Questions about baseline methods

[kvttt]

Hi,

Thanks for presenting such an awesome work with extensive ablation studies and detailed explanations and thanks for making the code available!

I have two clarification questions regarding the baseline methods:

1. For the method that you call "MLP", is the latent feature dimension fixed to be 566 for different networks corresponding to different observation sets?
2. For the method that you call "SH", you mentioned the maximum order is 8. I wonder if you use only the even order spherical harmonics or the full set of spherical harmonics up to order 8. And do you have a specific regularization factor $\lambda$ that you use for least square fitting of the spherical harmonics?

Thank you in advance!

Best,
kaibo