- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
Neural solvers for networked systems usually require retraining when topology changes, yet real changes typically affect only a few links. RecastNet exploits this by treating such changes as low-rank updates to the network Laplacian and handling them analytically rather than learning them. It first recovers the base-topology Laplacian inverse from the input–output Jacobian of a frozen neural solver, then combines it with the low-rank change to produce solutions for unseen topologies without retraining or fine-tuning. For linear systems, only directional inverse actions along the changed links are needed, keeping computation low-dimensional. For nonlinear systems, an auxiliary solver trained on Newton-step pairs enables zero-shot nonlinear recasting. The authors provide error bounds and recastability conditions, and validate on synthetic networks and real power grids, reporting accurate zero-shot adaptation and 18–150× inference speedups over numerical solvers.
Critical strengths
-
The paper addresses a real and widely encountered problem. In networked infrastructure, topology changes are often localized, while training neural solvers to cover many possible configurations can be expensive and may generalize poorly to unseen topologies. Moving topology adaptation from training to an analytical correction is an appealing and well-motivated perspective.
-
The solution-to-inverse insight in Theorem 1 is particularly elegant. It is architecture-agnostic and shows how a trained neural solver can provide useful structural information beyond its solution predictions. Combining this insight with low-rank topology updates and the SMW formula leads to a coherent method with substantial theoretical support.
-
The experiments cover a broad range of settings, including different network sizes, three synthetic graph families, four power-grid benchmarks, several nodal relations, and topology changes of varying severity. This breadth makes the empirical evaluation convincing within the tested regimes.
Critical weaknesses
- The paper lacks discussion of how the method degrades under high-rank or large-scale topology changes.
Questions for authors
-
As the rank r of the topology change grows (e.g., network partitioning, simultaneous failure of many links), how do RecastNet's error and computational cost change? Is there a characterizable "recastable rank limit" or break-even point? Could the authors provide experimental or theoretical analysis across different r/n ratios?
-
If the base-topology neural solver itself has limited accuracy, how does recasting performance degrade?
6 comments
Part[1/2] Response to Reviewer 6dhi 1.3
(III) Experimental evaluation under increasing topology-change ranks.
Next, we present new experimental results to complement the above rank-dependent error bound and complexity analysis. We evaluate accuracy and runtime on the 250-node ER network as the topology-change rank increases from to , where is the rank of the topology update and . All cases use the same frozen base-topology neural solver. Prediction accuracy is evaluated against the true target-topology solution.
Accuracy. The recast error grows gradually as the topology-change rank increases. As shown in the table below, raising from to --a fifteenfold increase in rank--raises the median error from to and the P95 error from to , corresponding to roughly a error increase in both cases. The aforementioned theoretical error bound characterizes this dependence through rank-dependent error accumulation and topology-dependent amplification. In these experiments, their combined effect leads to a modest increase in error, with the median error remaining below even at .
| RecastNet error (%) Median | RecastNet error (%) P95 | Base solver error (%) Median | Base solver error (%) P95 | RecastNet time (ms/case) | Numerical solver time (ms/case) | Speedup | ||
|---|---|---|---|---|---|---|---|---|
| 1 | 1.35 | 5.72 | 1.72 | 13.64 | 0.045 | 1.739 | ||
| 5 | 1.39 | 5.78 | 1.81 | 7.49 | 0.221 | 1.750 | ||
| 10 | 1.48 | 6.21 | 3.64 | 16.65 | 0.309 | 1.750 | ||
| 20 | 1.60 | 6.51 | 3.97 | 15.06 | 0.600 | 1.752 | ||
| 30 | 1.79 | 7.24 | 4.95 | 18.46 | 0.892 | 1.744 | ||
| 40 | 1.88 | 7.64 | 14.67 | 44.78 | 1.203 | 1.683 | ||
| 50 | 1.87 | 7.79 | 14.51 | 32.63 | 1.607 | 1.711 | ||
| 62 | 2.36 | 9.71 | 15.90 | 44.37 | 2.179 | 1.718 | ||
| 75 | 2.55 | 10.38 | 19.70 | 44.68 | 2.702 | 1.663 |
Computational cost. We use a batch size of , consistent with the original paper, and report average inference time per case, including data transfers. As increases from to , the average runtime of RecastNet increases from to ms. RecastNet remains comparable to the numerical solver at ( versus ms/case), while the numerical solver is faster at and . The measured runtime crossover therefore lies between the sampled and levels under this implementation and hardware setup.
This behavior is consistent with the complexity analysis: when considering fixed , the matrix-operation cost is . Increasing therefore gradually increases the cost to , and the correction systems are no longer much smaller than the original system. The computational benefit consequently diminishes as the update rank grows, as is also the case for classical SMW updates. The original runtime experiments focus on the low-rank regime motivated by localized outages and network reconfiguration. The additional results clarify that, although recasting maintains a median error below through , its runtime advantage does not extend across this entire range. We thank the reviewer for prompting this broader evaluation and have clarified this distinction in the revised manuscript.
