pith. sign in

arxiv: 2606.06094 · v1 · pith:J3CNXIIHnew · submitted 2026-06-04 · 💻 cs.AI · cs.LG· math.DS· physics.med-ph

Integrating Mechanistic and Data-Driven Models for Neurological Disorders through Differentiable Programming

Shah Pallav Dhanendrakumar , Saikat Pal , Sitikantha Roy This is my paper

Pith reviewed 2026-06-28 01:42 UTC · model grok-4.3

classification 💻 cs.AI cs.LGmath.DSphysics.med-ph
keywords hybrid modelingneurological disordersneural ODEsmechanistic modelsdifferentiable programmingbrain tumorsAlzheimer's diseasestroke
0
0 comments X

The pith

Hybrid models integrating differential equations and deep learning outperform standalone approaches in modeling neurological disorders.

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

The paper establishes that hybrid modeling strategies, which combine physics-based differential equation models with deep learning, provide superior characterization of neurological disorder evolution compared to purely mechanistic or data-driven methods. It reviews architectures such as parallel, series, and parallel-series configurations, highlighting residual modeling, Neural Ordinary Differential Equations, and solver-in-the-loop techniques. These hybrids promise to overcome limitations like simplification in mechanistic models and lack of interpretability in data-driven ones, enabling better diagnostics, prognosis, and treatment planning for disorders including brain tumors, Alzheimer's disease, and stroke. A sympathetic reader would care because this integration could lead to more accurate and personalized neurological care.

Core claim

Hybrid models that integrate governing differential equation based formulations and deep learning outperform standalone mechanistic or purely data driven approaches in characterizing the evolution of neurological disorders, with proposed configurations improving diagnosis accuracy, predicting disease progression, and informing treatment strategies.

What carries the argument

Hybrid architectures (parallel, series, parallel-series) that combine deep learning with physics-based solvers through approaches like residual modeling for missing physics, Neural Ordinary Differential Equations for continuous time dynamics, and neural approximations to accelerate solvers.

If this is right

  • Enhanced diagnosis accuracy for neurological conditions such as brain tumors, Alzheimer's disease, and stroke.
  • Better prediction of disease progression over time.
  • More informed treatment strategies based on personalized modeling.
  • Overcoming computational expense and simplification issues in mechanistic models.
  • Addressing data requirements and generalization problems in pure data-driven methods.

Where Pith is reading between the lines

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

  • Such hybrid methods could be tested on other dynamical systems in biology, like immune response modeling.
  • Development of user-friendly software libraries for these differentiable programming hybrids would accelerate adoption.
  • Clinical validation studies comparing hybrid predictions against long-term patient data would be needed to confirm benefits.

Load-bearing premise

That the hybrid configurations can be practically implemented and validated across neurological disorders without major data or integration barriers.

What would settle it

An empirical test on a dataset for Alzheimer's progression where the hybrid model fails to show measurable improvements in prediction accuracy or computational speed over the best existing pure approaches.

Figures

Figures reproduced from arXiv: 2606.06094 by Saikat Pal, Shah Pallav Dhanendrakumar, Sitikantha Roy.

Figure 1
Figure 1. Figure 1: Different approaches of adjoint state method [ [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Parallel or Residual Model (a) Architectural Configuration (b) Computational Graph for Training [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Neural ODEs Architecture (a) Evolution of hidden states governed through the ODE (b) Iterative [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Series or Solver in the Loop Model-Architectural Configuration [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Transformation of low fidelity solution trajectory (red) towards high [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Parallel-Series Architecture, the parallel path (Green), the series path (Red) [ [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: TumorTwin key components [105]. Kapteyn et al. [105] introduced TumorTwin, a modular Python framework for patient specific digital twins based on reaction diffusion PDEs initialized from imaging and treatment history. While the framework enables individualized forecasting, its reliance on simplified growth laws and coarse tumor burden measurements limits its ability to fully exploit high resolution spatial… view at source ↗
Figure 8
Figure 8. Figure 8: Hybrid Modeling for Missing or Incomplete Cerebrovascular Hemodynamical System [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Universal Differential Equations as a Common Modeling Language for Neuroscience (a)Uncover [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
read the original abstract

