Pith. sign in

REVIEW 3 major objections 5 minor 54 references

Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that imitation policies shaped as contractive dynamical systems recover out-of-sample states with a provable upper bound on loss, because every rollout contracts exponentially toward the expert trajectory.

desk verdict Useful architecture and strong OOS results, but the central contractivity guarantee and the OOS loss bound are not proven; the paper should not be accepted as-is. read the letter →

arxiv 2412.07544 v2 pith:FJONDMM6 submitted 2024-12-10 cs.LG cs.ROstat.ML

Amin Abyaneh , Mahrokh G. Boroujeni , Hsiu-Chin Lin , Giancarlo Ferrari-Trecate This is my paper · ORCID
classification cs.LGcs.ROstat.ML
keywords imitationlearningcontractivedynamicalsystemsout-of-samplerecoveryrecurrentequilibriumnetworkscouplinglayersneuralordinarydifferentialequationssoftdynamictimewarpingroboticmanipulation
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

The paper targets a specific failure mode of imitation learning: a policy trained on expert demonstrations can become unreliable when deployment starts from a state outside the training data. Its proposal, SCDS, is an imitation policy built as a contractive dynamical system, so that any rollout exponentially converges toward a single trajectory regardless of initial condition or perturbation. Contractivity is built into the architecture rather than enforced during training: a recurrent equilibrium network in a latent space, followed by a linear projection and invertible coupling layers, remains contractive for every parameter choice, which allows unconstrained gradient-based optimization. The paper derives an upper bound on the worst-case and expected out-of-sample mean-squared-error loss under an initial-state locality assumption, turning contractivity into a quantitative recovery certificate. A sympathetic reading is that this gives imitation learning a transient-phase guarantee that asymptotic stability alone does not provide.

What carries the argument

The load-bearing construction is the SCDS policy, a continuous-time dynamical system $\dot{z} = f_\theta(z)$ in a latent space, paired with an output map made of a linear projection and $K$ invertible coupling layers. The latent dynamics are a recurrent equilibrium network (REN), a nonlinear model whose parameterization guarantees contractivity at a chosen rate $\gamma$ for all parameters $\theta$; the coupling layers are bijective maps that preserve that contractivity, and the initial latent state is set with a pseudoinverse so the planned state approximates the observed initial state. This architecture lets the paper reduce the out-of-sample recovery question to a comparison of rollouts from nearby initial conditions, which contractivity then turns into an exponential-decay estimate used in Theorem 4.1.

What would settle it

Take $a=0$, $b=1$, $c=2$: the squared-error loss from $a$ to $c$ is $4$, while the sum of the losses via $b$ is $1+1=2$, so the triangle inequality used in the proof of Theorem 4.1 (Appendix A.2, Eq. (16)) fails for MSE. A complete falsifier would be an explicit contractive policy and an out-of-sample state whose measured MSE loss exceeds the claimed upper bound.

Watch

Extended reading notes

Core claim

The central claim is that the policy in Eq. (7) is contractive for every parameter choice, with the contraction rate preserved through the output transformation, and that this contractivity directly yields a deployable recovery guarantee. Theorem 4.1 states that, when the out-of-sample initial state lies in a multi-focal ellipse around the training initial states, its MSE loss is bounded by a weighted sum of in-sample rollout errors plus a term that shrinks with the contraction rate and with the size of the uncertainty region; Corollary 4.1.1 extends the same bound to any distribution over that region. The paper presents empirical evidence that these guarantees translate into lower out-of-sample error than stability-only baselines on LASA handwriting motions and Robomimic manipulation tasks, including simulated deployment on a Franka arm and a Jackal robot.

Load-bearing premise

The quantitative upper-bound theorem assumes that the squared-error loss satisfies the triangle inequality, but mean squared error does not: the cross term is dropped in the proof in Appendix A.2. If that assumption fails, the stated worst-case and expected-loss bounds do not follow from contractivity alone.