Part[2/2] Response to Reviewer 6dhi 2.1 & 2.2
Question 2 If the base-topology neural solver itself has limited accuracy, how does recasting performance degrade?
We thank the reviewer for this question. in the original submission characterizes how the base-topology neural solver's approximation errors propagate through recasting. Below, we add additional experiments to study how performance degrades as the base predictions become less accurate.
(I) Error propagation from the base neural solver.
Corollary 3 bounds the recast-solution error as where is the base prediction error and is the directional inverse-action recovery error. The first term captures the propagation of the base prediction error, while the second accounts for inaccuracies in the inverse information recovered from the neural solver. The coefficients and characterize how these errors are amplified under a particular topology change. We examine the resulting effect on recasting accuracy experimentally below.
(II) Recasting with increasingly inaccurate base predictions.
We use the IEEE 118-bus test system and evaluate DC power flow as the power injections progressively move beyond the training range of the frozen base-topology neural solver. This increases the base prediction error from to . At each input, we evaluate the same ten randomly sampled outages in each of the medium- and high-impact groups.
The results show a gradual loss of recasting accuracy as the base prediction becomes less accurate. In these tested cases, median recast errors remain close to the base error even when it exceeds , while high-impact outages produce greater additional error, particularly in P95. Thus, recasting can accommodate topology changes despite substantial base prediction errors, but its accuracy still depends on the quality of the base approximation.
| Base-topology error | Recast error (medium-impact outages) Median | Recast error (medium-impact outages) P95 | Recast error (high-impact outages) Median | Recast error (high-impact outages) P95 |
|---|---|---|---|---|
Part[1/2] Response to Reviewer 6dhi 1.1 & 1.2
We thank the reviewer for the constructive feedback and positive remarks on "a real and widely encountered problem," the "particularly elegant" solution-to-inverse insight, "substantial theoretical support," and the "breadth" of our experiments. We address each comment below and have highlighted the manuscript revisions in blue. We hope these revisions address your concerns.
We thank the reviewer for this important question. The original paper reports low recasting errors on ER, SBM, and BA networks with up to 20 of the edges removed (see Figure 3, Page 7 and Table 1, Page 8). Below, we extend this study in three aspects: a rank-dependent error bound that characterizes conditions for a "recastable rank limit," a rank-dependent computational complexity analysis, and experiments evaluating prediction error and runtime with up to 30 of the edges removed.
(I) Effect of topology-change rank on recasting error.
Eq. (8), Corollary 3, Page 5, bounds the recast-solution error by two terms: the contribution from the base prediction error ε0 and the contribution from the directional inverse-action recovery error EH,Δ. The latter measures how accurately the neural solver recovers the base-network inverse along the topology-change directions.
We extend this bound to make the dependence on the topology-change rank r explicit. If each recovered inverse-action column has error at most δH, then: ∥EH,Δ∥2≤∥EH,Δ∥F≤rδH. Substituting this relationship into Eq.(8) gives: ∥xG−xG∥2≤c0,Δ∥ε0∥2+cH,Δr,δH The bound identifies a square-root dependence on the topology-change rank: the inverse-action error contribution grows as r, scaled by the topology-dependent amplification coefficient cH,Δ. It therefore characterizes how increasing rank can degrade recasting accuracy, while accounting for differences among topology changes. We provide the derivation in Appendix C.2.3, Page 22 of the revised manuscript.
Is there a recastable rank limit? Our analysis suggests that there may not be a universal recastable rank limit. For a prescribed tolerance ϵ, requiring the right-hand side of the bound to be at most ϵ∥xG∥2 gives a sufficient condition for ϵ-recastability. However, equal-rank topology changes can produce substantially different errors because they affect different links and amplify approximation errors differently. In an extreme case, even a rank-one change that removes a bridge edge can disconnect the network and make LG singular. Thus, a rank threshold alone may not guarantee that recasting meets a prescribed accuracy tolerance.
(II) Effect of topology-change rank on computational cost.
For a rank-r topology change, Proposition 2 reduces inverse-action recovery and topology correction to linear solves of dimensions at most q:=psh+r and r, respectively. The inference cost of RecastNet is thus: Csingle(n,r)=Cfwd+Csens+O(rCJVP+nq2+q3+r3),
where Cfwd, Csens, and CJVP denote the costs of the base prediction, sensitivity evaluation, and one neural Jacobian-vector product (JVP), respectively. The nq2+q3 terms account for inverse-action recovery, and r3 accounts for the topology-correction solve. We add this complexity analysis to Appendix C.2.1, Page 21.