Advances in computational modeling, neuroimaging, and artificial intelligence are revolutionizing the modeling of neurological disorders for improved diagnostics, prognosis, and treatment planning. Mechanistic models provide valuable scientific insight into the disorders, but in practice they are often simplified with assumptions or computationally expensive and slow to solve. However, while purely data driven approaches provide speed and scalability, they require large, high quality data to train and generally suffer from interpretability and generalization issues. This perspective paper presents a structured overview of hybrid modeling strategies, which combine deep learning models with physics based solvers, and are categorized into parallel, series, and parallel-series architectures. Three main approaches that have been emphasized are residual modeling for missing or incomplete physics, Neural Ordinary Differential Equations (NODEs) for continuous time dynamics approximation, and solver in the loop that accelerates traditional solvers with neural approximations. These hybrid models integrate the governing differential equation based formulations and deep learning to characterize the evolution of neurological disorders, and promise advanced personalized neurological modeling. In addition, the study explores and proposes different hybrid configurations to improve diagnosis accuracy, predict disease progression, and inform treatment strategies across a range of neurological disorders. These capabilities outperform standalone mechanistic or purely data driven approaches, making hybrid modeling a powerful tool, especially in applications involving modeling the progression and treatment responses in neurological conditions such as brain tumors, Alzheimer's disease, and stroke.

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 / 2 minor

Summary. This perspective paper overviews hybrid modeling strategies that combine physics-based differential equation solvers with deep learning for neurological disorders. It categorizes architectures as parallel, series, and parallel-series; emphasizes residual modeling, Neural ODEs, and solver-in-the-loop techniques; and proposes configurations for brain tumors, Alzheimer's disease, and stroke. The central claim is that these hybrids integrate mechanistic and data-driven elements to outperform standalone mechanistic or purely data-driven approaches in diagnosis accuracy, progression prediction, and treatment planning.

Significance. A well-supported overview of hybrid architectures could usefully synthesize existing ideas for the neurology modeling community and highlight differentiable programming as an integration tool. However, because the manuscript supplies no derivations, implementations, benchmarks, or validation results, the asserted performance gains remain conceptual and the practical significance cannot yet be assessed.

major comments (2)
  1. [Abstract] Abstract: the statement that the proposed hybrid configurations 'outperform standalone mechanistic or purely data driven approaches' is presented as established fact, yet the manuscript contains no quantitative comparisons, error metrics, datasets, or validation studies to support this load-bearing claim.
  2. [Proposed hybrid configurations] Discussion of hybrid configurations for brain tumors, Alzheimer's, and stroke: the text proposes specific architectures but supplies no implementation details, solver integration code, or even schematic equations demonstrating how the governing DEs are coupled to the neural components, leaving the feasibility of the claimed integration unverified.
minor comments (2)
  1. The manuscript would benefit from explicit citations to prior hybrid modeling work in other domains (e.g., physics-informed neural networks or Neural ODE applications in systems biology) to situate the proposed neurological configurations.
  2. Terminology such as 'solver in the loop' and 'parallel-series architectures' is introduced without a dedicated definitions subsection or diagram, which may reduce accessibility for readers outside the immediate subfield.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed review of our perspective paper. We agree that the manuscript is conceptual in nature and does not contain new empirical validations or implementations. Below we address the major comments directly, with revisions planned where appropriate to better reflect the perspective format.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the statement that the proposed hybrid configurations 'outperform standalone mechanistic or purely data driven approaches' is presented as established fact, yet the manuscript contains no quantitative comparisons, error metrics, datasets, or validation studies to support this load-bearing claim.

    Authors: We agree the abstract wording overstates the claim as established fact. As this is a perspective paper synthesizing existing hybrid modeling ideas rather than reporting new experiments, we will revise the abstract to indicate that the proposed configurations 'have the potential to outperform' standalone approaches by combining mechanistic insight with data-driven flexibility, consistent with the cited literature on residual learning and Neural ODEs. revision: yes

  2. Referee: [Proposed hybrid configurations] Discussion of hybrid configurations for brain tumors, Alzheimer's, and stroke: the text proposes specific architectures but supplies no implementation details, solver integration code, or even schematic equations demonstrating how the governing DEs are coupled to the neural components, leaving the feasibility of the claimed integration unverified.