Editorial extensions

If this is right

  • Any rollout started from an out-of-sample state contracts exponentially toward rollouts from in-sample states, so the transient phase is controlled and the target is reached from unseen initial conditions.
  • Because contractivity holds for every parameter value, training can be done with unconstrained gradient descent, avoiding the accuracy-versus-stability trade-offs of constrained optimization.
  • The bound in Theorem 4.1 is computable from precomputed in-sample errors and inverse-distance weights, so it can serve as a deployment-time certificate for a learned policy.
  • State-only training removes the need for expert velocity measurements, eliminating a common source of accumulated error in imitation learning.
  • Higher contraction rates or tighter initial-state uncertainty regions directly tighten the certified worst-case and expected loss.

Reading between the lines

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

  • The bound's proof uses the triangle inequality for the squared-error loss, which mean squared error does not satisfy; a direct corollary is that the certificate would become rigorous if the loss were replaced by a true metric, such as Euclidean distance, or if the proof were amended to account for the missing cross term.
  • Because the bound is computable from the current state, one could extend SCDS to an online safety monitor that pre-empts rollouts whose certified loss exceeds a threshold, an application the paper does not discuss.
  • The inverse-distance weighting in Eq. (9) is a nearest-neighbour-style soft partition of the initial-state space; it would be natural to test whether alternative kernels or learned weights tighten the bound while preserving the zero-loss-at-demonstration property.
  • For multi-modal expert behaviors, such as the Robomimic tasks where demonstrations do not share a target, a single contractive system can only contract to an average behavior; an extension would need mixture or mode-conditional contractive policies.
Share X Bluesky LinkedIn Reddit HN

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. The paper proposes SCDS, a state-only imitation learning method whose policy is a continuous-time dynamical system composed of a contractive recurrent equilibrium network (REN) in a latent space, a linear projection, and a stack of RealNVP coupling layers. The authors claim that this composition is contractive for every parameter choice, enabling unconstrained optimization, and they derive upper bounds on the worst-case and expected out-of-sample loss (Theorem 4.1 and Corollary 4.1.1). The method is evaluated on LASA and Robomimic datasets, with additional deployment in the Isaac Lab simulator, reporting strong out-of-sample recovery compared with stable and non-stable baselines.

Significance. If the theoretical results were correct, this would be a valuable contribution: a contractive-by-construction policy class with unconstrained training and explicit worst-case out-of-sample guarantees would address a real gap in imitation learning. The empirical results are promising, the experiments are fairly extensive, and the authors ship code. However, the central theoretical claims rest on two independent mathematical errors—in the preservation of Euclidean contractivity through coupling layers and in the use of a triangle inequality for the MSE loss—so the main advertised guarantee is not established. The empirical findings may still be of interest, but the paper's headline contribution is the rigorous OOS bound, which is currently unsupported.

