Pith. sign in

REVIEW 3 major objections 5 minor 33 references

Forward Trajectory Steering for Hamilton-Jacobi Reachability Analysis

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Steering forward trajectories under the current value function's optimal control produces safety metrics competitive with, and sometimes better than, an MPC-guided solver, while cutting relative L2 error by up to 5.7x.

desk verdict A simple and mostly effective adaptive sampling scheme for PINN reachability, undermined by an overclaiming abstract and a potentially circular safety metric on the one adversarial benchmark. read the letter →

arxiv 2608.11480 v1 pith:UCUSMPLB submitted 2026-08-11 eess.SY cs.LGcs.SY

Sungje Park , Stephen Tu This is my paper · ORCID
classification eess.SYcs.LGcs.SY
keywords Hamilton-Jacobireachabilityphysics-informedneuralnetworksadaptivecollocationsamplingbackwardreachabletubesafetyvaluefunctionforwardtrajectorysteeringhigh-dimensionalcontrolstochasticdifferentialequations
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Hamilton-Jacobi reachability is the standard way to certify that a control system will avoid unsafe states, but solving its high-dimensional PDE is slow, and recent physics-informed neural network (PINN) solvers have needed complicated training pipelines to work. This paper claims that the key difficulty is not the network but where the PDE residual is sampled, and that a cheap adaptive sampling rule is enough. The rule is to take the current learned value function, compute the control and disturbance it says are optimal, roll the system forward under those signals with a little noise, and train on the states those rollouts visit. The resulting method, called STEER2REACH (S2R), matches or improves the safety metrics of a state-of-the-art model predictive control (MPC)-guided solver on four of five benchmarks and reduces relative L2 error by 1.3x, 2x, and 5.7x on the three problems with known ground truth, all without multi-stage training or MPC supervision. If this is right, the expensive machinery of curriculum learning and model-predictive-control-generated labels is not necessary for accurate high-dimensional reachability value functions.

What carries the argument

The central object is the forward-SDE sampling measure $\mu_\theta = \mathrm{Law}((X^{x;\theta}_t, t))$, where $X^{x;\theta}$ is the trajectory generated by Euler-Maruyama from the drift $f(x,u_\theta,d_\theta)$ plus additive noise $\sigma\,dW$, and $u_\theta,d_\theta$ are the control and disturbance induced by the current value function through the Hamiltonian argmax/argmin in equations (12)-(13). This measure replaces both the uniform collocation distribution of vanilla PINNs and the MPC-regression dataset of MPC-guided solvers. A second piece is the hard-constrained parameterization $V_\theta(x,t)=\ell(x)-(T-t)\rho(\phi_\theta(x,t))$, which enforces the terminal condition and the inequality $V_\theta\le\ell$ by construction and removes the boundary loss. The machinery works by continuously re-shaping the residual evaluation toward the part of state-time space the value function is actively controlling, and away from regions that do not matter.

What would settle it

Run S2R on a stiff or hybrid-dynamics control benchmark using an accurate implicit integrator for the rollouts instead of the Euler-Maruyama step with $N_{\mathrm{traj}}=50$; if the gap to MPC-DeepReach on safety metrics and RL2 closes, rollout fidelity is the mechanism, and if the gap persists, the limit lies elsewhere in the sampling strategy.

Watch

Extended reading notes

Core claim

The central claim is that a simple forward-trajectory sampling scheme can replace the complex sampling strategies used by current PINN-based reachability solvers. S2R defines a sampling distribution from the law of a stochastic differential equation whose drift uses the control and worst-case disturbance that the current value function $V_\theta$ would choose: $u_\theta$ maximizes and $d_\theta$ minimizes the inner product $\langle \nabla V_\theta, f(x,u,d)\rangle$. Euler-Maruyama integration of this SDE from uniformly sampled initial states, with Gaussian noise of strength $\sigma$, produces collocation points that concentrate where the value function is actively steering; the PINN residual is then evaluated at those points. The paper demonstrates on 2D, 3D, 7D, 13D, and 40D benchmarks that this procedure yields safety-level-set metrics (precision, IoU, true volume) competitive with MPC-DeepReach, and better relative L2 error where ground truth is available; on the 40D publisher-subscriber problem it is the best method by all three safety metrics. The one clear miss is the 7D F1Tenth hybrid-dynamics problem, which the paper attributes to rollout inaccuracy biasing the collocation measure. The paper also shows that a hard constraint $V_\theta(x,t) \le \ell(x)$, encoded through $V_\theta = \ell(x) - (T-t)\rho(\phi_\theta)$, improves both PINN and S2R performance but is not the main source of S2R's advantage.