    Authors: We acknowledge that the perspective format does not include code or full implementations. To improve clarity on feasibility, we will add schematic equations in the revised manuscript illustrating the coupling of governing differential equations with neural components (e.g., residual terms or Neural ODE formulations) for the brain tumor, Alzheimer's, and stroke configurations. revision: partial

Circularity Check

0 steps flagged

No circularity: purely descriptive perspective paper with no derivations or equations

full rationale

The paper is explicitly a perspective/overview piece that describes hybrid architectures (parallel/series/parallel-series), residual modeling, NODEs, and solver-in-the-loop concepts at a high level and proposes configurations for disorders such as brain tumors and Alzheimer's. No equations, derivations, fitted parameters, or quantitative results are presented anywhere in the text, so there is no derivation chain that could reduce to self-definition, fitted inputs, or self-citation. The outperformance statement is a forward-looking claim without any supporting math or data reduction steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

As a perspective overview paper, no new free parameters, axioms, or invented entities are introduced or relied upon beyond standard concepts already present in the cited literature on hybrid modeling.

pith-pipeline@v0.9.1-grok · 5784 in / 1166 out tokens · 45535 ms · 2026-06-28T01:42:54.214357+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

130 extracted references · 24 canonical work pages

  1. [1]

    Exploring human brain mechanics by combining experiments, modeling, and simula- tion

    Silvia Budday. “Exploring human brain mechanics by combining experiments, modeling, and simula- tion”. In:Brain Multiphysics5 (2023), p. 100076.URL:https://www.sciencedirect.com/s cience/article/pii/S266652202300014X

    2023

  2. [2]

    Fifty Shades of Brain: A Review on the Mechanical Testing and Modeling of Brain Tissue

    Silvia Budday et al. “Fifty Shades of Brain: A Review on the Mechanical Testing and Modeling of Brain Tissue”. In:Archives of Computational Methods in Engineering27 (2020), 1187–1230

    2020

  3. [3]

    Human Brain Disorders: A Review

    Naz Falaq and Siddique Yasir Hasan. “Human Brain Disorders: A Review”. In:The Open Biology Journal8 (2020).URL:https://www.openbiologyjournal.com/abs/10.2174/18741 96702008010006

    work page doi:10.2174/18741 2020

  4. [4]

    A review of brain injury at multiple time scales and its clinicopathological correlation through in silico modeling

    Abhilash Awasthi et al. “A review of brain injury at multiple time scales and its clinicopathological correlation through in silico modeling”. In:Brain Multiphysics6 (2024), p. 100090

    2024

  5. [5]

    June 2019.ISBN: 978-1-4939-9810-4

    Corina Drapaca and Sivabal Sivaloganathan.Mathematical Modelling and Biomechanics of the Brain. June 2019.ISBN: 978-1-4939-9810-4

    2019

  6. [6]

    Biomechanical modelling and computer aided simulation of deep brain retrac- tion in neurosurgery

    Abhilash Awasthi et al. “Biomechanical modelling and computer aided simulation of deep brain retrac- tion in neurosurgery”. In:Computer methods and programs in biomedicine197 (2020), p. 105688

    2020

  7. [7]

    Modeling chemical processes using prior knowledge and neural networks

    Michael L. Thompson and Mark A. Kramer. “Modeling chemical processes using prior knowledge and neural networks”. In:AIChE Journal40.8 (1994), pp. 1328–1340.URL:https://aiche.online library.wiley.com/doi/abs/10.1002/aic.690400806

    work page doi:10.1002/aic.690400806 1994

  8. [8]

    Combining Semi-Physical and Neural Network Modeling: An Example ofIts Usefulness

    U. Forssell and P. Lindskog. “Combining Semi-Physical and Neural Network Modeling: An Example ofIts Usefulness”. In:IFAC Proceedings Volumes30.11 (1997). IFAC Symposium on System Identifi- cation (SYSID’97), Kitakyushu, Fukuoka, Japan, 8-11 July 1997, pp. 767–770.ISSN: 1474-6670.URL: https://www.sciencedirect.com/science/article/pii/S1474667017429387

    1997

  9. [9]

    Differentiable modelling to unify machine learning and physical models for geosciences

    Chaopeng Shen et al. “Differentiable modelling to unify machine learning and physical models for geosciences”. In:Nature Reviews Earth & Environment4 (2023)

    2023

  10. [10]

    NeuralSim: Augmenting Differentiable Simulators with Neural Networks

    Eric Heiden et al. “NeuralSim: Augmenting Differentiable Simulators with Neural Networks”. In:2021 IEEE International Conference on Robotics and Automation (ICRA). 2021, pp. 9474–9481

    2021

  11. [11]

