Question for backward about UT

[SCUTykLin]

Hello:
That a great paper and very thanks for your opensource.
I have encountered an issue regarding the implementation of this paper. In your paper, u mentioned that:

. Instead, UT only acts as an acceleration structure to efficiently determine the particles that contribute to each pixel thus avoiding the need
for computing the backward pass through the non-linear projection function.

Therefore, is it unnecessary to implement the backpropagation of UT in CUDA?

[wilsonCernWq]

You are correct, in our formulation, computing the backward of UT is unnecessary.