Load-bearing premise

The load-bearing premise is that the simple forward-simulation rollouts, driven by the current value function's optimal control and worst-case disturbance, faithfully represent where the true system actually goes; if the rollouts are inaccurate, the PDE residual is evaluated in the wrong places and the training signal degrades, exactly as the paper reports for the F1Tenth benchmark.

Editorial extensions

If this is right

  • On the 2D vertical drone, 3D pursuit-evade, 13D quadrotor, and 40D publisher-subscriber benchmarks, S2R matches or exceeds the safety-level-set metrics (precision, IoU, true volume) of MPC-DeepReach; on the 40D problem it is the top performer.
  • Relative L2 error against ground truth is lower for S2R than for MPC-DeepReach by 1.3x (pursuit-evade), 2x (drone), and 5.7x (publisher-subscriber).
  • S2R's training loop is ordinary Adam on the PINN residual with the steered sampling distribution; it uses no MPC regression loss, no curriculum, and no supervised pre-training, and it can drop the boundary loss entirely when the hard-constrained parameterization is used.
  • The 7D F1Tenth hybrid problem is the exception: S2R trails MPC-DeepReach, and the paper shows that clipping, bounding, or filtering rollout states to keep them in the domain does not fix the bias introduced by rollout error.
  • S2R scales better than uniform PINN sampling in the publisher-subscriber family, where its relative L2 advantage grows as the state dimension increases from 2 to 500.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Not in the paper but directly suggested: any PDE solver whose accuracy hinges on collocation placement could adopt the same 'steer a surrogate process by the current solution, train where it goes' rule, so the idea is portable beyond HJ reachability.
  • The F1Tenth failure implies a testable boundary: the method's validity is tied to rollout fidelity, so stiff or hybrid systems that defeat the simple forward integrator are the natural first place to look for counterexamples.
  • Since the paper fixes the noise at $\sigma=0.01$ and $N_{\mathrm{traj}}=50$ across all benchmarks but its own ablation shows the best noise is problem-dependent, per-problem tuning is a plausible untested source of further gains.
  • The clipping/bounding/filtering results suggest a design principle for trajectory-based sampling measures: post-processing must preserve the measure induced by the true dynamics, a constraint that other adaptive sampling schemes would also need to respect.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. STEER2REACH (S2R) proposes a lightweight adaptive collocation-sampling scheme for physics-informed neural network (PINN) solvers of Hamilton-Jacobi reachability problems. In place of uniform, residual-based, or MPC-guided sampling, S2R generates collocation points by Euler-Maruyama rollouts of a forward SDE in which the control and disturbance are the argmax/argmin of the Hamiltonian computed from the current learned value function, plus injected noise. The training loop is formulated as repeated gradient descent on the PDE-residual loss under the induced sampling measure. The paper reports experiments on five benchmarks (2D vertical drone, 3D pursuit-evade, 7D F1Tenth, 13D quadrotor, and 40D publisher-subscriber), comparing S2R with vanilla PINNs, RAD-PINNs, and MPC-guided DeepReach. The central claim is that S2R matches or improves upon the state-of-the-art MPC-DeepReach on safety metrics while achieving lower relative L2 error on problems with ground truth, without multi-stage training or MPC-based supervision. The paper also contains a candid discussion of a substantial performance gap on F1Tenth and of the sensitivity of trajectory-rollout sampling to rollout accuracy.