    Artificial Intelligence and Disease Modeling: Focus on Neurological Disorders

    Benjamin Ribba et al. “Artificial Intelligence and Disease Modeling: Focus on Neurological Disorders”. In:Clinical Pharmacology & Therapeutics115.6 (2024), pp. 1208–1211.URL:https://ascpt.o nlinelibrary.wiley.com/doi/abs/10.1002/cpt.3253

    work page doi:10.1002/cpt.3253 2024

  12. [12]

    Solving Forward and Inverse Problems with Differentiable Physics and Deep Learning

    Philipp Marcel Holl. “Solving Forward and Inverse Problems with Differentiable Physics and Deep Learning”. PhD thesis. Technische Universit¨at M¨unchen, 2024.URL:https://mediatum.ub.tu m.de/doc/1739592/1739592.pdf

    arXiv 2024

  13. [13]

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

    Maziar Raissi, Paris Perdikaris, and George Em Karniadakis. “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differen- tial equations”. In:Journal of Computational Physics378 (2019), pp. 686–707.URL:https://doi .org/10.1016/j.jcp.2018.10.045

    work page doi:10.1016/j.jcp.2018.10.045 2019

  14. [14]

    Mohammad Sarabian, Hessam Babaee, and Kaveh Laksari.Physics-informed neural networks for im- proving cerebral hemodynamics predictions. 2021. arXiv:2108.11498.URL:https://arxiv.o rg/abs/2108.11498. 20

    arXiv 2021

  15. [15]

    Parameter Estimation in Brain Dynamics Models from Resting-State fMRI Data using Physics-Informed Neural Networks

    Roberto C. Sotero, Jose M. Sanchez-Bornot, and Iman Shaharabi-Farahani. “Parameter Estimation in Brain Dynamics Models from Resting-State fMRI Data using Physics-Informed Neural Networks”. In:2024 8th International Conference on Biomedical Engineering and Applications (ICBEA). 2024, pp. 73–79

    2024

  16. [16]

    Estimating the Excitatory-Inhibitory Balance from Electrocorticography Data using Physics-Informed Neural Networks

    Roberto Carlos Sotero and Jose Sanchez-Bornot. “Estimating the Excitatory-Inhibitory Balance from Electrocorticography Data using Physics-Informed Neural Networks”. In: New York, NY , USA: Asso- ciation for Computing Machinery, 2025, 113–118.ISBN: 9798400717536.URL:https://doi.org /10.1145/3700666.3700701

    work page doi:10.1145/3700666.3700701 2025

  17. [17]

    Thuerey et al.Physics-based Deep Learning

    N. Thuerey et al.Physics-based Deep Learning. 2025. arXiv:2109.05237.URL:https://arxi v.org/abs/2109.05237

    arXiv 2025

  18. [18]

    A review and perspective on hybrid modeling methodologies

    Artur M. Schweidtmann, Dongda Zhang, and Moritz von Stosch. “A review and perspective on hybrid modeling methodologies”. In:Digital Chemical Engineering10 (2024), p. 100136.ISSN: 2772-5081. URL:https://www.sciencedirect.com/science/article/pii/S2772508123000 546

    2024

  19. [19]

    Amig ´o.Differentiable programming and its applications to dynamical systems

    Adri ´an Hern´andez and Jos´e M. Amig ´o.Differentiable programming and its applications to dynamical systems. 2020. arXiv:1912.08168.URL:https://arxiv.org/abs/1912.08168

    arXiv 2020

  20. [20]

    Facundo Sapienza et al.Differentiable Programming for Differential Equations: A Review. 2024. arXiv: 2406.09699.URL:https://arxiv.org/abs/2406.09699

    arXiv 2024

  21. [21]

    Bharath Ramsundar, Dilip Krishnamurthy, and Venkatasubramanian Viswanathan.Differentiable Physics: A Position Piece. 2021. arXiv:2109.07573.URL:https://arxiv.org/abs/2109.07573

    arXiv 2021

  22. [22]

    Generative Models of Brain Dynamics

    Mahta Ramezanian-Panahi et al. “Generative Models of Brain Dynamics”. In:Frontiers in Artificial Intelligence5 (2022).ISSN: 2624-8212.URL:https://www.frontiersin.org/journals/a rtificial-intelligence/articles/10.3389/frai.2022.807406

    work page doi:10.3389/frai.2022.807406 2022

  23. [23]

