- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes a diffusion-based data imputation method called DoobImp that does not require artificial masking during training. It learns the unconditional distribution of the data and applies observational conditions at the inference stage using Doob's h-transform to recover missing data; meanwhile, through iterative imputation and retraining, the model is able to learn directly from incomplete data.
Critical strengths
- The ideal Doob transform has a clear probabilistic interpretation, whereas the practical algorithm uses Tweedie estimation, posterior precision approximation, and the Jacobian to convert the theory into a computable guidance term.
- The experiments are sufficient and, from the perspective of performance comparison, provide relatively strong support for the effectiveness of DoobImp.
Critical weaknesses
- The motivation may overstate the novelty of mask-free diffusion imputation. NewImp and DiffPuter have already explored imputation without additional artificial masking.
- There is no sufficient direct comparison with existing conditional sampling methods such as DPS and RePaint under the same unconditional backbone.
Questions for authors
- What specific limitations of existing mask-free imputation methods, particularly NewImp and DiffPuter, motivate the proposed framework? (Weakness 1)
- Can the authors demonstrate whether the advantage of DoobCond actually comes from the special properties of Doob's h-transform, or merely from adding an observation-consistency gradient to the reverse diffusion? (Weakness 2)
4 comments
Response to Reviewer NKAe (Part 2/2)
Q2. Can the authors demonstrate whether the advantage of DoobCond actually comes from the special properties of Doob's h-transform, or merely from adding an observation-consistency gradient to the reverse diffusion?
A2. Doob's -transform plays a key role in constructing DoobCond's conditioning term, beyond simply adding an observation-consistency gradient. Lemma 1 identifies the conditioning term required by Doob's -transform. In the Gaussian case, Equation (8) specifies how the posterior precision weights the observation residual and how the denoiser Jacobian maps it to the missing coordinates. Equations (9)–(10) provide the practical approximation used by DoobCond, giving the update a probabilistic basis. In Table 37 of the manuscript (lines 2453–2465), both samplers retain an observation-consistency gradient, allowing us to isolate the contribution of the precision weighting derived from Doob's -transform.Replacing temperature-only weighting (which already includes an observation-consistency gradient) with the diagonal precision derived from Doob's -transform reduced mean MAE from 0.291 to 0.272 and RMSE from 0.548 to 0.518. All three datasets improved.
For comparison with an established method, we additionally evaluated DoobCond against DPS [4], which also differentiates an observation residual through the denoiser. On the same frozen DoobFit checkpoints, DoobCond lowered both MAE and RMSE in 27/27 pairs. Each cell below gives mean MAE / RMSE across MCAR, MAR, and MNAR at 30% missingness, with three seeds per mechanism.
| Dataset | DoobCond | DPS [4] |
|---|---|---|
| Letter | 0.284 / 0.445 | 0.451 / 0.643 |
| Gesture | 0.263 / 0.563 | 0.358 / 0.759 |
| News | 0.262 / 0.533 | 0.423 / 0.768 |
The DPS comparison further demonstrates the advantage of DoobCond's full conditioning update over a sampler that already uses an observation-consistency gradient in these settings. A3 below provides the results by dataset and mechanism, including RePaint.
Q3. There is insufficient direct comparison with existing conditional sampling methods such as DPS and RePaint under the same unconditional backbone.
A3. We additionally evaluated DPS [4] and the RePaint-style sampler [2, 3] on exactly the same frozen DoobFit checkpoints as DoobCond. RePaint here is DiffPuter's inference implementation. We held the test samples, split, 30% mask, and normalization fixed and averaged 20 draws for each point prediction. Each cell reports mean MAE / RMSE over three seeds.
| Dataset | Mechanism | DoobCond | DPS [4] | RePaint-style [2, 3] |
|---|---|---|---|---|
| Letter | MCAR | 0.270 / 0.423 | 0.447 / 0.635 | 0.280 / 0.441 |
| Letter | MAR | 0.297 / 0.465 | 0.452 / 0.644 | 0.301 / 0.474 |
| Letter | MNAR | 0.284 / 0.447 | 0.454 / 0.650 | 0.292 / 0.461 |
| Gesture | MCAR | 0.255 / 0.549 | 0.355 / 0.751 | 0.274 / 0.808 |
| Gesture | MAR | 0.263 / 0.559 | 0.350 / 0.744 | 0.271 / 0.574 |
| Gesture | MNAR | 0.270 / 0.581 | 0.370 / 0.783 | 0.281 / 0.610 |
| News | MCAR | 0.263 / 0.531 | 0.425 / 0.763 | 0.281 / 0.551 |
| News | MAR | 0.261 / 0.525 | 0.418 / 0.759 | 0.276 / 0.660 |
| News | MNAR | 0.263 / 0.544 | 0.425 / 0.781 | 0.289 / 1.975 |
DoobCond achieved lower MAE and RMSE than DPS in 27/27 matched dataset–mechanism–seed comparisons and lower MAE and RMSE than RePaint in 27/27. This comparison isolates the contribution of the conditional sampling method under the same backbone, supporting DoobCond's superior performance in imputation tasks.
Thank you again for reviewing our paper. We hope these responses address your concerns, and we welcome further discussion.
