pith. sign in

arxiv: 2606.09108 · v1 · pith:OQSN7PYZnew · submitted 2026-06-08 · 💻 cs.RO · cs.LG

RAM: Reachability Across Morphologies

Tim Walter , Xinyu Chen , Jonathan K\"ulz , Matthias Althoff This is my paper

Pith reviewed 2026-06-27 16:40 UTC · model grok-4.3

classification 💻 cs.RO cs.LG
keywords reachabilitymorphologyimplicit neural representationrobot workspaceself-collisionsforward kinematicsdifferentiable surrogaterobotics
0
0 comments X

The pith

A morphology-conditioned neural network serves as a fast differentiable surrogate for robot pose reachability that generalizes to unseen body shapes.

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

Robotic tasks from design to motion planning require accurate knowledge of reachable poses, yet current workspace methods remain slow, imprecise, or fixed to one robot shape. RAM trains an implicit neural representation on a dataset of thirty billion forward-kinematics samples to predict reachability while automatically handling self-collisions. The resulting model runs in nanoseconds, achieves an 86 percent F1 score, and generalizes to morphologies never seen during training. When used inside gradient-based loops it accelerates morphology optimization by one order of magnitude and trajectory optimization by two orders of magnitude. A sympathetic reader would therefore treat RAM as a drop-in replacement for expensive geometric or sampling-based reachability checks across families of robots.

Core claim

We introduce Reachability Across Morphologies (RAM): a morphology-conditioned, implicit neural representation that acts as a fast, differentiable surrogate for pose reachability, generalising to unseen morphologies while inherently accounting for self-collisions. Trained solely on forward-kinematics data, the model delivers an F1-score of 86 percent at nanosecond inference, outperforming the baseline by 14 percent and reducing inference time by three orders of magnitude.

What carries the argument

morphology-conditioned implicit neural representation that maps robot shape parameters together with a candidate pose to a reachability probability

If this is right

  • Reachability queries become feasible inside real-time control or simulation loops.
  • Gradient-based morphology optimization can be performed at interactive speeds.
  • Trajectory optimization benefits from two-order-of-magnitude speed-ups without sacrificing collision awareness.
  • A single trained model replaces per-morphology workspace computations for families of robots.
  • Differentiability enables end-to-end learning pipelines that include reachability constraints.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same implicit-representation approach could be extended to other robot properties such as force transmission or stability margins.
  • Real-time deployment inside reinforcement-learning environments would allow agents to learn morphology-aware policies without explicit collision engines.
  • Dataset scaling laws observed for forward kinematics might apply to other kinematic or dynamic properties if similar volume of data can be generated.
  • Integration with differentiable simulators would let designers jointly optimize morphology, actuation, and control under reachability constraints.

Load-bearing premise

Samples generated only from forward kinematics contain enough information for the network to learn both generalization across shapes and correct self-collision behavior.

What would settle it

Run the trained model on a held-out morphology whose self-collision patterns differ from the training distribution and measure whether it assigns high reachability scores to poses that physically intersect.

Figures

Figures reproduced from arXiv: 2606.09108 by Jonathan K\"ulz, Matthias Althoff, Tim Walter, Xinyu Chen.

Figure 1
Figure 1. Figure 1: We present RAM, a surrogate model for robot reachability. Trained on a diverse dataset of [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example link types with columns representing length-only, offset-only, and both. The [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hierarchical rejection sampling of morphologies. Each stage samples new parameters and [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example of a work￾space slice with reachable poses indicated by dark pix￾els. The over-approximation trend in RAM is consistent with its train￾ing data and the implicit smoothing of the MLP architecture [17], resulting in a majority of false positives among misclassified sam￾ples, as reported in Section K. We also evaluated RAM out-of-distribution on morphologies with {1-4, 8, 9} DoF. RAM maintained F1-sco… view at source ↗
Figure 5
Figure 5. Figure 5: Result of a morphology optimisation compared to the initial morphology. Task poses are [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Many stages of the robotic lifecycle, from morphology synthesis to operation, rely fundamentally on the reachable workspace. However, current methods for approximating workspaces are slow, imprecise, or tied to a single morphology. We introduce Reachability Across Morphologies (RAM): a morphology-conditioned, implicit neural representation that acts as a fast, differentiable surrogate for pose reachability, generalising to unseen morphologies while inherently accounting for self-collisions. To train RAM, we publish a large-scale dataset of $3\cdot10^{10}$ samples generated solely from forward kinematics. Experiments show that our model achieves an $ F_1$-score of $86\%$ at nanosecond inference, outperforming the baseline by $14\%$ while reducing inference time by three orders of magnitude. We further demonstrate speed-ups of one and two orders of magnitude for gradient-based morphology and trajectory optimisation, respectively. Website: https://timwalter.github.io/ram.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces Reachability Across Morphologies (RAM), a morphology-conditioned implicit neural representation trained on 3·10^10 forward-kinematics samples that serves as a fast, differentiable surrogate for pose reachability. It claims to generalize to unseen morphologies while inherently accounting for self-collisions, reports an F1-score of 86% at nanosecond inference (14% above baseline), and demonstrates speed-ups in gradient-based morphology and trajectory optimization.

Significance. If the generalization and self-collision claims hold with proper validation, the work would supply a practical, morphology-agnostic tool for workspace queries that could accelerate morphology synthesis, planning, and optimization loops in robotics by orders of magnitude.

major comments (2)
  1. [Abstract] Abstract: the central claim that RAM 'inherently accounting for self-collisions' lacks support in the described training procedure. The dataset is generated solely from forward kinematics, which produces only positive (reachable) end-effector poses without mesh intersection tests, negative labels, or any collision signal; consequently the implicit function can at best reproduce the kinematic workspace and cannot discover collision-induced unreachability.
  2. [Abstract] Abstract and experiments section: aggregate F1 and timing numbers are reported without any description of the train/test split, morphology sampling strategy, self-collision labeling procedure (if any), or ablation on generalization to held-out morphologies; these omissions make it impossible to verify that the reported performance actually supports the generalization claim.
minor comments (1)
  1. [Abstract] The abstract states the dataset size as $3\cdot10^{10}$ but supplies no information on how morphologies were sampled or how the implicit representation is conditioned on morphology parameters.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful review and constructive feedback on our work. We address each major comment below and commit to revisions where the concerns are valid.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that RAM 'inherently accounting for self-collisions' lacks support in the described training procedure. The dataset is generated solely from forward kinematics, which produces only positive (reachable) end-effector poses without mesh intersection tests, negative labels, or any collision signal; consequently the implicit function can at best reproduce the kinematic workspace and cannot discover collision-induced unreachability.

    Authors: We agree that the referee's analysis is correct. The dataset generation uses only forward kinematics to produce positive reachable samples and includes no mesh intersection tests, negative labels, or collision signals. Consequently, the model approximates the kinematic workspace and cannot inherently account for self-collisions. We will revise the abstract, introduction, and any related claims to remove or qualify the self-collision statement and accurately describe the model's scope as a kinematic reachability surrogate. revision: yes

  2. Referee: [Abstract] Abstract and experiments section: aggregate F1 and timing numbers are reported without any description of the train/test split, morphology sampling strategy, self-collision labeling procedure (if any), or ablation on generalization to held-out morphologies; these omissions make it impossible to verify that the reported performance actually supports the generalization claim.

    Authors: We acknowledge the omissions in the abstract and the need for clearer experimental details. While the full manuscript describes the overall dataset of 3·10^10 FK samples and reports aggregate metrics, it does not sufficiently detail the train/test split, morphology sampling strategy for held-out testing, or include an explicit ablation on generalization. We will expand the abstract with a concise description of these elements, confirm that no self-collision labeling was used, and add a dedicated ablation subsection in the experiments to demonstrate performance on unseen morphologies. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The paper presents an empirical neural network trained on an externally generated forward-kinematics dataset of 3·10^10 samples and evaluated on held-out generalization metrics. No equations, parameters, or claims in the provided text reduce reported performance (F1-score, inference time) to quantities defined by the model itself or by self-citation chains. The central claim of a morphology-conditioned implicit representation is an approximation learned from independent data rather than a self-referential construction. The self-collision accounting is an empirical assertion about what the trained model captures, not a definitional reduction of the result to its inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Abstract-only review; full architecture, loss, and data-generation details unavailable.

free parameters (1)
  • Neural network architecture and hyperparameters
    Standard learned parameters of any implicit representation; exact values and selection procedure not stated.
axioms (1)
  • domain assumption Forward-kinematics samples alone suffice to label reachability including self-collisions for arbitrary morphologies
    Training and generalization rest on this assumption; location: implicit in the sentence 'generated solely from forward kinematics'.

pith-pipeline@v0.9.1-grok · 5689 in / 1279 out tokens · 23961 ms · 2026-06-27T16:40:08.247424+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

65 extracted references · 7 canonical work pages

  1. [1]

    J. Külz, S. Ha, and M. Althoff. A Design Co-Pilot for Task-Tailored Manipulators. arXiv:2509.13077, 2025

    work page arXiv 2025

  2. [2]

    Y . Dai, S. He, X. Nie, X. Rui, S. Li, and S. He. Research on Reconfiguration Strategies for Self-reconfiguring Modular Robots: A Review.Journal of Intelligent & Robotic Systems, 110 (2), 2024

    2024

  3. [3]

    M. Qin, J. Brawer, and B. Scassellati. Robot tool use: A survey.Frontiers in Robotics and AI, 9, 2023

    2023

  4. [4]

    Deroo, B

    B. Deroo, B. Pousett, E. Aertbeliën, W. Decré, and H. Bruyninckx. The Need for Task-Specific Execution in Robot Manipulation: Skill Design for Energy-Efficient Control. InIEEE Int. Conf. on Automation Science and Engineering (CASE), pages 1–7, 2023

    2023

  5. [5]

    Schiele, S

    S. Schiele, S. Baumgartner, S. Laudahn, and T. C. Lueth. Automated design of task specific additively manufacturable coupled serial chain mechanisms for tracing predefined planar tra- jectories. InProc. of the IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), pages 3047–3052, 2022

    2022

  6. [6]

    E. M. Hoffman, D. Costanzi, G. Fadini, N. Miguel, A. D. Prete, and L. Marchionni. Address- ing Reachability and Discrete Component Selection in Robotic Manipulator Design Through Kineto-Static Bi-Level Optimization.IEEE Robotics and Automation Letters, 10(3):2263– 2270, 2025

    2025

  7. [7]

    S. Ha, S. Coros, A. Alspach, J. Kim, and K. Yamane. Computational co-optimization of design parameters and motion trajectories for robotic systems.The International Journal of Robotics Research, 37(13-14):1521–1536, 2018

    2018

  8. [8]

    R. S. Hartenberg and J. Denavit.Kinematic Synthesis of Linkages. McGraw-Hill, 1964

    1964

  9. [9]

    Shirafuji and K

    S. Shirafuji and K. Shimamura. Kinematic Synthesis of a Serial Manipulator Using Gradient- Based Optimization on Lie Groups.IEEE Robotics and Automation Letters, 10(3):2550–2557, 2025

    2025

  10. [10]

    K. M. Lynch and F. C. Park.Modern Robotics. Cambridge University Press, 2017

    2017

  11. [11]

    D. L. Pieper.The Kinematics of Manipulators under Computer Control. Stanford University, 1969

    1969

  12. [12]

    Zacharias, C

    F. Zacharias, C. Borst, and G. Hirzinger. Capturing robot workspace structure: Representing robot capabilities. InProc. of the IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), pages 3229–3236, 2007

    2007

  13. [13]

    Chirikjian and I

    G. Chirikjian and I. Ebert-Uphoff. Numerical convolution on the Euclidean group with ap- plications to workspace generation.IEEE Transactions on Robotics and Automation, 14(1): 123–136, 1998

    1998

  14. [14]

    Vahrenkamp, T

    N. Vahrenkamp, T. Asfour, G. Metta, G. Sandini, and R. Dillmann. Manipulability analysis. InProc. of the IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids), pages 568–573, 2012. 10

    2012

  15. [15]

    Y . Han, J. Pan, M. Xia, L. Zeng, and Y .-J. Liu. Efficient SE(3) Reachability Map Generation via Interplanar Integration of Intra-planar Convolutions. InProc. of the IEEE Int. Conf. on Robotics and Automation (ICRA), pages 1854–1860, 2021

    2021

  16. [16]

    Leibrandt, L

    K. Leibrandt, L. da Cruz, and C. Bergeles. Designing Robots for Reachability and Dexterity: Continuum Surgical Robots as a Pretext Application.IEEE Transactions on Robotics, 39(4): 2989–3007, 2023

    2023

  17. [17]

    Sitzmann, J

    V . Sitzmann, J. N. Martel, A. W. Bergman, D. B. Lindell, and G. Wetzstein. Implicit neural rep- resentations with periodic activation functions. InProc. of the Int. Conf. on Neural Information Processing Systems (NeurIPS), pages 7462–7473, 2020

    2020

  18. [18]

    Mescheder, M

    L. Mescheder, M. Oechsle, M. Niemeyer, S. Nowozin, and A. Geiger. Occupancy networks: Learning 3D reconstruction in function space. InProc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), pages 4460–4470, 2019

    2019

  19. [19]

    Limoyo, F

    O. Limoyo, F. Mari ´c, M. Giamou, P. Alexson, I. Petrovi´c, and J. Kelly. Generative Graphical Inverse Kinematics.IEEE Transactions on Robotics, 41:1002–1018, 2025

    2025

  20. [20]

    Diankov.Automated Construction of Robotic Manipulation Programs

    R. Diankov.Automated Construction of Robotic Manipulation Programs. PhD thesis, Carnegie Mellon University, 2018

    2018

  21. [21]

    Ostermeier, J

    D. Ostermeier, J. Külz, and M. Althoff. Automatic Geometric Decomposition for Analytical Inverse Kinematics.IEEE Robotics and Automation Letters, 10(10):9964–9971, 2025

    2025

  22. [22]

    Yasutake, Z

    C. Yasutake, Z. Kingston, and B. Plancher. HJCD-IK: GPU-accelerated inverse kinematics through batched hybrid jacobian coordinate descent. arXiv:2510.07514, 2025

    work page arXiv 2025

  23. [23]

    Beeson and B

    P. Beeson and B. Ames. TRAC-IK: An open-source library for improved solving of generic inverse kinematics. InProc. of the IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids), pages 928–935, 2015

    2015

  24. [24]

    Sundaralingam, S

    B. Sundaralingam, S. K. S. Hari, A. Fishman, C. Garrett, K. Van Wyk, V . Blukis, A. Millane, H. Oleynikova, A. Handa, F. Ramos, N. Ratliff, and D. Fox. CuRobo: Parallelized collision- free robot motion generation. InProc. of the IEEE Int. Conf. on Robotics and Automation (ICRA), pages 8112–8119, 2023

    2023

  25. [25]

    von Oehsen, A

    T. von Oehsen, A. Fabisch, S. Kumar, and F. Kirchner. Comparison of Distal Teacher Learning with Numerical and Analytical Methods to Solve Inverse Kinematics for Rigid-Body Mecha- nisms. arXiv:2003.00225, 2020

    work page arXiv 2003

  26. [26]

    B. Ames, J. Morgan, and G. Konidaris. IKFlow: Generating diverse inverse kinematics solu- tions.IEEE Robotics and Automation Letters, 7(3):7177–7184, 2022

    2022

  27. [27]

    Bensadoun, S

    R. Bensadoun, S. Gur, N. Blau, and L. Wolf. Neural inverse kinematic. InProc. of the Int. Conf. on Machine Learning (ICML), pages 1787–1797, 2022

    2022

  28. [28]

    Abdel-Malek and J

    K. Abdel-Malek and J. Yang. Workspace boundaries of serial manipulators using manifold stratification.The International Journal of Advanced Manufacturing Technology, 28(11): 1211–1229, 2006

    2006

  29. [29]

    R. F. Cavelli, P. D. Cen Cheng, and M. Indri. Modeling the Reachability Space of Robotic Manipulators through Ellipsoid Equations.Journal of Intelligent & Robotic Systems, 111(3), 2025

    2025

  30. [30]

    J. J. Park, P. Florence, J. Straub, R. Newcombe, and S. Lovegrove. DeepSDF: Learning con- tinuous signed distance functions for shape representation. InProc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), pages 165–174, 2019. 11

    2019

  31. [31]

    Mildenhall, P

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng. NeRF: Representing scenes as neural radiance fields for view synthesis. arXiv:2003.08934, 2020

    work page arXiv 2003

  32. [32]

    M. Z. Irshad, M. Comi, Y .-C. Lin, N. Heppert, A. Valada, R. Ambrus, Z. Kira, and J. Tremblay. Neural Fields in Robotics: A Survey. arXiv:2410.20220, 2024

    work page arXiv 2024

  33. [33]

    Y . Li, Y . Zhang, A. Razmjoo, and S. Calinon. Representing Robot Geometry as Distance Fields: Applications to Whole-body Manipulation. arXiv:2307.00533, 2024

    work page arXiv 2024

  34. [34]

    Y . Li, X. Chi, A. Razmjoo, and S. Calinon. Configuration Space Distance Fields for Manipu- lation Planning. InProc. of Robotics: Science and Systems (RSS), volume 20, 2024

    2024

  35. [35]

    Gienger, C

    A. Gienger, C. Stein, A. P. R. Lauer, O. Sawodny, and C. Tarín. Data-Based Reachability Anal- ysis and Optimized Robot Positioning for Co-Design of Construction Processes. InIEEE/SICE Int. Symp. on System Integration (SII), pages 1247–1252, 2024

    2024

  36. [36]

    Murooka, I

    M. Murooka, I. Kumagai, M. Morisawa, and F. Kanehiro. Learning Differentiable Reachability Maps for Optimization-Based Humanoid Motion Generation. InProc. of the IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids), pages 1–8, 2025

    2025

  37. [37]

    Kim and J

    S. Kim and J. Perez. Learning Reachable Manifold and Inverse Mapping for a Redundant Robot manipulator. InProc. of the IEEE Int. Conf. on Robotics and Automation (ICRA), pages 4731–4737, 2021

    2021

  38. [38]

    X. Chen, J. Külz, and M. Althoff. Generating robot capability maps with neural fields. InProc. of Robotics: Science and Systems (RSS), 2024

    2024

  39. [39]

    E. S. Conkur and R. Buckingham. Clarifying the definition of redundancy as used in robotics. Robotica, 15(5):583–586, 1997

    1997

  40. [40]

    Craig.Introduction to Robotics

    J. Craig.Introduction to Robotics. Pearson International, 2013

    2013

  41. [41]

    R. Reiter. On Closed World Data Bases. In H. Gallaire and J. Minker, editors,Logic and Data Bases, pages 55–76. Springer US, 1978

    1978

  42. [42]

    N. W. Johnson.Geometries and Transformations. Cambridge University Press, 2018

    2018

  43. [43]

    Y . Zhou, C. Barnes, J. Lu, J. Yang, and H. Li. On the Continuity of Rotation Representations in Neural Networks. InProc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), pages 5738–5746, 2019

    2019

  44. [44]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber. Long Short-Term Memory.Neural Computation, 9(8): 1735–1780, 1997

    1997

  45. [45]

    Schuck, A

    M. Schuck, A. von Rohr, and A. P. Schoellig. Scipy.spatial.transform: Differentiable Framework-Agnostic 3D Transformations in Python. InProc. of the Int. Conf. on Neural Information Processing Systems (NeurIPS), 2025

    2025

  46. [46]

    Spearman

    C. Spearman. The proof and measurement of association between two things.The American Journal of Psychology, 15(1):72–101, 1904

    1904

  47. [47]

    N. Mantel. The detection of disease clustering and a generalized regression approach.Cancer research, 27(2):209–220, 1967

    1967

  48. [48]

    Simonyan, A

    K. Simonyan, A. Vedaldi, and A. Zisserman. Deep inside convolutional networks: Visual- ising image classification models and saliency maps. InProc. of the Int. Conf. on Learning Representations (ICLR), 2014

    2014

  49. [49]

    Dosovitskiy and T

    A. Dosovitskiy and T. Brox. Inverting visual representations with convolutional networks. In Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), pages 4829– 4837, 2016. 12

    2016

  50. [50]

    Mahendran and A

    A. Mahendran and A. Vedaldi. Understanding deep image representations by inverting them. InProc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), pages 5188– 5196, 2015

    2015

  51. [51]

    Rader, T

    J. Rader, T. Lyons, and P. Kidger. Optimistix: Modular optimisation in JAX and Equinox. arXiv:2402.09983, 2024

    work page arXiv 2024

  52. [52]

    D. P. Kingma and J. Ba. Adam: A method for stochastic optimization. InProc. of the Int. Conf. on Learning Representations (ICLR), 2015

    2015

  53. [53]

    D. C. Liu and J. Nocedal. On the limited memory BFGS method for large scale optimization. Mathematical Programming, 45(1):503–528, 1989

    1989

  54. [54]

    Yoshikawa

    T. Yoshikawa. Analysis and control of robot manipulators with redundancy. InProc. of the Int. Symp. on Robotics Research (ISRR), pages 735–747, 1983

    1983

  55. [55]

    M. E. Muller. A note on a method for generating points uniformly on n-dimensional spheres. Communications of the ACM, 2(4):19–20, 1959

    1959

  56. [56]

    Shoemake

    K. Shoemake. Uniform random rotations. InGraphics Gems III, pages 124–132. Academic Press Professional, Inc., 1992

    1992

  57. [57]

    F. C. Park. Distance Metrics on the Rigid-Body Motions with Applications to Mechanism Design.Journal of Mechanical Design, 117(1):48–54, 1995

    1995

  58. [58]

    Žefran, V

    M. Žefran, V . Kumar, and C. Croke. Choice of Riemannian Metrics for Rigid Body Kinematics. InProc. of the ASME Biennial Mechanisms Conf., 1996

    1996

  59. [59]

    Akiba, S

    T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama. Optuna: A Next-generation Hyperpa- rameter Optimization Framework. InProceedings of the 25th ACM (SIGKDD) International Conference on Knowledge Discovery & Data Mining, pages 2623–2631, 2019

    2019

  60. [60]

    R. A. Fisher. Frequency distribution of the values of the correlation coefficient in samples from an indefinitely large population.Biometrika, 10(4):507–521, 1915

    1915

  61. [61]

    P. Jaccard. étude comparative de la distribution florale dans une portion des Alpes et des Jura. Bull Soc Vaudoise Sci Nat, 37:547–579, 1901

    1901

  62. [62]

    L. Ruff, R. Vandermeulen, N. Goernitz, L. Deecke, S. A. Siddiqui, A. Binder, E. Müller, and M. Kloft. Deep one-class classification. InProc. of the Int. Conf. on Machine Learning (ICML), volume 80, pages 4393–4402, 2018

    2018

  63. [63]

    Tancik, P

    M. Tancik, P. P. Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ra- mamoorthi, J. T. Barron, and R. Ng. Fourier features let networks learn high frequency func- tions in low dimensional domains. InProc. of the Int. Conf. on Neural Information Processing Systems (NeurIPS), pages 7537–7547, 2020

    2020

  64. [64]

    Saragadam, D

    V . Saragadam, D. LeJeune, J. Tan, G. Balakrishnan, A. Veeraraghavan, and R. G. Baraniuk. Wire: Wavelet implicit neural representations. InProc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), pages 18507–18516, 2023

    2023

  65. [65]

    E. Jang, S. Gu, and B. Poole. Categorical reparameterization with gumbel-softmax. InProc. of the Int. Conf. on Learning Representations (ICLR), 2017. 13 A Scissor Self-collisions Scissor self-collisions arise when adjacent capsules rotate into an anti-parallel configuration through a shared joint, causing their bodies to overlap. For each jointi, letc pre...

    2017