Significance. If the empirical claims are valid, S2R is a practically meaningful simplification: it removes the need for MPC regression data, curriculum learning, and multi-stage fine-tuning while achieving relative L2 errors that are 1.3x to 5.7x lower than MPC-DeepReach on the three benchmarks with ground truth, and comparable or better safety metrics on four of the five problems. The paper deserves credit for releasing code, for running ablations on the hard constraint parameterization and on the two new hyperparameters sigma and dt, for a scalability study up to 500 dimensions, and for explicitly discussing the F1Tenth failure rather than omitting it. The main unresolved concern is that the safety evaluation is not independent of the learned value function: the 'true' safety labels are generated using the disturbance that is optimal for the learned value, not for the true value. The RL2 comparisons, which are independent of that issue, provide the strongest evidence for the usefulness of the method.

major comments (3)
  1. [Sec. V-A, Eq. (13)] The safety ground-truth label is computed as h*_theta(x) = 1{J(x,0,u_theta,d_theta)>0}, where d_theta is the argmin in Eq. (13) obtained from the learned value function V_theta. Since D is a box, d_theta is bang-bang in the sign of the gradient inner product, so a modest error in V_theta can switch the chosen disturbance from adversarial to non-adversarial and thereby inflate the reported precision, IoU, and TV. The paper calls d_theta an approximation but gives no error bound, no sensitivity analysis, and no comparison with the true worst-case disturbance d* on the three benchmarks where ground-truth V is available (vertical drone, pursuit-evade, publisher-subscriber). This issue is load-bearing because the central claim of 'competitive or improved safety metrics' rests on these labels. I request a re-evaluation of the safety metrics using a disturbance that is independent of the learned network (e.g., the true d* from ground-truth V, or an independent adversarial-search procedure), or, at minimum, a quantitative sensitivity study showing that the reported metrics are stable under perturbations of V_theta.
  2. [Sec. V-A, metric definitions] The metric called 'Precision' is defined as TP/(TP+FN), which is recall, not precision. Standard precision is TP/(TP+FP). As defined, this metric does not penalize false positives and can therefore make an over-optimistic safety classifier look better than it is. Since precision is one of the headline safety metrics in Table I, the formula should be corrected (or the metric renamed), and ideally the true precision should also be reported. This is a factual error in the evaluation protocol, and it directly affects the interpretation of the numerical comparisons.
  3. [Abstract, Sec. V-C, Sec. V-D] The abstract states that S2R 'achieves competitive--and in some cases improved--performance on safety metrics across a range of reachability benchmarks.' This is contradicted by the F1Tenth row of Table I, where S2R reports precision 0.7500 vs 0.9794 and IoU 0.7364 vs 0.9603 against MPC-DeepReach, and by the authors' own Sec. V-D discussion. The paper is transparent about this in the body, but the abstract and the conclusion should be amended to scope the claim, e.g., by explicitly excluding F1Tenth or by saying that performance is competitive on four of the five benchmarks. The F1Tenth discussion also states that 'any inaccuracies in the rollouts directly affect where the PDE residual is evaluated,' which is a fundamental limitation of the trajectory-steered sampling premise that should be reflected in the paper's claims.
minor comments (5)
  1. [Algorithm 1, line 21] The line 'Define L_S2R,j(theta) = L_VI(theta_j) + lambda L_bnd(theta_j)' is notationally inconsistent because the right-hand side does not depend on the argument theta. It should read L_VI(theta) + lambda L_bnd(theta) evaluated at theta = theta_j, or the theta argument should be removed.
  2. [Appendix B, publisher-subscriber] The publisher-subscriber problem is described as defining a 'target set' in the final paragraph of Appendix B, while the paper elsewhere refers to L as the 'failure set.' Since the paper's parameterization and value-function inequality V(x,t) <= l(x) are specific to the avoid formulation, please clarify whether publisher-subscriber is treated as a reach or avoid problem and reconcile the terminology.
  3. [Sec. V-B] The implementation paragraph says the methods are 'effectively under-sampled compared to MPC-DeepReach' because MPC-DeepReach also trains on regression data. This is a reasonable point, but the phrase 'We match the total number of collocation points' is confusing when immediately followed by a statement that the methods are under-sampled; please clarify which quantity is matched and which is not.
  4. [Sec. V-D] The three post-processing strategies (clipping, bounding, filtering) are described qualitatively, but no quantitative results are shown. A short table or figure quantifying the resulting bias and the failure to recover performance would strengthen the discussion, especially because this section is the main evidence for the method's limitation on hybrid dynamics.
  5. [Table I caption] The runtime unit is given as 'Runtime[m]' but never defined; please state explicitly that m denotes minutes, and consider reporting on comparable hardware and with comparable total training examples.

Circularity Check

1 steps flagged · score 4.0 of 10

Training loss is non-circular, but the safety ground truth for the disturbance benchmark is defined using the learned value function's own d_theta.