References
[1] Zhichao Chen et al. Rethinking the Diffusion Models for Missing Data Imputation: A Gradient Flow Perspective. NeurIPS, 2024.
[2] Hengrui Zhang et al. DiffPuter: Empowering Diffusion Models for Missing Data Imputation. ICLR, 2025.
[3] Andreas Lugmayr et al. RePaint: Inpainting Using Denoising Diffusion Probabilistic Models. CVPR, 2022.
[4] Hyungjin Chung et al. Diffusion Posterior Sampling for General Noisy Inverse Problems. ICLR, 2023.
Thank you for the clarifications and additional experiments. The discussion of NewImp and DiffPuter better explains the method’s positioning, while the matched-backbone comparisons with DPS and RePaint strengthen the empirical support for DoobCond. These responses address my concerns to a reasonable extent. I maintain my original rating of 3 (Weak Acceptance).
Thank you for your positive feedback and for acknowledging our clarifications and additional experiments. We sincerely appreciate your thoughtful review and constructive comments throughout the discussion.
Response to Reviewer NKAe (Part 1/2)
We thank you for the time you spent reviewing our paper and for recognizing our framework based on Doob's h-transform and the strong empirical performance on a variety of imputation tasks. We also appreciate your detailed questions, which help us improve the completeness and clarity of the paper. We now address each question in turn.
A1. Thank you for helping us clarify this point. In the first place, we do not claim to be the first to propose mask-free diffusion imputation models. We acknowledge prior work in the manuscript (lines 76–79; lines 101–104). Instead, our motivation starts from missingness distribution shift, which we address through DoobCond complemented by DoobFit. This is supported by the experiments in Appendix C.5 (Figure 8) and the real-data experiments below. We now explain in detail how our formulation based on Doob's h-transform and its adaptation to imputation distinguish DoobImp from NewImp [1] and DiffPuter [2].
NewImp. Algorithm 1 of NewImp [1] alternates between fitting a joint score to the current completed matrix and updating its missing entries by a Wasserstein gradient flow, with a negative-entropy term that favors more concentrated imputations. In contrast, DoobCond provides reusable conditioning: given a frozen unconditional diffusion backbone, DoobCond takes a new sample's observed values and mask and draws its missing entries without updating the backbone (Algorithm 1). DoobFit calls the same sampler to complete its training samples before the next refit (Algorithm 2). NewImp's authors also note that their kernel computation scales roughly quadratically with the number of samples and that their real-valued support assumption limits direct application to categorical variables [1, Appendix G.1]. DoobCond computes its update from the current sample through the frozen denoiser (Equation (10)), without evaluating a kernel over pairs of samples. Our tabular implementation handles categorical variables by binary encoding and decodes the imputed values for evaluation (Appendix D.1; Table 1).
DiffPuter. Our main difference from DiffPuter [2] lies in the conditioning term. DiffPuter's Theorem 1 characterizes conditional sampling in the limit Δt→0. In the proof, Equation (11) of [2, Appendix B.1] makes the following approximation inside the reverse transition, where y denotes the clean observed entries:
\approx p_\theta(x_{t-\Delta t}^{\mathrm{mis}}\mid x_t).$$ The right-hand side **drops the explicit dependence on $y$**. Instead, using Doob's $h$-transform, our Lemma 1 derives how $y$ changes the reverse drift through $\nabla_x\log h_\eta(x,t)$ (Equation (6)). DoobCond approximates its missing-coordinate component: the denoiser Jacobian maps the precision-weighted discrepancy between $y$ and its predicted value back to the missing entries (Eqs. (8)–(10)). Holding DiffPuter's own weights fixed, switching from its default RePaint-style sampler to DoobCond lowered both errors in 27/27 pairs. The table below shows this comparison on frozen DiffPuter weights. Cells give mean MAE / RMSE across MCAR, MAR, and MNAR at 30% missingness, with three paired seeds per mechanism. | Dataset | RePaint [2, 3] | DoobCond | |---|---:|---:| | Letter | 0.330 / 0.510 | 0.322 / 0.495 | | Gesture | 0.322 / 0.760 | 0.307 / 0.687 | | News | 0.299 / 0.784 | 0.278 / 0.573 | These results demonstrate **consistently lower imputation errors with DoobCond in the tested settings**. To reconcile reverse updates of the missing block with refreshed observations, DiffPuter's implementation runs 20 resampling rounds at each nonterminal noise level. DoobCond retains the explicit dependence on $y$ through its conditioning term derived from Doob's $h$-transform and applies this term before refinement. On Letter at 30% MCAR, our default sampler used five same-level refinement steps and achieved a lower measured cost: it took 0.555 ms per sample per posterior draw versus 2.171 ms for DiffPuter on the same RTX A6000. A complete training iteration averaged 9.618 min for DoobImp versus 16.436 min for DiffPuter (Table 42-43). The tabular benchmark in the manuscript gives a second comparison in which both methods reuse their weights as missingness changes. DoobImp had lower average MAE and RMSE in 9/9 training–test mechanism combinations (Appendix E.3). For models trained at 30% missingness and tested at 50%, Table 40 reports MAE / RMSE of 0.341 / 0.623 for DoobImp and 0.393 / 4.546 for DiffPuter. These results further demonstrate DoobImp's stronger imputation performance under the tested changes in missingness mechanism and rate.