- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
The authors propose two neural architectures, LegendreNet and ContactNet, for learning contact maps that are invertible on regular domains.
LegendreNet composes strict contact blocks, while ContactNet combines them with prolonged base maps to model general contact dynamics.
The paper establishes strict universality and gives sufficient conditions for approximating smooth contact Hamiltonian flow maps on compact sets.
The inverse of the forward-trained map is then used to recover initial profiles from complete ordered one-dimensional folded fronts.
Critical strengths
-
The modular approximation theorem provides a useful criterion for constructing expressive contact-preserving networks.
-
The folded-front example provides a concrete illustration of reusing a forward-trained contact map for initial-profile recovery.
Critical weaknesses
-
For the triangular-base ContactNet used in the inverse experiments, universality for general contact Hamiltonian flow maps relies on the -density of its base-map compositions, which is not established in the paper. Could the authors identify any non-strict target classes for which the current results provide approximation guarantees without this assumption?
-
Although the paper provides derivative formulas and low-dimensional timing results, it does not assess how derivative-evaluation overhead and the capacity needed to maintain a fixed approximation accuracy change with dimension
Questions for authors
-
How do computation, memory, and required depth scale with state dimension for the triangular and neural-flow variants?
-
What geometric assumptions or additional measurements would enable momentum recovery in two or more dimensions?
-
How does the proposed framework scale to long-horizon prediction in real-world applications, such as damped mechanical systems, in terms of stability and computational cost?
2 comments
Thank you for your thoughtful and thorough rebuttal. I sincerely appreciate the significant effort you have put into addressing my concerns, especially the additional theoretical results, experiments, and detailed analyses.
After carefully reading your responses, I realize that I may have misunderstood certain aspects of your proposed methodology. Your explanations and additional results have helped clarify these points and given me a better understanding of the contributions. I have decided to increase my score by one point.
Response to GBcM
We thank the reviewer for the careful reading and for the precise question about Corollary 4.3(b). It led to a stronger result: the triangular ContactNet is universal with no base-density hypothesis (see the general response). Below we answer that question directly, then the questions on dimension and horizon.
W1 (universality of the triangular ContactNet). The general response gives the stronger answer. LegendreNet atoms composed with positive x-conditioned action shears are already Cr-dense, so Corollary 4.3(b) now holds with no base-density hypothesis.
The reviewer's question also has a direct answer within the submitted paper. For H=H0(x,p)+γ(x)u+β(x), which includes our damped oscillator, the flow of γ(x)u+β(x) is an exact action shear defined on all of J1(Rn). A refined Lie–Trotter product of a strict flow (Theorem 4.1) and this shear, with Lemma C.1, gives the guarantee without the new theorem. The new theorem extends this to every smooth contact Hamiltonian. It shows that brackets of strict Hamiltonians with x1u generate all of them. The full proof, including the localization to a compact tube and the negative-time substeps, is in the new Appendix C.1.
W2 and Q1 (cost, memory and depth as n grows). The triangular blocks were designed to avoid n×n solves. Each needs only scalar fields and their gradients on Rn, and its inverse is in closed form at the same cost. The u-conditioned spatial block adds a rank-one correction instead of a linear solve. Only the neural-flow base of ContactNet-ODE needs an (n+1)×(n+1) Jacobian, with an O(n3) solve. We timed each block at width 96 for n=1 to 64 (Appendix J.3, Table 25). Time and memory grow near-linearly in n for the triangular blocks, and a full stage goes from 2.0 to 46 µs per state. Inverses cost the same as forward passes, and the neural-flow base costs 15 to 40 times a triangular stage.
A bound on depth at fixed accuracy remains open. The theorem is qualitative, like the universality results for coupling flows. We did measure depth on the one-step map of H=∣p∣2/2+u2/4 for n=1,2,4 and 8. At three stages, the error barely moves across widths 8 to 128, by less than a factor of two for n≥2. It does grow with n, from between 2.3×10−7 and 1.05×10−6 at n=1, depending on width, to about 1.4×10−5 at n=8. Doubling the number of stages cuts the error 6.6 times at n=4 and 2.5 times at n=8. Four times the training cuts it only 1.3 to 1.4 times. So at these sizes, the number of composed stages, not the width of each field, has to grow with n. This rests on one doubling at one width with two seeds, so we report it as an observation, not a rate (Table 26).
Q2 (momentum recovery in two or more dimensions). The relation DsU=PDsX fixes P wherever DsX is invertible, with no extra assumption. At a fold point, DsX has corank one with kernel v. Differentiating along v gives P⋅D2X(v,v)=D2U(v,v), which supplies the missing component when D2X(v,v) is not in the range of DsX. Where that fails, as at the isolated cusps of a generic 2D front, P follows by continuity, provided the singular set has empty interior. This gives an n-dimensional version of Proposition 6.1 (new Proposition H.2, Appendix H.3). In practice, the 2D inverse can now run from the projected front alone. It stays finite on all 11 test profiles, with median initial-front error 4.5×10−3 against 2.3×10−5 when momentum is supplied, limited by the momentum lift near caustics.
Q3 (long horizons on damped mechanical systems). Cost grows linearly with the horizon. Each step is one map evaluation with constant memory at inference. Training uses one-step pairs, so its cost does not depend on the horizon. Every step is an exact contactomorphism on its regular domain, so what accumulates over a rollout is fit error, not loss of structure. The conformal factor is a product over steps, so long-horizon dissipation accuracy is set by the per-step factor error, which we now report (Appendix J.1).
We ran the saved oscillator models to ten periods (800 steps, Appendix J.4), and every model stays finite on the test sets. Damped state MSE at ten periods is 2.3×10−5 for GCF, 2.8×10−4 for the MLP, 7.4×10−4 for ContactNet, 1.4×10−3 for RealNVP and 0.97 for LegendreNet. In the conservative case, MSE is 0.083 for GCF, 0.36 for ContactNet, 0.80 for LegendreNet, 3.5 for the MLP and 182 for RealNVP. On the wide-range test set, GCF diverges in every seed, between steps 58 and 92. A ContactNet step costs 0.11 ms at batch 32, against 0.8 to 1.1 ms for GCF.