  1. fitted input called prediction [Sec. V-A, Evaluation Metrics (definition of h*_theta; Eq. (13))]
    "Note for the subset of problems where the disturbance set is non-empty, we utilize the disturbance d_theta(x,t) that is optimal for the learned value function to approximate the optimal disturbance d*(x,t; u_theta). ... we view V_theta as a binary classifier hhat_theta(x)=1{V_theta(x,0)>0} that approximates the ground truth h*_theta(x)=1{J(x,0,u_theta,d_theta)>0}."

    The 'ground-truth' safety label h*_theta is constructed from d_theta, which is itself the argmin in Eq. (13) computed from the learned V_theta. Therefore the TV, precision, and IoU metrics for the benchmark with a non-empty disturbance set measure whether the sign of V_theta agrees with trajectories rolled out under V_theta's own chosen adversary, not whether the controller u_theta is safe against the true worst-case disturbance d* of Eq. (8). If V_theta is wrong, d_theta can fail to be adversarial and inflate the apparent safety volume; no error bound or sensitivity analysis is provided. This makes the safety comparison partially self-referential, though it does not affect the RL2 comparisons or the no-disturbance benchmarks.

full rationale

The core S2R objective is the standard HJI-VI residual (9) evaluated on collocation points sampled from forward SDE rollouts; u_theta and d_theta enter only through the Hamiltonian and the sampling measure, so the value function is not defined in terms of the fitted parameters. The reported RL2 errors on 2D drone, 3D pursuit-evade, and 40D pub-sub are measured against external ground-truth value functions (17), and the no-disturbance benchmarks (drone, F1Tenth, quadrotor, pub-sub) have well-defined safety labels independent of theta. The cited [16] self-citation motivates trajectory-based sampling but is not used to prove the empirical claims. The one genuinely self-referential part is the safety ground truth h*_theta in Sec. V-A for the disturbance benchmark: it uses d_theta derived from the learned V_theta instead of the true worst-case d*, so the TV, precision, and IoU numbers for pursuit-evade are a consistency check rather than a certified worst-case safety evaluation. This localized evaluation circularity does not undermine the main training derivation or the external RL2 comparisons.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The ledger captures the two method-specific hyperparameters (noise and integration step), the chosen constraint function, and the core modeling premises: fidelity of the forward rollouts, approximation capacity of the network, and convergence of the repeated-gradient scheme. Standard HJ reachability theory is inherited from prior literature and not charged to the paper.

free parameters (4)
  • Noise injection sigma = 0.01, fixed across all benchmarks in Table I
    Introduced in Alg. 1; Fig. 4b shows the optimal sigma is problem-dependent, so the fixed value is a hand-chosen hyperparameter that affects RL2.
  • Integration step Delta t (N_traj=50) = Delta t = T/50 per problem
    Second new hyperparameter of S2R; Fig. 4c shows RL2 varies with Delta t and smaller step does not always help.
  • Constraint function rho = quadratic for most problems, softplus for F1Tenth
    In Eq. (16), Table II shows TV and RL2 depend on this choice; no principled selection rule is given.
  • Training schedule (lr, B_traj, B_pde) = lr 1e-4/5e-5/1e-5; B_traj=512; B_pde=4096
    Standard training hyperparameters matched to MPC-DeepReach where possible; they affect results but are not the paper's conceptual contribution.
assumptions (5)
  • standard math V(x,t) is the viscosity solution of the HJI-VI (6) and satisfies V(x,t) <= l(x).
    Used to set up the loss in Secs. II and IV-C; standard reachability theory from [1], [2].
  • domain assumption Neural networks can approximate the value function well enough that minimizing the PDE residual yields a useful controller.
    Assumed by all PINN baselines in Sec. III; no approximation guarantee is supplied.
  • ad hoc to paper Euler-Maruyama rollouts with value-function-induced controls and disturbances faithfully approximate the true dynamics-induced sampling measure mu_theta.
    This is the load-bearing premise of S2R; Sec. V-D shows it fails for F1Tenth, causing the main performance gap.
  • domain assumption The optimal disturbance of the learned value function d_theta approximates the true worst-case disturbance d* in evaluation.
    Stated in Sec. V-A; needed to compute safety metrics when disturbances are present.
  • ad hoc to paper Repeated gradient descent on the performative loss (15) converges to a useful value function.
    No convergence proof is given; direct minimization with smoothed gradients was unstable (Sec. IV-B), so the paper relies on RGD empirically.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Forward Trajectory Steering for Hamilton-Jacobi Reachability Analysis." pith.science (2026). https://pith.science/paper/UCUSMPLB

