4 problems

Training-trajectory conjecture. The entire training trajectory might impact the final model quality.

Training-trajectory conjecture. The entire training trajectory, rather than only the final loss, is important for the quality of the trained diffusion model.

EFS has a forward optimization step whose time complexity is quadratic in the number of training samples. Stochastic optimization conjecture. Stochastic optimization techniques cou…

KODE's kernel sufficiency conjecture. This problem is difficult for KODE because the MMD between the reference and target samples is very small and the chosen kernel $K$ is not suf…