- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes two hybrid methods for tabular data imputation, NuclearForest and SoftForest, which combine low-rank matrix completion with a single Random Forest refinement pass. NuclearForest employs a modified Singular Value Thresholding (SVT) algorithm with column-mean warm starting and an adaptive step-size rule, while SoftForest uses the existing SoftImpute algorithm for initialization. The central motivation is that a structured low-rank initialization can reduce the need for iterative Random Forest refinement while maintaining competitive imputation accuracy. The authors evaluate the proposed methods against classical imputation baselines on two real-world datasets under different missingness settings, using reconstruction, structural, statistical, downstream predictive, and runtime metrics. The results demonstrate substantial computational savings relative to iterative MissForest in the evaluated settings, although the proposed methods do not consistently achieve the best imputation accuracy across all metrics and missingness mechanisms.
Critical strengths
1.The proposed methods substantially reduce computational runtime compared with iterative MissForest. 2.The paper provides a comprehensive evaluation of imputation quality using complementary metrics, component ablation studies, and supplementary runtime analyses.
Critical weaknesses
- The methodological novelty is limited, and the adaptive step-size mechanism lacks sufficient justification.
- Despite the authors' explanation, the lack of comparisons with neural-network-based imputation methods weakens the claim of broad robustness and competitiveness.
Questions for authors
1.Why do the authors change the conventional SVT step size from (1.2/p) to (1.2p)? What is the theoretical justification for this choice, and what advantages does it offer over other admissible initial step sizes? 2.How were the contraction and expansion factors of 0.9 and 1.05 determined? Could the authors provide justification for these choices or a corresponding sensitivity analysis? 3.The authors state that no step-size contractions were triggered during the experiments, making the adaptive update equivalent to a predetermined multiplicative growth schedule. What additional benefits does the residual-driven mechanism offer over a simpler deterministic schedule? 4.The convergence proof is established for a zero-initialized SVT iteration, whereas the implemented algorithm uses column-mean warm starting. Does the convergence guarantee also apply to the implemented warm-started algorithm?
5 comments
Reviewer: Q2. How were the contraction and expansion factors of 0.9 and 1.05 determined? Could the authors provide justification for these choices or a corresponding sensitivity analysis?
We thank the reviewer for this suggestion and have added a sensitivity analysis to the revised manuscript. The constants 0.9 and 1.05 are common multiplicative adaptation factors. They do not affect the convergence guarantee. The proof of Proposition 1 extends directly to any contraction factor and any expansion factor above 1, provided only finitely many contractions occur. Only the contraction factor enters the result, through the lower bound . The sensitivity analysis covers all missing rates (10–90%), 10 repeats and four settings (metabolomics MCAR/MAR and MNAR, housing MCAR and MAR).
For initial step and expansion factor, we tested initial steps of , , , , , and , and expansion factors of 1.01, 1.02, 1.05, 1.1 and 1.2. In both cases, the NRMSE changed by at most 0.002 at any missing rate in any setting. On metabolomics, the step grows geometrically to the cap within 3–52 of the 1000 iterations (11 with the paper's factor 1.05), so all variants run with essentially the same step. On housing, SVT converges within about 11 iterations to the unique solution of the same convex problem, regardless of the step.
Table A. Sensitivity to the initial step size (cap 2p, expansion factor 1.05, contraction factor 0.9). NRMSE averaged over 10 repeats.