major comments (3)
  1. [App. A.2, Eq. (16)] Theorem 4.1's proof uses the 'triangle inequality for MSE' to claim ℓ(ŷ, y^m) ≤ ℓ(ŷ, ŷ^m) + ℓ(ŷ^m, y^m). Mean squared error does not satisfy the triangle inequality: for scalar values 0, 1, 2, MSE(0,2)=4 while MSE(0,1)+MSE(1,2)=2. The missing cross term is not bounded in the proof, so the displayed upper bound in Theorem 4.1 and the bound in Corollary 4.1.1 do not follow from the preceding inequalities. This is a load-bearing step for the paper's main 'rigorous' deployment guarantee.
  2. [App. A.1, Proposition 2.1 and Eq. (13)] The preservation of Euclidean contractivity through the output map is not established. First, Lemma A.1 gives ||Pv||²/||v||² ≥ σ_min(P)², hence ||v|| ≤ ||Pv||/σ_min(P), so the denominator in Eq. (13) should be σ_min(P), not σ_min(P)². Second, the paper explicitly advocates latent dimensions Nz > Ny for expressiveness, in which case P ∈ R^{Ny×Nz} necessarily has a nontrivial null space, σ_min(P)=0, and the displayed bound is vacuous. Third, the citation to Manchester & Slotine for 'bijective maps preserve contractivity' concerns contraction in a pulled-back metric, not a single global Euclidean α in the norm of Definition 2.1. A RealNVP coupling layer such as (x1,x2) ↦ (x1, x2 e^{x1}) is bijective but not bi-Lipschitz; combined with a contracting linear latent flow, the ratio of output-trajectory distance to e^{-γt} times the initial output distance can be made arbitrarily large, so no finite α independent of the initial condition satisfies Eq. (2). Consequently the premise of Theorem 4.1—that the composed policy ϕθ is contractive in the output space—is not guaranteed for the proposed architecture.
  3. [Sec. 5, 'Upper bound' paragraph and Corollary 4.1.1] The constant α in the reported upper bound is not a verified certificate: the paper states it is 'approximated empirically using a Monte Carlo approach' after training. Since Proposition 2.1 does not provide a computable α from the network parameters, the numerical values labeled L^{MSE}_{ub} are post-hoc estimates rather than upper bounds certified by the theory. The claim that the bound 'rigorously establish[es] the reliability of our method in deployment' is therefore stronger than what is delivered.
minor comments (5)
  1. [Definition 2.1] Definition 2.1 allows an arbitrary Lp norm, but the paper then fixes the L2 norm; this should be stated in the definition itself to avoid ambiguity in Eq. (2) and in the proofs.
  2. [Sec. 2.3, Eq. (6)] When Nz > Ny, P cannot have full column rank, so z(0) is only a least-squares approximation and ŷ(0) ≠ y0 in general; the paper acknowledges this, but the subsequent contractivity statements for output trajectories starting from exact y0 require an additional argument that is not provided.
  3. [Assumption 4.1 and App. A.2] The set in Eq. (12) is called a 'multi-focal ellipse', but for M>2 it is a convex multi-center region rather than an ellipse in the usual sense; also App. A.2 refers to it as a 'conic initial set', which appears to be a typo.
  4. [Theorem 4.1] The geometric-series factor (e^{-2γ}-1)/(e^{-2γ/H}-1) is positive because both numerator and denominator are negative; the authors should state this explicitly to avoid confusion about the sign of the bound.
  5. [Sec. 5, Table 1] The L^{MSE}_{ub} values would be more informative if the paper reported the number of Monte Carlo samples, the variance of the α estimate, and a comparison of the bound with the observed OOS loss under the same distribution; the claim that the bound is 'relatively tight' is not quantified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the OOS bound is conditional on an imported contractivity guarantee, and the bound's terms are not equivalent to its own fitted inputs by construction.

full rationale

Walking the derivation chain: (i) contractivity of the latent REN is imported from Martinelli et al. (2023, Thm 3; App. D.2-D.3), a prior theorem with an explicit proof and a stated algebraic construction. Although a coauthor of the present paper is on that cited work, the citation is real evidence rather than a restatement of the present paper's target result, so it does not constitute circularity. (ii) Proposition 2.1 (App. A.1) is meant to transport contractivity through the output map; the argument uses Lemma A.1 and cites Manchester and Slotine for bijective maps. Whatever the validity of that step as a mathematical claim (a potential correctness risk), it is not a reduction of the conclusion to the paper's own fitted quantities. (iii) Theorem 4.1 (App. A.2) starts from the definition of L and the contractivity inequality, and its bound has two terms: the weighted training-loss term (i) and an uncertainty term (ii) built from R and alpha. The bound is not made true by construction: alpha is approximated after training rather than solved for to force the inequality, and the training-loss term is an input to the bound, not a renamed version of the OOS loss being predicted. The proof does contain an invalid use of a triangle inequality for MSE and a possible misapplication of Assumption 4.1 in Eq. 19; these are correctness defects, not circularity. No equation in the paper sets the predicted OOS loss equal to a fitted parameter or to a self-cited theorem by definition.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The architecture introduces no new physical entities. The main theoretical claim rests on imported contractivity theorems, a domain assumption on initial states, and an invalid subadditivity premise for MSE.

free parameters (3)
  • gamma (contraction rate) = [1.0, 18.6], learnable
    Contraction rate controls the exponential convergence speed and appears in the theoretical bound; it is either set by hand or learned with an augmented Lagrangian term (App. C.1).
  • R (multi-focal ellipse radius) = not fixed; OOS samples drawn from a hyper-sphere of radius 0.1||y0|| around training starts
    Radius of the multi-focal ellipse in Assumption 4.1; it scales the second term of the upper bound and is chosen by the evaluation protocol.
  • alpha (contraction constant) = estimated via Monte Carlo after training (Sec. 5.2)
    Constant in the contractivity inequality (Def. 2.1); the paper approximates it empirically to compute the reported upper bound, rather than deriving it.
assumptions (4)
  • standard math The REN parameterization of Martinelli et al. (2023) guarantees that the latent dynamics are contractive with rate gamma for any parameters.
    Used without proof in Eq. 3; the paper relies on external Theorems D.1 and D.2 (App. D).
  • standard math Bijective maps preserve contractivity with the same rate gamma (Manchester & Slotine 2015).
    Invoked in Proposition 2.1; external result, not proved in the paper.
  • domain assumption Assumption 4.1: the OOS initial state lies in a multi-focal ellipse with sum of distances to training initial states bounded by R.
    Defines the region in which the loss bound is claimed to hold.
  • ad hoc to paper The MSE loss is subadditive (triangle inequality).
    Used in App. A.2 Eq. 16 to split the loss; false for squared L2 distances, so the proof of Theorem 4.1 breaks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery." pith.science (2026). https://pith.science/paper/FJONDMM6

@misc{pith2026241207544,
  author       = {Pith},
  title        = {Pith review of: Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FJONDMM6}},
  note         = {Machine review of arXiv:2412.07544}
}
read the original abstract

Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamical systems guarantees convergence to a desired state, it often overlooks transient behavior. We propose a framework for learning policies modeled by contractive dynamical systems, ensuring that all policy rollouts converge regardless of perturbations, and in turn, enable efficient OOS recovery. By leveraging recurrent equilibrium networks and coupling layers, the policy structure guarantees contractivity for any parameter choice, which facilitates unconstrained optimization. We also provide theoretical upper bounds for worst-case and expected loss to rigorously establish the reliability of our method in deployment. Empirically, we demonstrate substantial OOS performance improvements for simulated robotic manipulation and navigation tasks.

Figures

Figures reproduced from arXiv: 2412.07544 by the authors.

Figure 1
Figure 1. Policy rollouts generated by contractive and stable policies. While both policies eventually reach the tar￾get, the contractive policy closely mim￾ics the expert in the transient phase. One such certificate is asymptotic stability (Devaney, 2021), which ensures that all trajectories eventually converge to the same equilibrium state, regardless of the initial condition or perturbations (Khansari-Zadeh & Billard, 2011… view at source ↗
Figure 2
Figure 2. Overview of the SCDS training scheme. The policy structure (top box) consists of a REN, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Uniform weighting averages the deviation from both demonstrations, hence, L(y a 0 ; θ) ̸= 0. In contrast, Eq. 9 results in L(y a 0 ; θ) = 0. For y c 0 , the weights in Eq. 9 assign higher importance to the closer y b 0 ’s demonstration, which is intuitive. where λm(yˆ0) ∈ [0, 1] is the weight assigned to the difference from the m-th demonstration, and these weights satisfy the condition PM m=1 λm(yˆ0) = 1. A naive c… view at source ↗
Figures from the paper (18 more)
Figure 4
Figure 4. Figure 4: In-sample and OOS policy rollouts for selected 2D tasks in the LASA dataset. The training [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Comparing SCDS to the selected baselines on in-sample and OOS rollouts in the 2D task [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: In-sample and OOS policy rollouts for the 6D [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Simulation rollouts with Franka arm and Jackal mobile robots in Isaac Lab. Both robots [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Contractive behavior of the latent state. The figure illustrates the rollouts over time for [PITH_FULL_IMAGE:figures/full_fig_p022_8.png]
Figure 9
Figure 9. Figure 9: SCDS is tested over additional motions in the LASA dataset. In all cases, the induced [PITH_FULL_IMAGE:figures/full_fig_p023_9.png]
Figure 10
Figure 10. Figure 10: Out-of-sample recovery for initial states sampled from regions with different volumes [PITH_FULL_IMAGE:figures/full_fig_p023_10.png]
Figure 11
Figure 11. Figure 11: Policies trained to have low (first row) and high (second row) contraction rates. SCDS [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]
Figure 12
Figure 12. Figure 12: Training results for a sample motion in the LASA dataset are plotted for various latent [PITH_FULL_IMAGE:figures/full_fig_p024_12.png]
Figure 13
Figure 13. Figure 13: The effect of replacing the contractive REN with a non-contractive model, specifically a [PITH_FULL_IMAGE:figures/full_fig_p025_13.png]
Figure 14
Figure 14. Figure 14: Distance between induced trajectories for the policies using a contractive REN (left) and [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]
Figure 15
Figure 15. Figure 15: SCDS results for the Snake dataset demonstrate its ability to handle a more complex [PITH_FULL_IMAGE:figures/full_fig_p027_15.png]
Figure 16
Figure 16. Figure 16: Simulation and real-world deployment pipelines. The trained policy can be deployed to [PITH_FULL_IMAGE:figures/full_fig_p028_16.png]
Figure 17
Figure 17. Figure 17: Left: Invertible coupling layer structure implemented as a part of RealNVP. The affine [PITH_FULL_IMAGE:figures/full_fig_p028_17.png]
Figure 18
Figure 18. Figure 18: Efficient parallel testing of out-of-sample behavior in the Isaac Lab robotics simulator. [PITH_FULL_IMAGE:figures/full_fig_p031_18.png]
Figure 19
Figure 19. Figure 19: Expert demonstrations in the LASA dataset. [PITH_FULL_IMAGE:figures/full_fig_p032_19.png]
Figure 20
Figure 20. Figure 20: Position and orientation data for different tasks in the Robomimic dataset. [PITH_FULL_IMAGE:figures/full_fig_p033_20.png]
Figure 21
Figure 21. Figure 21: Data collection setup for Lift, Can, Square, and transport tasks (Mandlekar et al., 2021). [PITH_FULL_IMAGE:figures/full_fig_p033_21.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 48 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Apprenticeship learning via inverse reinforcement learning

    Pieter Abbeel and Andrew Y Ng. Apprenticeship learning via inverse reinforcement learning. In International C onference on M achine L earning , pp.\ 1, 2004

    work page 2004

  3. [3]

    Learning L yapunov-stable polynomial dynamical systems through imitation

    Amin Abyaneh and Hsiu-Chin Lin. Learning L yapunov-stable polynomial dynamical systems through imitation. In 7th Annual Conference on Robot Learning, 2023

    work page 2023

  4. [4]

    Globally stable neural imitation policies

    Amin Abyaneh, Mariana Sosa Guzmán, and Hsiu-Chin Lin. Globally stable neural imitation policies. In IEEE International Conference on Robotics and Automation (ICRA), pp.\ 15061--15067, 2024

    work page 2024

  5. [5]

    Neural dynamic policies for end-to-end sensorimotor learning

    Shikhar Bahl, Mustafa Mukadam, Abhinav Gupta, and Deepak Pathak. Neural dynamic policies for end-to-end sensorimotor learning. Advances in Neural Information Processing Systems, 33: 0 5058--5069, 2020

    work page 2020

  6. [6]

    Learning stable dynamical systems using contraction theory

    Caroline Blocher, Matteo Saveriano, and Dongheui Lee. Learning stable dynamical systems using contraction theory. In 14th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI), pp.\ 124--129, 2017

    work page 2017

  7. [7]

    A unified framework for walking and running of bipedal robots

    Mahrokh Ghoddousi Boroujeni, Elham Daneshman, Ludovic Righetti, and Majid Khadiv. A unified framework for walking and running of bipedal robots. In 2021 20th International Conference on Advanced Robotics (ICAR), pp.\ 396--403. IEEE, 2021

    work page 2021

  8. [8]

    A PAC-Bayesian Framework for Optimal Control with Stability Guarantees

    Mahrokh Ghoddousi Boroujeni, Clara Luc \' a Galimberti, Andreas Krause, and Giancarlo Ferrari-Trecate. A P A C - B ayesian framework for optimal control with stability guarantees. arXiv preprint arXiv:2403.17790, 2024

    work page Pith review arXiv 2024

Show all 54 references
  1. [9]

    Neural ordinary differential equations

    Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordinary differential equations. Advances in N eural I nformation P rocessing S ystems , 31, 2018

    2018

  2. [10]

    Diffusion policy: V isuomotor policy learning via action diffusion

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: V isuomotor policy learning via action diffusion. The International Journal of Robotics Research, pp.\ 02783649241273668, 2023

    2023

  3. [11]

    Soft- DTW : a differentiable loss function for time-series

    Marco Cuturi and Mathieu Blondel. Soft- DTW : a differentiable loss function for time-series. In International C onference on M achine L earning , pp.\ 894--903. PMLR, 2017

    2017

  4. [12]

    Safe control with learned certificates: A survey of neural lyapunov, barrier, and contraction methods for robotics and control

    Charles Dawson, Sicun Gao, and Chuchu Fan. Safe control with learned certificates: A survey of neural lyapunov, barrier, and contraction methods for robotics and control. IEEE Transactions on Robotics, 39 0 (3): 0 1749--1767, 2023

    2023

  5. [13]

    An introduction to chaotic dynamical systems

    Robert L Devaney. An introduction to chaotic dynamical systems. CRC press, 2021

    2021

  6. [14]

    Density estimation using real NVP

    Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. Density estimation using real NVP . In International Conference on Learning Representations, 2017

    2017

  7. [15]

    A physically-consistent bayesian non-parametric mixture model for dynamical system learning

    Nadia Figueroa and Aude Billard. A physically-consistent bayesian non-parametric mixture model for dynamical system learning. In 2nd Annual Conference on Robot Learning, pp.\ 927--946, 2018

    2018

  8. [16]

    Learning robust rewards with adversarial inverse reinforcement learning

    Justin Fu, Katie Luo, and Sergey Levine. Learning robust rewards with adversarial inverse reinforcement learning. In International Conference on Learning Representations, 2018

    2018

  9. [17]

    Generative adversarial imitation learning

    Jonathan Ho and Stefano Ermon. Generative adversarial imitation learning. Advances in N eural I nformation P rocessing S ystems , 29: 0 4572–4580, 2016

    2016

  10. [18]

    Imitation learning: A survey of learning methods

    Ahmed Hussein, Mohamed Medhat Gaber, Eyad Elyan, and Chrisina Jayne. Imitation learning: A survey of learning methods. ACM Computing Surveys (CSUR), 50 0 (2): 0 1--35, 2017

    2017

  11. [19]

    Exact indexing of dynamic time warping

    Eamonn Keogh and Chotirat Ann Ratanamahatana. Exact indexing of dynamic time warping. Knowledge and I nformation S ystems , 7: 0 358--386, 2005

    2005

  12. [20]

    Learning stable nonlinear dynamical systems with G aussian mixture models

    S Mohammad Khansari-Zadeh and Aude Billard. Learning stable nonlinear dynamical systems with G aussian mixture models. IEEE Transactions on Robotics, 27 0 (5): 0 943--957, 2011

    2011

  13. [21]

    Learning control L yapunov function to ensure stability of dynamical system-based robot reaching motions

    S Mohammad Khansari - Zadeh and Aude Billard. Learning control L yapunov function to ensure stability of dynamical system-based robot reaching motions. Robotics and Autonomous Systems, 62 0 (6): 0 752--765, 2014

    2014

  14. [22]

    Learning stable deep dynamics models

    J Zico Kolter and Gaurav Manek. Learning stable deep dynamics models. Advances in N eural I nformation P rocessing S ystems , 32, 2019

    2019

  15. [23]

    D ART : Noise injection for robust imitation learning

    Michael Laskey, Jonathan Lee, Roy Fox, Anca Dragan, and Ken Goldberg. D ART : Noise injection for robust imitation learning. In 1st Conference on R obot L earning , 2017

    2017

  16. [24]

    Winfried Lohmiller and Jean-Jacques E. Slotine. On contraction analysis for non-linear systems. Autom., 34: 0 683--696, 1998

    1998

  17. [25]

    Manchester and Jean-Jacques E

    Ian R. Manchester and Jean-Jacques E. Slotine. Control contraction metrics: Convex and intrinsic criteria for nonlinear feedback design. IEEE Transactions on Automatic Control, 62: 0 3046--3053, 2015