    First-Principles, Data-Based, and Hybrid Modeling and Optimization of an Industrial Hydrocracking Unit

    N. Bhutani, G. P. Rangaiah, and A. K. Ray. “First-Principles, Data-Based, and Hybrid Modeling and Optimization of an Industrial Hydrocracking Unit”. In:Industrial & Engineering Chemistry Research 45.23 (2006), pp. 7807–7816.URL:https://doi.org/10.1021/ie060247q

    work page doi:10.1021/ie060247q 2006

  24. [24]

    Hybrid model of a physics-based model and machine learning for real-time estimation of unmeasurable parts: Mapping from measurable to unmeasurable variables

    Tatsuya Kaneko et al. “Hybrid model of a physics-based model and machine learning for real-time estimation of unmeasurable parts: Mapping from measurable to unmeasurable variables”. In:Ocean Engineering261 (2022), p. 112123.ISSN: 0029-8018.URL:https://www.sciencedirect.co m/science/article/pii/S0029801822014342

    2022

  25. [25]

    2022.URL: https://openreview.net/forum?id=famc03Gg231

    Philipp Holl, Nils Thuerey, and Vladlen Koltun.Physical Gradients for Deep Learning. 2022.URL: https://openreview.net/forum?id=famc03Gg231

    2022

  26. [26]

    Integrating physics-based modeling with machine learning: A survey

    Jared Willard et al. “Integrating physics-based modeling with machine learning: A survey”. In:arXiv preprint arXiv:2003.049191.1 (2020), pp. 1–34.URL:https://arxiv.org/abs/2003.0491 9v4

    arXiv 2003

  27. [27]

    Society for Industrial and Applied Mathematics, 2011

    Uwe Naumann.The Art of Differentiating Computer Programs. Society for Industrial and Applied Mathematics, 2011. eprint:https://epubs.siam.org/doi/pdf/10.1137/1.978161197 2078.URL:https://epubs.siam.org/doi/abs/10.1137/1.9781611972078

    work page doi:10.1137/1.978161197 2011

  28. [28]

    Mathieu Blondel and Vincent Roulet.The Elements of Differentiable Programming. 2025. arXiv:240 3.14606.URL:https://arxiv.org/abs/2403.14606

    arXiv 2025

  29. [29]

    Automatic Differentiation in Machine Learning: a Survey

    Atilim Gunes Baydin et al. “Automatic Differentiation in Machine Learning: a Survey”. In:Journal of Machine Learning Research18.153 (2018), pp. 1–43.URL:http://jmlr.org/papers/v 18/17-468.html

    2018

  30. [30]

    Andreas Griewank and Andrea Walther.Evaluating Derivatives. Second. Society for Industrial and Applied Mathematics, 2008. eprint:https://epubs.siam.org/doi/pdf/10.1137/1. 9780898717761.URL:https://epubs.siam.org/doi/abs/10.1137/1.978089871 7761

    work page doi:10.1137/1 2008

  31. [31]

    Stanford University, 2024.URL:https://purl.stanford.edu/pw437cb2999

    Rastko Ciric.Differentiable Programming for Mapping the Function of the Human Brain. Stanford University, 2024.URL:https://purl.stanford.edu/pw437cb2999. 21

    2024

  32. [32]

    Understanding the Mechanical Properties of Pituitary Adenomas for Optimized Surgery

    Umesh Gautam et al. “Understanding the Mechanical Properties of Pituitary Adenomas for Optimized Surgery”. In:Journal of Biomedical Materials Research Part A113.6 (2025), e37940

    2025

  33. [33]

    Biomechanical Characterization of Epileptic Brain for Surgical Planning of In- tractable Epilepsy

    Umesh Gautam et al. “Biomechanical Characterization of Epileptic Brain for Surgical Planning of In- tractable Epilepsy”. In:Journal of Biomedical Materials Research Part A114.1 (2026), e70017

    2026

  34. [34]

    Mechanical characterization and standardization of silicon scalp and dura surrogates for neurosurgical simulation

    Ramandeep Singh et al. “Mechanical characterization and standardization of silicon scalp and dura surrogates for neurosurgical simulation”. In:World Neurosurgery169 (2023), e197–e205

    2023

  35. [35]

    Material parameter identification for brain tissue using open-source platforms-GIBBON and FEBio