| Initial step | 10% | 20% | 30% | 40% | 50% | 60% | 70% | 80% | 90% |
|---|---|---|---|---|---|---|---|---|---|
| Metabolomics, MCAR/MAR | |||||||||
| 0.5p | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.126 | 1.250 | 1.377 | 1.325 |
| 0.8p | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.378 | 1.326 |
| 1.0p | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.378 | 1.326 |
| 1.1p | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.378 | 1.326 |
| 1.2p (paper) | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.378 | 1.327 |
| 1.5p | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.379 | 1.327 |
| 1.9p | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.379 | 1.327 |
| Max. difference | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.001 | 0.001 | 0.002 | 0.002 |
| Metabolomics, MNAR | |||||||||
| 0.5p | 2.277 | 2.687 | 2.626 | 2.633 | 2.820 | 2.865 | 2.981 | 3.176 | 3.333 |
| 0.8p | 2.277 | 2.687 | 2.627 | 2.633 | 2.821 | 2.865 | 2.981 | 3.176 | 3.334 |
| 1.0p | 2.277 | 2.687 | 2.627 | 2.633 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.1p | 2.277 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.2p (paper) | 2.277 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.5p | 2.277 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.9p | 2.277 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| Max. difference | 0.001 | 0.000 | 0.001 | 0.001 | 0.001 | 0.001 | 0.001 | 0.001 | 0.001 |
| Housing, MCAR | |||||||||
| 0.5p | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.005 | 1.004 | 1.003 | 1.007 |
| 0.8p | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.006 | 1.004 | 1.003 | 1.007 |
| 1.0p | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.006 | 1.005 | 1.004 | 1.008 |
| 1.1p | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.006 | 1.005 | 1.004 | 1.008 |
| 1.2p (paper) | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.007 | 1.005 | 1.004 | 1.008 |
| 1.5p | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.007 | 1.005 | 1.004 | 1.008 |
| 1.9p | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.007 | 1.005 | 1.005 | 1.008 |
| Max. difference | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.001 | 0.002 | 0.002 | 0.001 |
| Housing, MAR | |||||||||
| 0.5p | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.036 | 1.058 | 1.081 | 1.087 |
| 0.8p | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.058 | 1.080 | 1.086 |
| 1.0p | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.086 |
| 1.1p | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.086 |
| 1.2p (paper) | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.085 |
| 1.5p | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.085 |
| 1.9p | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.085 |
| Max. difference | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.001 | 0.001 | 0.002 |
We continue our response to Q2 here.
Table B. Sensitivity to the expansion factor (initial step 1.2p, cap 2p, contraction factor 0.9). NRMSE averaged over 10 repeats.
| Expansion factor | 10% | 20% | 30% | 40% | 50% | 60% | 70% | 80% | 90% |
|---|---|---|---|---|---|---|---|---|---|
| Metabolomics, MCAR/MAR | |||||||||
| 1.01 | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.126 | 1.250 | 1.377 | 1.325 |
| 1.02 | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.378 | 1.326 |
| 1.05 (paper) | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.378 | 1.327 |
| 1.1 | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.379 | 1.327 |
| 1.2 | 0.736 | 0.810 | 0.874 | 0.938 | 1.027 | 1.127 | 1.251 | 1.379 | 1.327 |
| Max. difference | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.001 | 0.001 | 0.001 | 0.002 |
| Metabolomics, MNAR | |||||||||
| 1.01 | 2.277 | 2.687 | 2.626 | 2.633 | 2.820 | 2.865 | 2.981 | 3.176 | 3.333 |
| 1.02 | 2.277 | 2.687 | 2.627 | 2.633 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.05 (paper) | 2.277 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.1 | 2.277 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| 1.2 | 2.278 | 2.687 | 2.627 | 2.632 | 2.821 | 2.865 | 2.982 | 3.176 | 3.334 |
| Max. difference | 0.001 | 0.000 | 0.000 | 0.001 | 0.001 | 0.001 | 0.001 | 0.000 | 0.001 |
| Housing, MCAR | |||||||||
| 1.01 | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.006 | 1.005 | 1.004 | 1.007 |
| 1.02 | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.006 | 1.005 | 1.004 | 1.008 |
| 1.05 (paper) | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.007 | 1.005 | 1.004 | 1.008 |
| 1.1 | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.007 | 1.005 | 1.004 | 1.008 |
| 1.2 | 1.019 | 1.013 | 1.013 | 1.011 | 1.007 | 1.007 | 1.005 | 1.004 | 1.008 |
| Max. difference | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.001 | 0.001 | 0.000 |
| Housing, MAR | |||||||||
| 1.01 | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.086 |
| 1.02 | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.086 |
| 1.05 (paper) | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.085 |
| 1.1 | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.085 |
| 1.2 | 1.036 | 1.024 | 1.021 | 1.024 | 1.027 | 1.035 | 1.057 | 1.080 | 1.085 |
| Max. difference | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.001 |
For the contraction factor, on the real datasets no contraction is triggered, so this factor does not influence the reported results. It acts as a safeguard when the step is too large. To test it, we started from Cai et al.'s step size and tested contraction factors of 0.5, 0.7, 0.9 and 0.95. With 0.5, 0.7 and 0.9, every one of the 360 runs recovered and converged. With 0.95, 52 of 360 runs failed, presumably because the steps shrink too slowly. Among the factors that worked, 0.9 gave the lowest error on metabolomics MCAR with 90% missing data (NRMSE 1.638, compared with 1.659 for 0.7 and 1.679 for 0.5). We note that the differences are small, with 0.9 providing competitive performance among the evaluated settings.
