- Rating
- 3
- Confidence
- 3
- Informativeness
- 3
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper points out and proved the numerical singularities of hard endpoint (h-transform-based) bridges from the perspective of probability theory. Based on it, the authors prescribe the non-degenerate Gaussian terminal marginal to overcome the singularity. They construct a fully closed-form, singularity-free, and -free diffusion bridge framework and unify some previous bridges like DDBMs, GOUB, and UniDB. Three different image restoration experiments demonstrate SOTA performance of SDDBMs.
Critical strengths
-
The paper is theoretical and the detailed proofs of the Theorems and Propositions are complete and clearly described in the Appendix. It provides an explanation of UniDB via probabilistic interpretation.
-
SDDBMs demonstrate superior performance across different image restoration tasks under four standard metrics.
Critical weaknesses
-
Some experimental details might not be aligned well with the proposed theory. A central contribution of SDDBMs is the proposed non-degenerate terminal distribution . However, their experiments were conducted by initializing the reverse process with , which omits the unavailable source-dependent term and the terminal noise (Appendix H). Dropping unavailable does make sense since it is unavailable, but the lack of justification for the terminal noise would weaken the paper's contribution.
-
The proposed SDDBMs framework recovers DDBMs, but the authors provided no theoretical derivations on the formulation of SDDBMs on VE, VP process and no empirical evidence on whether SDDBMs could improve original DDBMs.
Questions for authors
-
The qualitative visual comparisons with baselines (GOUB and UniDB) are somewhat limited. The results among GOUB, UniDB, and the proposed SDDBMs should be compared in Fig. 3. More qualitative visual comparison examples would better emphasize the strengths of SDDBMs. Moreover, some examples in Fig. 3 and Fig. 8 are redundant. The appendix should provide additional diverse visual comparisons.
-
The Number of Function Evaluations (NFEs) or sampling steps for their SDDBMs and baselines are missed. Could the authors clarify the NFEs used during inference?
-
Proposition 3.1 would benefit from a rewriting. As it is, the main idea is to show the singularity of diffusion bridges. I could understand the injection ( for any ) leads to the conclusion. However, in this version of Proposition 3.1, neither singularity nor the relation to diffusion bridges are emphasized.
-
Lack of quantitative evaluation for the ablation on hyperparameter . Although the authors have claimed that ``different values lead to comparable performance", the specific results should be also provided to validate the claim, which would be more convincing.
10 comments
W2: The proposed SDDBMs framework recovers DDBMs, but the authors provided no theoretical derivations on the formulation of SDDBMs on VE, VP process and no empirical evidence on whether SDDBMs could improve original DDBMs.
Thank you for raising this concern. In fact, the soft VP and VE formulations can be derived directly from Eq. (14) by choosing , for VP, and , for VE, respectively. With the general parameterization , the resulting soft VP and VE processes are given by
and
g_t^2 \frac{\hat{x}-x_t} {\hat{\sigma}^2+\bar{\sigma}_{t:T}^2}dt + g_tdw_t,$$ respectively. Furthermore, following the $x_0$-free reformulation in Section 3.4, the corresponding $x_0$-free forward SDEs can be expressed as: $$dx_t=\Bigg[g_t^2\left(\frac{e^{-\bar{\theta}_{t:T}}(\alpha-e^{-\bar{\theta}_{0:T}})}{\phi_t}-\frac12\right)x_t+\frac{g_t^2\beta e^{-\bar{\theta}_{0:T}}}{\phi_t}x^\star\Bigg]dt+\eta_t d\bar w_t$$ and $$dx_t= \Bigg[ \frac{g_t^2}{\phi_t} \left((\alpha-1)x_t+\beta x^\star\right) \Bigg]dt+\eta_t d\bar w_t,$$ respectively, where $\phi_t$ and $\eta_t$ are defined in Eqs.(18) and (21), with the corresponding VP and VE parameter settings. When $\hat{\sigma}^2\to0$, $\alpha=0$, and $\beta=1$, we have $\eta_t\to g_t$, and these two processes recover DDBM-VP and DDBM-VE in the hard-endpoint limit, respectively. Regarding the empirical evidence, due to time constraints, we have so far completed the experimental evaluation of SDDBMs-VE. As shown in Table 1, SDDBMs-VE substantially outperforms both DDBM-VE and UniDB-VE across all four evaluation metrics on the image deraining task. Moreover, it also achieves better performance than DDBM-VP and UniDB-VP across all four metrics. These results provide direct empirical evidence that SDDBMs can improve upon the original DDBMs. __Table 1. Comparison of UniDB-VP and UniDB-VE with the original DDBM-VP and DDBM-VE on image deraining.__ | Method| **PSNR ↑** | **SSIM ↑** | **LPIPS ↓** | **FID ↓** | |:---:|:---:|:---:|:---:|:---:| |DDBMs-VP|29.58|0.828|0.113|35.46| |UniDB-VP| 30.11 | 0.8414 | 0.102 | 33.17 | |DDBMs-VE|29.34|0.7654|0.185|43.22| |UniDB-VE| 29.46 | 0.7671 | 0.185| 42.57 | |SDDBMs-VE|**31.17**|**0.8655**|**0.086**|**26.19**| We are also working on the experimental evaluation of SDDBMs-VP and will report the corresponding results and incorporate them into the revised manuscript once the experiments are completed.Q1: The qualitative visual comparisons with baselines (GOUB and UniDB) are somewhat limited. The results among GOUB, UniDB, and the proposed SDDBMs should be compared in Fig.3. More qualitative visual comparison examples would better emphasize the strengths of SDDBMs. Moreover, some examples in Fig.3 and Fig.8 are redundant. The appendix should provide additional diverse visual comparisons.