    Abhilash Awasthi, Umesh Gautam, and Sitikantha Roy. “Material parameter identification for brain tissue using open-source platforms-GIBBON and FEBio”. In:Recent Advances in Applied Mechanics: Proceedings of Virtual Seminar on Applied Mechanics (VSAM 2021). Springer. 2022, pp. 495–502

    2021

  36. [36]

    MatNLI: An open-source MATLAB-based solver for the non-linear inversion in elastography

    Abhilash Awasthi et al. “MatNLI: An open-source MATLAB-based solver for the non-linear inversion in elastography”. In:Advances in Engineering Software181 (2023), p. 103476

    2023

  37. [37]

    A Comparison of Gradient-Based Optimization Techniques in Magnetic Res- onance Elastography

    Abhilash Awasthi et al. “A Comparison of Gradient-Based Optimization Techniques in Magnetic Res- onance Elastography”. In:Asian Conference on Mechanics of Functional Materials and Structures. Springer. 2022, pp. 371–381

    2022

  38. [38]

    Tikhonov and Vasily Y

    Andrey N. Tikhonov and Vasily Y . Arsenin.Solutions of Ill-Posed Problems. Wiley, 1977

    1977

  39. [39]

    Diego Renner and Georgios Kissas.Accelerated Patient-Specific Calibration via Differentiable Hemo- dynamics Simulations. 2024. arXiv:2412.14572.URL:https://arxiv.org/abs/2412. 14572

    arXiv 2024

  40. [40]

    Inverse identification of region-specific hyperelastic material parameters for hu- man brain tissue

    Jan Hinrichsen et al. “Inverse identification of region-specific hyperelastic material parameters for hu- man brain tissue”. In:Biomechanics and Modeling in Mechanobiology22.5 (2023), pp. 1729–1749. URL:https://doi.org/10.1007/s10237-023-01739-w

    work page doi:10.1007/s10237-023-01739-w 2023

  41. [41]

    Rhys Newbury et al.A Review of Differentiable Simulators. 2024. arXiv:2407.05560.URL:https ://arxiv.org/abs/2407.05560

    arXiv 2024

  42. [42]

    Ado Farsi, Nacime Bouziani, and David A Ham.Missing Physics Discovery through Fully Differen- tiable Finite Element-Based Machine Learning. 2025. arXiv:2507.15787.URL:https://arxiv .org/abs/2507.15787

    arXiv 2025

  43. [43]

    Automatic Differentiation and Applications

    Hans-Christoph Fischer. “Automatic Differentiation and Applications”. In:Scientific Computing with Automatic Result Verification. Ed. by E. Adams and U. Kulisch. V ol. 189. Mathematics in Science and Engineering. Elsevier, 1993, pp. 105–142.URL:https://www.sciencedirect.com/scienc e/article/pii/S0076539208628441

    1993

  44. [44]

    Yu-Hsueh Fang et al.A Step-by-step Introduction to the Implementation of Automatic Differentiation

  45. [45]

    arXiv:2402.16020 [cs.LG].URL:https://arxiv.org/abs/2402.16020

    arXiv

  46. [46]

    Learning representations by back- propagating errors

    David E. Rumelhart, Geoffrey E. Hinton, and Ronald J. Williams. “Learning representations by back- propagating errors”. In:Nature323.6088 (1986), pp. 533–536.URL:https://www.nature.com /articles/323533a0

    1986

  47. [47]

    John Wiley & Sons, 1994

    Paul John Werbos.The roots of backpropagation: from ordered derivatives to neural networks and political forecasting. John Wiley & Sons, 1994

    1994

  48. [48]

    A comparison of the continuous and discrete adjoint approach extended based on the standard lattice Boltzmann method in flow field inverse optimization problems

    Mohamad Hamed Hekmat and Masoud Mirzaei. “A comparison of the continuous and discrete adjoint approach extended based on the standard lattice Boltzmann method in flow field inverse optimization problems”. In:Acta Mechanica227.4 (2016), pp. 1025–1050.URL:https : / / doi . org / 10 . 1007/s00707-015-1509-x

    2016

  49. [49]

    Multi-Directional Dynamic Model for Traumatic Brain Injury Detection

    Kaveh Laksari et al. “Multi-Directional Dynamic Model for Traumatic Brain Injury Detection”. In: Journal of Neurotrauma37.7 (2020). PMID: 31856650, pp. 982–993.URL:https : / / doi . org /10.1089/neu.2018.6340

    work page doi:10.1089/neu.2018.6340 2020

  50. [50]