    2015

  18. [26]

    GTI: Learning to Generalize across Long-Horizon Tasks from Human Demonstrations

    Ajay Mandlekar, Danfei Xu, Roberto Martín-Martín, Silvio Savarese, and Li Fei-Fei. GTI: Learning to Generalize across Long-Horizon Tasks from Human Demonstrations . In Proceedings of Robotics: Science and Systems, Corvalis, Oregon, USA, July 2020

    2020

  19. [27]

    What matters in learning from offline human demonstrations for robot manipulation

    Ajay Mandlekar, Danfei Xu, Josiah Wong, Soroush Nasiriany, Chen Wang, Rohun Kulkarni, Li Fei-Fei, Silvio Savarese, Yuke Zhu, and Roberto Mart \' n-Mart \' n. What matters in learning from offline human demonstrations for robot manipulation. In 5th Annual Conference on Robot Le...

    2021

  20. [28]

    Learning to optimize with convergence guarantees using nonlinear system theory

    Andrea Martin and Luca Furieri. Learning to optimize with convergence guarantees using nonlinear system theory. IEEE Control Systems Letters, 2024

    2024

  21. [29]

    Manchester, Luca Furieri, and Giancarlo Ferrari-Trecate

    Daniele Martinelli, Clara Lucía Galimberti, Ian R. Manchester, Luca Furieri, and Giancarlo Ferrari-Trecate. Unconstrained parametrization of dissipative and contracting neural ordinary differential equations. In 2023 62nd IEEE Conference on Decision and Control (CDC), pp.\ 304...

    2023

  22. [30]

    Kochenderfer

    Kunal Menda, Katherine Driggs-Campbell, and Mykel J. Kochenderfer. Ensemble D A gger: A bayesian approach to safe imitation learning. In 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp.\ 5041--5048, 2019

    2019

  23. [31]

    Orbit: A unified simulation framework for interactive robot learning environments

    Mayank Mittal, Calvin Yu, Qinxi Yu, Jingzhou Liu, Nikita Rudin, David Hoeller, Jia Lin Yuan, Ritvik Singh, Yunrong Guo, Hammad Mazhar, Ajay Mandlekar, Buck Babich, Gavriel State, Marco Hutter, and Animesh Garg. Orbit: A unified simulation framework for interactive robot learni...

    2023

  24. [32]

    Neural contractive dynamical systems

    Hadi Beik Mohammadi, S ren Hauberg, Georgios Arvanitidis, Nadia Figueroa, Gerhard Neumann, and Leonel Rozo. Neural contractive dynamical systems. In The Twelfth International Conference on Learning Representations, 2024

    2024

  25. [33]

    Learning robot motions with stable dynamical systems under diffeomorphic transformations

    Klaus Neumann and Jochen J Steil. Learning robot motions with stable dynamical systems under diffeomorphic transformations. Robotics and Autonomous Systems, 70: 0 1--15, 2015

    2015

  26. [34]

    Normalizing flows for probabilistic modeling and inference

    George Papamakarios, Eric Nalisnick, Danilo Jimenez Rezende, Shakir Mohamed, and Balaji Lakshminarayanan. Normalizing flows for probabilistic modeling and inference. Journal of Machine Learning Research, 22 0 (57): 0 1--64, 2021

    2021

  27. [35]

    P y T orch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  28. [36]

    Variational discriminator bottleneck: Improving imitation learning, inverse RL , and GAN s by constraining information flow

    Xue Bin Peng, Angjoo Kanazawa, Sam Toyer, Pieter Abbeel, and Sergey Levine. Variational discriminator bottleneck: Improving imitation learning, inverse RL , and GAN s by constraining information flow. In International Conference on Learning Representations, 2019

    2019

  29. [37]

    R. Penrose. On best approximate solutions of linear matrix equations. Mathematical Proceedings of the Cambridge Philosophical Society, 52 0 (1): 0 17–19, 1956. doi:10.1017/S0305004100030929

    1956 doi

  30. [38]