@misc{pith2026260811480,
  author       = {Pith},
  title        = {Pith review of: Forward Trajectory Steering for Hamilton-Jacobi Reachability Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UCUSMPLB}},
  note         = {Machine review of arXiv:2608.11480}
}
read the original abstract

Hamilton-Jacobi (HJ) reachability provides a mathematically rigorous framework for safe control of dynamical systems, but its practical application is bottlenecked by the computational complexity of solving Hamilton-Jacobi-Isaacs variational inequality PDEs in high dimensions. Physics-informed neural networks (PINNs) have recently emerged as a promising alternative to classical mesh-based solvers, yet their performance is highly sensitive to the choice of collocation sampling. In order to learn accurate safety value functions, existing PINNs-based HJ reachability solvers must rely on complex training pipelines and auxiliary supervision. In this work, we propose STEER2REACH (S2R), a PINNs-based HJ reachability solver that requires minimal modification on top of standard PINNs training. S2R's key contribution is a lightweight, low-overhead adaptive collocation sampling distribution constructed by steering forward trajectories using a combination of the optimal control and disturbance signals induced by the current value function, with injected stochastic exploration noise. We demonstrate that despite its simplicity, S2R achieves competitive--and in some cases improved--performance on safety metrics while reducing relative L2 error across a range of reachability benchmarks compared with SoTA MPC-guided HJ reachability solvers, all without requiring multi-stage training or MPC-based supervision.

Figures

Figures reproduced from arXiv: 2608.11480 by the authors.

Figure 1
Figure 1. A plot of sample trajectories from the 2D vertical drone problem [7]. The green dotted lines represent the floor and ceiling that the system should avoid, the black line represents the learned safety value function zero-level contour, and the blue lines illustrate sampled trajectories. For each sample trajectory, the green and red dots represent the start and end points, respectively. of physics-informed neural netw… view at source ↗
Figure 2
Figure 2. Comparison plots of the learned safety value function Vθ(x, t) at t = 0 with the zero level set outlined as a black line. The green dotted line in (a) represent the floor and ceiling that the system should avoid and the white dotted line represents the ground truth zero level contour. We include the RL2 error of each method in the title for reference. where ρ : R 7→ R⩾0 is a fixed (non-learnable) function with non-n… view at source ↗
Figure 3
Figure 3. A plot of the predicted volume (PV) vs. true volume (TV) for each problem and model pair. The dashed black line represents the ideal TV = PV scaling. where RAD-PINNs achieves a better TV but a significant worse IOU and RL2 compared to MPC-DeepReach and S2R. We also see that both MPC-DeepReach and S2R have significantly more accurate level-sets than PINNs and RAD￾PINNs (e.g., higher IOU) even without explicit post-ho… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Ablation study plots for STEER2REACH. In (b) and (c), the outlined points represent the values used for S2R in Table I. constraint (i.e., ρ(x) = x). Changing ρ in (16) does not affect the boundary constraint Vθ(x, T) = ℓ(x), and hence this study focuses on the benefits…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 26 canonical work pages

  1. [1]

    A time-dependent hamilton- jacobi formulation of reachable sets for continuous dynamic games,

    I. Mitchell, A. Bayen, and C. Tomlin, “A time-dependent hamilton- jacobi formulation of reachable sets for continuous dynamic games,” IEEE Trans. Autom. Control, vol. 50, no. 7, pp. 947–957, 2005

    work page 2005

  2. [2]

    Hamilton-jacobi reachability: A brief overview and recent advances,

    S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin, “Hamilton-jacobi reachability: A brief overview and recent advances,” inProc. IEEE Conf. Decision and Control (CDC), 2017

    work page 2017

  3. [3]

    Dgm: A deep learning algorithm for solving partial differential equations,

    J. Sirignano and K. Spiliopoulos, “Dgm: A deep learning algorithm for solving partial differential equations,”J. Comput. Phys., vol. 375, pp. 1339–1364, 2018

    work page 2018

  4. [4]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” Journal of Computational Physics, vol. 378, pp. 686–707, 2019

