1: \begin{abstract}

2: Diffusion models have emerged as the de facto choice for generating visual signals.

3: However, training a single model to predict noise across various levels poses significant challenges, necessitating numerous iterations and incurring significant computational costs.

4: Various approaches, such as loss weighting strategy design and architectural refinements, have been introduced to expedite convergence.

5: In this study, we propose a novel approach to design the noise schedule for enhancing the training of diffusion models.

6: Our key insight is that the importance sampling of the logarithm of the Signal-to-Noise ratio ($\log \text{SNR}$), theoretically equivalent to a modified noise schedule, is particularly beneficial for training efficiency when increasing the sample frequency around $\log \text{SNR}=0$.

7: We empirically demonstrate the superiority of our noise schedule over the standard cosine schedule.

8: Furthermore, we highlight the advantages of our noise schedule design on the ImageNet benchmark, showing that the designed schedule consistently benefits different prediction targets.

9: 

10: 

11: \end{abstract}

12: