- Rating
- 3
- Confidence
- 1
- Informativeness
- 3
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This manuscript addresses the generalization challenge of radio frequency fingerprint recognition in practical applications, with a particular focus on the significant drop in recognition performance in unseen environments. To tackle this issue, the authors reformulate the source-only RF domain generalization problem through decision-evidence geometry and demonstrate that the margins between true and rival classes constitute decision-complete evidence coordinates. Within these coordinates, each identity occupies a loose region composed of decision-equivalent states, rather than a single point. Ultimately, the authors introduce a MarginCell regularizer, which constructs relative evidence coordinates using the class mean spacing within a mini-batch and combines Dispersion Reward and Boundary Barrier through a unified potential function. Experiments conducted on three public datasets, WiSig, UCLA-CORES, and LoRa, demonstrate that this method achieves significant performance improvements over existing approaches.
Critical strengths
Unlike conventional patchwork papers, this manuscript is grounded in the geometric completeness of pairwise intervals at the classifier output; it analyzes the anisotropic curvature contraction effect of the cross-entropy Hessian matrix, offering a strong theoretical foundation. Furthermore, the manuscript presents an extensive range of experiments.
Critical weaknesses
Why does intentionally introducing a negative relative shift () in the source domain actually enhance the reserves for the target domain?
Questions for authors
(1) When applying symmetric dispersion rewards under the zero-mean constraint, will it cause half of the samples to shift in the negative direction and reduce their bottleneck decision-making reserves?
(2) The theorem in the method section assumes that the class mean interval is a fixed constant. However, in Sections 4.1 and 4.3, the author specifically emphasizes that Notably, the minibatch reference is computed dynamically and remains differentiable.
(3) Classification errors only occur when , and a positive offset represents a stronger classification margin than the class mean. How will performance change if an asymmetric potential function is adopted?
(4) The author can consider the performance changes under different batch sizes |B|, that is, the sensitivity of the number of samples per class within the small batch and the class scalability.
5 comments
Q2. Class-Level vs. Minibatch Reference
We appreciate this careful reading and are glad to clarify how the theoretical analysis in Section 3 connects to the dynamic implementation in Section 4. Rather than requiring to stay constant across training epochs, the phrase "at fixed class-mean margins" in Section 3 means holding the class-mean level fixed at a given classifier state to study mean-preserving deviations around it ( with ). As noted in the Remark of Appendix A.3, Theorem 3.3 characterizes the local decision geometry around the current reference state; since the classifier changes at every step, training tracks this reference online through the minibatch mean. The minibatch coordinates satisfy exactly, so Theorem 3.3 holds at the current reference at every step. Theorems 3.2 and 3.4 use only as a positive scale: Theorem 3.2 holds for any positive reference, and Theorem 3.4, which concerns the trained model, requires only that source and target share the same positive reference. In our experiments, this reference is computed once as the final model's class-mean margins over the entire source training set.
Differentiability of the reference does not alter this analysis either. Moving margins along a mean-preserving direction leaves the minibatch mean unchanged, so backpropagating through the reference introduces no first- or second-order term along these directions. The local analysis, including the release threshold , is therefore identical to the fixed-reference case (full proof added to Appendix A.3 of the revised manuscript). We keep the reference differentiable because is invariant to a uniform rescaling of a class's margins (, up to ). With a differentiable reference, the regularizer has zero gradient along this rescaling direction, so it shapes only the relative evidence within a class and leaves the margin scale to cross-entropy. A detached reference would ignore this dependence: the regularizer would then produce a spurious gradient that rescales the margins without reducing the regularizer itself. Q4 reports the accuracy of the online estimate: the reference error is small at every batch size, and performance remains robust to it.
Q3. Performance under Asymmetric Potentials
Following this constructive suggestion, we evaluated asymmetric variants on UCLA-CORES, changing only the potential: the reward is either symmetric or positive-only, and the barrier either two-sided or on the negative side only. All variants underperform the symmetric design, especially on hard identities:
| Reward / barrier | Acc | WEnv | TIA@20 | Source coords (%) | Source coords (%) |
|---|---|---|---|---|---|
| Symmetric / two-sided (MarginCell) | 76.74±2.86 | 75.17±1.21 | 53.44±3.21 | 14.85 | 7.38 |
| Symmetric / negative side only | 74.23±5.15 | 72.20±4.83 | 43.45±9.87 | 25.44 | 30.78 |
| Positive-only / two-sided | 74.68±2.14 | 72.70±2.20 | 45.45±10.99 | 3.81 | 4.60 |
| Positive-only / negative side only | 74.85±5.00 | 73.72±4.50 | 43.26±11.86 | 7.99 | 24.51 |
The reviewer is right that classification errors occur only when , and the upper barrier does not treat positive offsets as errors. Yet both bounding and release need to act on both sides, and each of the first two variants isolates one of them.
- For bounding, coordinates are measured from the class mean, so positive and negative deviations must balance: unbounded positive deviations must be offset by deeper negative ones, pushing the negative tail toward the decision boundary, while the extra positive margin adds no reserve, since reserve is governed only by the weakest rival, . Removing the upper barrier confirms this: positive excursions rise from 7.38% to 30.78%, negative excursions follow from 14.85% to 25.44%, and TIA@20 drops by 10.0 points.
- For release, CE contracts evidence toward a single point from every direction, and Theorem 3.3 requires the reward to counteract this contraction in every direction. With both barriers kept, a positive-only reward leaves the negative side contracted; since couples positive and negative offsets, this stalls the release on both sides (fewest deviations: 3.81% and 4.60%). The evidence thus stays nearly point-like and, as in W1, relies on a few source-stable directions that unseen shifts can perturb, dropping mean TIA@20 by 8.0 points, so the loss comes from unreleased freedom rather than from unsafe deviations.