    2019

  5. [5]

    Linear supervision for nonlinear, high-dimensional neural control and differential games,

    W. Sharpless, Z. Feng, S. Bansal, and S. Herbert, “Linear supervision for nonlinear, high-dimensional neural control and differential games,” inProc. Learn. Dyn. Control Conf. (L4DC), 2025

    work page 2025

  6. [6]

    Madr: Mpc-guided adversarial deepreach,

    R. Teoh, S. Tonkens, W. Sharpless, A. Yang, Z. Feng, S. Bansal, and S. Herbert, “Madr: Mpc-guided adversarial deepreach,”arXiv preprint arXiv:2510.18845, 2025

    arXiv 2025

  7. [7]

    Bridging model predictive control and deep learning for scalable reachability analysis,

    Z. Feng, L. Qiu, and S. Bansal, “Bridging model predictive control and deep learning for scalable reachability analysis,”arXiv preprint arXiv:2505.03830, 2025

    arXiv 2025

  8. [8]

    Predictive limitations of physics- informed neural networks in vortex shedding,

    P.-Y . Chuang and L. A. Barba, “Predictive limitations of physics- informed neural networks in vortex shedding,”arXiv preprint arXiv:2306.00230, 2023

    arXiv 2023

Show all 33 references
  1. [9]

    DeepReach: A deep learning approach to high-dimensional reachability,

    S. Bansal and C. Tomlin, “DeepReach: A deep learning approach to high-dimensional reachability,” inProc. IEEE Int. Conf. Robot. Autom. (ICRA), 2021

    2021

  2. [10]

    Exact imposition of safety boundary conditions in neural reachable tubes,

    A. Singh, Z. Feng, and S. Bansal, “Exact imposition of safety boundary conditions in neural reachable tubes,” inProc. IEEE Int. Conf. Robot. Autom. (ICRA), 2025

    2025

  3. [11]

    Implicit neural representations with periodic activation functions,

    V . Sitzmann, J. Martel, A. Bergman, D. Lindell, and G. Wetzstein, “Implicit neural representations with periodic activation functions,” in Adv. Neural Inf. Process. Syst. (NeurIPS), 2020

    2020

  4. [12]

    Characterizing possible failure modes in physics-informed neural networks,

    A. Krishnapriyan, A. Gholami, S. Zhe, R. Kirby, and M. W. Mahoney, “Characterizing possible failure modes in physics-informed neural networks,” inAdv. Neural Inf. Process. Syst. (NeurIPS), 2021

    2021

  5. [13]

    Solving high-dimensional partial differential equations using deep learning,

    J. Han, A. Jentzen, and W. E, “Solving high-dimensional partial differential equations using deep learning,”Proc. Natl. Acad. Sci. U.S.A, vol. 115, no. 34, pp. 8505–8510, 2018

    2018

  6. [14]

    Forward–backward stochastic neural networks: deep learn- ing of high-dimensional partial differential equations,

    M. Raissi, “Forward–backward stochastic neural networks: deep learn- ing of high-dimensional partial differential equations,” inPeter Carr Gedenkschrift, Res. Adv. Math. Finance, 2024, pp. 637–655

    2024

  7. [15]

    Interpolating between bsdes and pinns: Deep learning for elliptic and parabolic boundary value problems,

    N. N ¨usken and L. Richter, “Interpolating between bsdes and pinns: Deep learning for elliptic and parabolic boundary value problems,”J. Mach. Learn., vol. 2, no. 1, pp. 31–64, 2023

    2023

  8. [16]

    Integration matters for learning PDEs with backward SDEs,

    S. Park and S. Tu, “Integration matters for learning PDEs with backward SDEs,” inAdv. Neural Inf. Process. Syst. (NeurIPS), 2025

    2025

  9. [17]

    Isl 2 physics informed loss always suitable for training physics informed neural network?

