF.py#L1036

Could you elaborate more? Thank you.

Best,

[poetrywanderer]

Hi,

Thanks for your interest. Actually, the choice of bandwidth H is empirical. Here we follow Silverman's rule of thumb, Hsqrt = $\sigma{rgb} (\frac{4}{(d+2)n})^{\frac{1}{d+4}}$. But we found -1/7 works better here.

You can find more explanation here: https://en.wikipedia.org/wiki/Multivariate_kernel_density_estimation

Best,

[songlin]

I see. Thank you.

[songlin]

@poetrywanderer Hi, can you tell me how to acquire the NLL metric in Table. 1? Or if possible, just point me to the code for computing it. Thank you.

[csrqli]

@poetrywanderer Hi, can you tell me how to acquire the NLL metric in Table. 1? Or if possible, just point me to the code for computing it. Thank you.

Looking for the same thing tooo