Inference Step: Use of Equation 5 over Equation 6

vsomnath / aligned_diffusion_bridges

Aligned Diffusion Schroedinger Bridges (UAI 2023)

https://arxiv.org/abs/2302.11419

MIT License

31 stars 0 forks source link

Inference Step: Use of Equation 5 over Equation 6 #3

Open HanLiii opened 3 months ago

HanLiii commented 3 months ago

Hi Vignesh,

Thank you for your excellent work. I have a question regarding the inference step: why should we use Equation 5 instead of Equation 6? It seems that Equation 5 is more concise. Could you clarify this choice?

vsomnath commented 4 weeks ago

Hi, sorry for the very delayed response here. Are you referring to the latest version on arxiv (https://arxiv.org/abs/2302.11419) for Eq. 5 and Eq. 6?

We use Eq. 5 instead of Eq. 6 because the drift in Eq. 5 is not dependent on x_1 (which we are trying to predict), while Eq 6 has the h-transform term dependent on x_1.