    C. Wang, S. Li, D. He, and L. Wang, “Isl 2 physics informed loss always suitable for training physics informed neural network?” inAdv. Neural Inf. Process. Syst. (NeurIPS), 2022

    2022

  10. [18]

    Efficient training of physics-informed neural networks via importance sampling,

    M. A. Nabian, R. J. Gladstone, and H. Meidani, “Efficient training of physics-informed neural networks via importance sampling,”Comput.- Aided Civ. Infrastruct. Eng., vol. 36, no. 8, pp. 962–977, 2021

    2021

  11. [19]

    A comprehensive study of non-adaptive and residual-based adaptive sampling for physics- informed neural networks,

    C. Wu, M. Zhu, Q. Tan, Y . Kartha, and L. Lu, “A comprehensive study of non-adaptive and residual-based adaptive sampling for physics- informed neural networks,”Comput. Methods Appl. Mech. Eng., vol. 403, p. 115671, 2023

    2023

  12. [20]

    Reflected solutions of backward sde’s, and related obstacle problems for pde’s,

    N. El Karoui, C. Kapoudjian, E. Pardoux, S. Peng, and M.-C. Quenez, “Reflected solutions of backward sde’s, and related obstacle problems for pde’s,”Ann. Probab., vol. 25, no. 2, pp. 702–737, 1997

    1997

  13. [21]

    Stochastic differential games with reflection and related obstacle problems for isaacs equations,

    R. Buckdahn and J. Li, “Stochastic differential games with reflection and related obstacle problems for isaacs equations,”Acta Math. Appl. Sin., Engl. Ser., vol. 27, no. 4, pp. 647–678, 2011

    2011

  14. [22]

    Numerical simulation of bsdes using empirical regression methods: theory and practice,

    E. Gobet and J.-P. Lemor, “Numerical simulation of bsdes using empirical regression methods: theory and practice,”arXiv preprint arXiv:0806.4447, 2008

    2008 arXiv

  15. [23]

    A primal–dual algorithm for bsdes,

    C. Bender, N. Schweizer, and J. Zhuo, “A primal–dual algorithm for bsdes,”Mathematical Finance, vol. 27, no. 3, pp. 866–901, 2017

    2017

  16. [24]

    Deep signature algorithm for multidimensional path-dependent options,

    E. Bayraktar, Q. Feng, and Z. Zhang, “Deep signature algorithm for multidimensional path-dependent options,”SIAM J. Financial Math., vol. 15, no. 1, pp. 194–214, 2024

    2024

  17. [25]

    Performative prediction,

    J. Perdomo, T. Zrnic, C. Mendler-D ¨unner, and M. Hardt, “Performative prediction,” inProc. Int. Conf. Mach. Learn. (ICML), 2020

    2020

  18. [26]

    A forward reachability perspective on control barrier functions and discount factors in reachability analysis,

    J. J. Choi, D. Lee, B. Liet al., “A forward reachability perspective on control barrier functions and discount factors in reachability analysis,” arXiv preprint arXiv:2310.17180, 2023

    2023

  19. [27]

    DeepReach Repository,

    A. Lin, Z. Feng, J. Borquez, and S. Bansal, “DeepReach Repository,”

  20. [28]

    JAX: composable transformations of Python+NumPy programs,

    J. Bradbury, R. Frostig, P. Hawkinset al., “JAX: composable transformations of Python+NumPy programs,” 2018. [Online]. Available: http://github.com/jax-ml/jax

    2018

  21. [29]

    Generalizing safety be- yond collision-avoidance via latent-space reachability analysis,

    K. Nakamura, L. Peters, and A. Bajcsy, “Generalizing safety be- yond collision-avoidance via latent-space reachability analysis,”arXiv preprint arXiv:2502.00935, 2025

    2025

  22. [30]

    Claude Code,

    Anthropic, “Claude Code,” https://www.anthropic.com/, 2026

    2026

  23. [31]

    OpenAI, “Codex,” https://openai.com/codex/, 2026

    2026

  24. [32]

    ChatGPT,

    OpenAI, “ChatGPT,” https://chatgpt.com/, 2026. APPENDIX A. RAD Algorithm Algorithm 2RAD Sampling [19] Input:V θ(x,t),B pde,N cand,k,c Output:{(x b,tb)} Bpde b=1 (xi,ti)∼U(Ω×[t 0,tf]), i= 1,...,N cand ri =|R Vθ(xi,ti)| qi = (ri)k ¯q= 1 Ncand PNcand m=1qm wi = qi ¯q+c pi = wiPNc...

    2026

  25. [2024]

    Available: https://github.com/smlbansal/deepreach

    [Online]. Available: https://github.com/smlbansal/deepreach

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.