    ALVINN : An autonomous land vehicle in a neural network

    Dean A Pomerleau. ALVINN : An autonomous land vehicle in a neural network. In Advances in N eural I nformation P rocessing S ystems , volume 1, pp.\ 305--313, 1988

    1988

  31. [39]

    Euclideanizing flows: Diffeomorphic reduction for learning stable dynamical systems

    Muhammad Asif Rana, Anqi Li, Dieter Fox, Byron Boots, Fabio Ramos, and Nathan Ratliff. Euclideanizing flows: Diffeomorphic reduction for learning stable dynamical systems. In Learning for Dynamics and Control, pp.\ 630--639. PMLR, 2020

    2020

  32. [40]

    Learning partially contracting dynamical systems from demonstrations

    Harish Ravichandar, Iman Salehi, and Ashwin Dani. Learning partially contracting dynamical systems from demonstrations. In 1st Annual Conference on Robot Learning, 2017

    2017

  33. [41]

    Recent advances in robot learning from demonstration

    Harish Ravichandar, Athanasios S Polydoros, Sonia Chernova, and Aude Billard. Recent advances in robot learning from demonstration. Annual review of control, robotics, and autonomous systems, 3 0 (1): 0 297--330, 2020

    2020

  34. [42]

    Recurrent equilibrium networks: Flexible dynamic models with guaranteed stability and robustness

    Max Revay, Ruigang Wang, and Ian R Manchester. Recurrent equilibrium networks: Flexible dynamic models with guaranteed stability and robustness. IEEE Transactions on Automatic Control, 2023

    2023

  35. [43]

    Efficient reductions for imitation learning

    St \'e phane Ross and Drew Bagnell. Efficient reductions for imitation learning. In Proceedings of the thirteenth international conference on artificial intelligence and statistics, pp.\ 661--668. JMLR Workshop and Conference Proceedings, 2010

    2010

  36. [44]

    Learning contracting vector fields for stable imitation learning

    Vikas Sindhwani, Stephen Tu, and Mohi Khansari. Learning contracting vector fields for stable imitation learning. arXiv preprint arXiv:1804.04878, 2018

    2018 arXiv

  37. [45]

    Learning deep dynamical systems using stable neural O D E s

    Andreas Sochopoulos, Michael Gienger, and Sethu Vijayakumar. Learning deep dynamical systems using stable neural O D E s. In Proceedings of the 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2024

    2024

  38. [46]

    A family of nonparametric density estimation algorithms

    Esteban G Tabak and Cristina V Turner. A family of nonparametric density estimation algorithms. Communications on Pure and Applied Mathematics, 66 0 (2): 0 145--164, 2013

    2013

  39. [47]

    Behavioral cloning from observation

    Faraz Torabi, Garrett Warnell, and Peter Stone. Behavioral cloning from observation. In Proceedings of the 27th International Joint Conference on Artificial Intelligence, pp.\ 4950--4957, 2018

    2018

  40. [48]

    Contraction theory for nonlinear stability analysis and learning-based control: A tutorial overview

    Hiroyasu Tsukamoto, Soon-Jo Chung, and Jean-Jaques E Slotine. Contraction theory for nonlinear stability analysis and learning-based control: A tutorial overview. Annual Reviews in Control, 52: 0 135--169, 2021

    2021

  41. [49]

    P. Erdös I. Vincze. On the approximation of convex, closed plane curves by multifocal ellipses. Journal of Applied Probability, 19 0 (A): 0 89–96, 1982

    1982

  42. [50]

    Learning R iemannian stable dynamical systems via diffeomorphisms

    Jiechao Zhang, Hadi Beik Mohammadi, and Leonel Rozo. Learning R iemannian stable dynamical systems via diffeomorphisms. In 6th Annual Conference on Robot Learning, 2022

    2022

  43. [51]

    Maximum entropy inverse reinforcement learning

    Brian D Ziebart, Andrew L Maas, J Andrew Bagnell, Anind K Dey, et al. Maximum entropy inverse reinforcement learning. In Association for the Advancement of Artificial Intelligence, volume 8, pp.\ 1433--1438, 2008

    2008

  44. [52]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  45. [53]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  46. [54]

    Lift”, “Can

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.