Table C. Sensitivity to the contraction factor when the initial step is too large (SVT started from , 1000 iterations, all 360 masks). In the paper setting no contraction occurs, so the factor has no effect there. A run fails if its residual becomes non-finite or does not fall below 1.
| Contraction factor | 0.5 | 0.7 | 0.9 (paper) | 0.95 |
|---|---|---|---|---|
| Failed runs | 0/360 | 0/360 | 0/360 | 52/360 |
| NRMSE, metab. MCAR 90% | 1.679 | 1.659 | 1.638 | diverges |
We added these three tables to the Appendix of the revised manuscript.
We appreciate the time and care the reviewer has devoted to our work. Your comments have helped us refine our ideas and improve the presentation. We have incorporated your suggestions into the revised manuscript, with all changes highlighted in green for ease of review.
We thank the reviewer for this observation. The step size was changed to guarantee admissibility. We keep Cai et al.'s constant 1.2, replace 1/p by p, and cap the step at 2p, which keeps every step below 2. The results are insensitive to this constant, because initial steps from 0.5p to 1.9p change the NRMSE by at most 0.002 (see Q2).
Cai et al. chose 1.2/p for large, randomly generated, low-rank matrices (up to 30,000 × 30,000) observed at low sampling ratios. Their heuristic justification is that, under random sampling and suitable incoherence assumptions, the squared Frobenius norm of the observed part of a fixed low-rank matrix is approximately p times that of the full matrix. This motivates a step size proportional to 1/p to accelerate convergence. However, Cai et al. explicitly acknowledge that this argument is heuristic.
In contrast, the real metabolomics data used in our paper are medium-sized and approximately low-rank (after standardization, the top 10 principal components explain 65% of the variance). In this setting, the 1/p compensation is less suitable, and a large step can make SVT diverge, and even an admissible step close to 2 can make SVT fit the noise in the observed entries within a fixed iteration budget. We added a statement to the Appendix of the revised manuscript to clarify this distinction.
With respect to the theoretical justification, we make the following observations. Cai et al.'s convergence theorem guarantees convergence for step sizes below 2. The value 1.2/p exceeds 2 as soon as more than 40% of the entries are missing (p<0.6), so the guarantee does not hold there. Our steps start at 1.2p and are capped at 2p, which is always below 2. In fact, 2p is the largest step proportional to p that is admissible for every p.
Our experiments are consistent with this. With Cai et al.'s step 1.2/p, SVT diverged whenever 50% or more of the data were missing in the MCAR/MAR settings (150 of 270 runs in three settings). The residual overflowed at 60–90% missingness (120 runs) and grew above 1088 within 1000 iterations at 50% (30 runs). It converged whenever the step stayed below 2.
We have also added a statement to the Introduction and the Appendix of the revised manuscript to make this argument explicit.
Within the admissible range, the choice of the initial step has essentially no effect. Initial steps from 0.5p to 1.9p change the NRMSE by at most 0.002 at any missing rate in all four settings. The advantage of the rule is therefore that it needs no tuning of the initial step. Details are given in our answer to Q2.