Thank you for this helpful suggestion. We have revised Fig.3 to directly compare GOUB, UniDB, and SDDBMs on the same examples for image deraining, inpainting, and super-resolution.
We have also revised Figs.7 and 8 in the appendix to include more diverse qualitative comparisons and eliminate the overlap with Fig.3. These revisions provide a more comprehensive visual comparison between SDDBMs and the baselines.
Q2: The Number of Function Evaluations (NFEs) or sampling steps for their SDDBMs and baselines are missed. Could the authors clarify the NFEs used during inference?
We apologize for omitting the number of sampling steps. Following previous works, we use 100 sampling steps () for sddbms during inference. We have now includes the setting in Appendix G of the revised manuscript.
Q3: Proposition 3.1 would benefit from a rewriting. As it is, the main idea is to show the singularity of diffusion bridges. I could understand the injection ( for any ) leads to the conclusion. However, in this version of Proposition 3.1, neither singularity nor the relation to diffusion bridges are emphasized.
Thank you for your helpful suggestion. We have appropriately rewritten Proposition 3.1 in the revised manuscript to explicitly emphasize the terminal singularities of diffusion bridges. Specifically, we first show that bounded drift coefficients yield a non-degenerate Gaussian terminal distribution that retains dependence on the initial state. We then prove that either complete source forgetting or terminal covariance collapse necessarily requires the linear drift coefficient to become unbounded near the terminal time. Since hard endpoint bridges simultaneously require complete source forgetting and terminal covariance collapse, they necessarily exhibit drift singularities at the terminal boundary. The corresponding proofs have also been updated in Appendix E.1.
Q4:Lack of quantitative evaluation for the ablation on hyperparameter . Although the authors have claimed that "different values lead to comparable performance", the specific results should be also provided to validate the claim, which would be more convincing.
Thanks for the helpful suggestion. we have conducted an additional quantitative ablation study on the hyperparameter for image deraining task. The results are presented in Table2.
Table 2. Effect of hyper-parameter on image deraining.
| PSNR ↑ | SSIM ↑ | LPIPS ↓ | FID ↓ | |
|---|---|---|---|---|
| 32.27 | 0.906 | 0.043 | 16.39 | |
| 32.38 | 0.906 | 0.043 | 15.68 | |
| default | 32.31 | 0.906 | 0.042 | 16.48 |
| 32.01 | 0.9041 | 0.045 | 18.19 |
Here, the "default" setting refers to a specific choice of and , with , that makes , rather than the value used by default in our experiments.
The results show that the performance remains relatively consistent across different values of , indicating that SDDBMs are not highly sentitve to hyperparameter within the proper range (compared to and ). In general, values close to or slightly blew the default setting tend to yield better performance, while larger values lead to a degradation in performance. In our experiments, we set for image deraining and adopt the default setting for both image inpainting and super-resolution.
Thank you for this insightful question. We apologize for the insufficient explanation of the terminal initialization strategy in the original manuscript. Below, we provide a detailed explanation of why omitting the terminal noise during inference is reasonable and mathematically justified.
Under our experimental setting μ=x⋆, the terminal distribution of SDDBMs is given by N(aTx0+(bT+cT)x⋆,σ2I). During inference, x0 is unavailable, and its coefficient aT is sufficiently small under our experimental parameter settings (∣aT∣≤0.0036). We therefore neglect the source-dependent contribution aTx0 and consider the approximate terminal distribution p~T(xT∣x⋆)=N((bT+cT)x⋆,σ2I). In image-to-image tasks, the degraded image x⋆ serves as guidance for the reverse process, which starts from a terminal state xT and aims to recover the corresponding clean image x0. Therefore, we need to determine a terminal state xT that best represents the terminal distribution. Under the expected squared error criterion, the optimal choice is xT=(bT+cT)x⋆.
Specifically, let m=(bT+cT)x⋆ and xT∼N(m,σ2I). We seek the terminal state z that minimizes the expected squared error with respect to the terminal random variable XT:
z⋆=argminzEp~T[∣∣XT−z∣∣22] Using the Gaussian reparameterization xT=m+σϵ, where ϵ∼N(0,I), and noting that E[ϵ]=0 and E[∣∣ϵ∣∣22]=d, we obtain
\left[ ||\mathbf{X}_T-\mathbf{z}||_2^2 \right] =||\mathbf{z}-\mathbf{m}||_2^2+d\sigma^2,$$ where $d$ denotes the dimensionality of the image. Since $d\sigma^2$ is independent of $\mathbf{z}$, the expected squared error is uniquely minimized when $\mathbf{z}=\mathbf{m}$. Therefore, we initialize the reverse sampling process with $\mathbf{x}_T=(b_T+c_T)\mathbf{x}^\star$, which is the optimal point estimate of the terminal distribution under the expected squared error criterion. We have also provided a detailed explanation of this initialization strategy in the revised manuscript. We hope this clarification addresses your question, and we welcome any further discussion.