About the equation 5

[hawkinglai]

Thanks for your contribution about the training-free framework in point cloud analysis.

I have a question that is the equation 5, I see that your equation is describing the function is equal to sin/cosine something with parameter alpha over beta.

However, in the code of models/point_nn.py line 92, I saw that you are using the div_embed = torch.div(self.beta xyz.unsqueeze(-1), dim_embed) It seems like the code are running a function with a parameter beta over alpha. Why is not should be: div_embed = torch.div(dim_embed, self.beta xyz.unsqueeze(-1))

I am confused whether it is mismatch with your paper, or just I am not familiar with pytorch or matrix computation. I would appreciate it if you could take your value time to reply to my question.

[KLM1109]

It seems to need to exchange Alpha and Beta？ And how to understand sin&cos equation's odd numbers and even numbers?

[hawkinglai]

I guess, it should be: Sequence | Index of token | matrix 0 matrix 1 ..... matrix even matrix odd x | 0 | sin cos sin cos y | 1 | sin cos sin cos z | 2 | sin cos sin cos

then you need to add the position token to the knn_x.