Bounding both sides and releasing both sides are therefore each necessary, which is exactly what the symmetric cell of Theorem 3.2 provides.
Q4. Batch Sensitivity
We appreciate this practical suggestion. To test the sensitivity to and to the per-class count , we swept from 512 to 32 on WiSig (2→2 d.), reducing the median from 85 to 5 (each cell reports ERM / MarginCell):
| Median | Reference deviation (%) | Acc | WEnv | TIA@20 | |
|---|---|---|---|---|---|
| 32 | 5 | 9.65 | 88.08/94.32 | 87.17/91.67 | 77.75/83.35 |
| 64 | 11 | 6.67 | 84.58/94.70 | 82.50/92.80 | 70.25/84.55 |
| 128 | 21 | 4.58 | 83.25/98.17 | 83.17/96.50 | 50.00/94.70 |
| 200 | 33 | 3.62 | 86.83/97.07 | 86.33/95.30 | 61.00/91.50 |
| 256 | 43 | 3.18 | 83.50/95.15 | 83.00/91.33 | 51.00/85.70 |
| 400 | 67 | 2.52 | 85.58/93.98 | 82.67/90.53 | 57.50/82.80 |
| 512 | 85 | 2.21 | 82.75/91.83 | 82.17/89.13 | 49.00/75.85 |
As shrinks, the deviation of the minibatch reference from the full-source reference grows from 2.21% to 9.65%, yet MarginCell consistently outperforms ERM in Acc, WEnv, and TIA@20 at every batch size (by 6.2–14.9, 4.5–13.3, and 5.6–44.7 points), maintaining 94.32% accuracy even at . This is comparable to the per-class regime of the 58-class UCLA-CORES at (, 8.70% reference deviation), where MarginCell still exceeds ERM by 11.4 Acc points. Complementing the analysis in Q2, this shows empirically that the gain of MarginCell is robust to the batch size and the resulting reference error. We have added the full results and analysis to Appendix C.8 in the revised manuscript.
Thanks so much for your support and insightful suggestions for improving our work! We are eager to address any remaining concerns you might have.
Dear Reviewer hEFM,
Thank you for your constructive review. We are sincerely grateful for your recognition of the theoretical grounding of our analysis, including the curvature contraction of the cross-entropy Hessian, and the extensiveness of our experiments. We provide a point-by-point response to each point below; revisions in the manuscript are highlighted in blue.
We appreciate this insightful question and are glad to clarify the underlying intuition. Because coordinates are centered at the class reference (∑irik=0), releasing evidence around the mean naturally produces negative deviations as the counterpart to positive ones, while the gain in target reserve comes from the bounded release as a whole rather than from the negative offsets alone. Crucially, the reserve cost on the negative side is controlled: reserve depends only on the worst rival coordinate, s=1+minkrik, which the barrier penalizes below −τ, so every in-cell sample retains s≥1−τ, well away from the decision boundary at r=−1. The barrier guards precisely the low tail that governs risk, namely the samples whose reserve approaches zero and whose decisions become fragile.
To see how this overall release improves target reserve, since sT≥sS−D∞ (Theorem 3.4), once the low tail is bounded, target reserve is determined by how far the shift moves the evidence. Intuitively, compressing each identity's evidence to a single point forces the classifier to over-rely on the feature directions that happen to be most stable in the source; once an unseen environment perturbs those directions, the evidence moves sharply. Allowing bounded variation spreads this reliance across more directions, so that no single shifted direction moves the evidence as much. Theoretically, we have added Proposition 3.5 in Section 3.5, with the complete proof in Appendix A.5, to formalize how bounded dispersion lowers target error. It shows that source dispersion is stationary at the minimum-dispersion state; therefore, whenever unseen shifts misalign with source variation, a bounded mean-preserving release increases source dispersion only at second order while reducing shift-induced variance at first order.
Comparing MarginCell with PointAlign, which compresses evidence toward r=0 without any release, confirms this (UCLA-CORES, Target Pr[s>0] is the target accuracy):
Both keep the source low tail equally small, so the difference lies in the release: MarginCell retains a higher target reserve at both thresholds, e.g., 63.01% vs. 55.57% of samples with s>0.5. Bounding the worst case preserves the source reserve; releasing the rest carries that reserve to the target domain.
Building on W1, we are glad to clarify how the bottleneck reserve is protected: because coordinates are referenced to the mean, some samples inevitably shift in the negative direction. This shift, however, is bounded: it consumes only surplus reserve, never brings the bottleneck reserve to a dangerous level, and in fact leaves fewer samples at risk. The reward operates only inside the cell and saturates at its boundary, so it pushes down samples near the class reference, whose reserve is close to 1, without pushing them out of the cell. For samples that have already fallen below −τ, the reward exerts no further downward pull, and the barrier instead pulls them back. Methods without such a floor have no mechanism to keep weak samples from dropping further. Consequently, the source reserve distribution exhibits a "lowered middle, raised floor" profile. On UCLA-CORES, the fraction of samples with reserve below 0.75 rises from 14.26% under ERM to 23.08%, whereas the fraction below 0.5 drops from 2.67% to 0.96%, and the 5% quantile of s rises from 0.604 to 0.704. More samples spread mildly, while the severely weakened samples, which actually govern the bottleneck reserve, become fewer.