    Finite Element Algorithms for Computational Biomechanics of the Brain

    Adam Wittek, Grand Roman Joldes, and Karol Miller. “Finite Element Algorithms for Computational Biomechanics of the Brain”. In: ed. by Karol Miller. Cham: Springer International Publishing, 2019, pp. 243–272.URL:https://doi.org/10.1007/978-3-030-04996-6_10. 22

    work page doi:10.1007/978-3-030-04996-6_10 2019

  51. [51]

    The mechanical behaviour of brain tissue: Large strain response and constitutive modelling

    M. Hrapko et al. “The mechanical behaviour of brain tissue: Large strain response and constitutive modelling”. In:Biorheology43.5 (2006), pp. 623–636.URL:https : / / doi . org / 10 . 1177 / 0006355X2006043005004

    2006

  52. [52]

    Constitutive Modeling of Brain Tissue: Current Perspectives

    Rijk de Rooij and Ellen Kuhl. “Constitutive Modeling of Brain Tissue: Current Perspectives”. In:Ap- plied Mechanics Reviews68.1 (Jan. 2016), p. 010801.ISSN: 0003-6900.URL:https://doi.org /10.1115/1.4032436

    work page doi:10.1115/1.4032436 2016

  53. [53]

    Christopher Rackauckas et al.Universal Differential Equations for Scientific Machine Learning. 2021. arXiv:2001.04385.URL:https://arxiv.org/abs/2001.04385

    Pith/arXiv arXiv 2021

  54. [54]

    A hybrid neural network-first principles approach to process modeling

    Dimitris C. Psichogios and Lyle H. Ungar. “A hybrid neural network-first principles approach to process modeling”. In:AIChE Journal38.10 (1992), pp. 1499–1511.URL:https://aiche.onlinelibr ary.wiley.com/doi/abs/10.1002/aic.690381003

    work page doi:10.1002/aic.690381003 1992

  55. [55]

    Integrating Neural Networks with First Principles Models for Dynamic Modeling

    Hong-Te Su et al. “Integrating Neural Networks with First Principles Models for Dynamic Modeling”. In:IFAC Proceedings Volumes25.5 (1992). 3rd IFAC Symposium on Dynamics and Control of Chemi- cal Reactors, Distillation Columns and Batch Processes (DYCORD+ ’92), Maryland, USA, 26-29 April, pp. 327–332.ISSN: 1474-6670.URL:https://www.sciencedirect.com/scien...

    1992

  56. [56]

    Universal Differential Equations Applied to Bioprocesses

    Jyler Menard. “Universal Differential Equations Applied to Bioprocesses”. PhD thesis. 2023.URL: https://uwaterloo.ca/computational-mathematics/sites/default/files/u ploads/documents/menard_jyler_0.pdf

    2023

  57. [57]

    Universal Differential Equations as a Common Modeling Language for Neuroscience

    Ahmed ElGazzar and Marcel van Gerven. “Universal Differential Equations as a Common Modeling Language for Neuroscience”. In:arXiv preprint arXiv:2403.14510(2024).URL:https://arxiv.o rg/abs/2403.14510

    arXiv 2024

  58. [58]

    Current state and open problems in universal dif- ferential equations for systems biology

    Maren Philipps, Nina Schmid, and Jan Hasenauer. “Current state and open problems in universal dif- ferential equations for systems biology”. In:npj Systems Biology and Applications11.1 (2025), p. 101. URL:https://doi.org/10.1038/s41540-025-00550-w

    work page doi:10.1038/s41540-025-00550-w 2025

  59. [59]

    Combining first principles modelling and artificial neural networks: a general framework

    R. Oliveira. “Combining first principles modelling and artificial neural networks: a general framework”. eng. In:Computers & chemical engineering28.5 (2004), pp. 755–766.ISSN: 0098-1354

    2004

  60. [60]

    Incorporating Unmodeled Dynamics Into First-Principles Models Through Machine Learning

    Ward Quaghebeur, Ingmar Nopens, and Bernard De Baets. “Incorporating Unmodeled Dynamics Into First-Principles Models Through Machine Learning”. In:IEEE Access9 (2021), pp. 22014–22022

    2021

  61. [61]

    Hybrid modeling of first-principles and machine learning: A step-by-step tutorial review for practical implementation

    Parth Shah et al. “Hybrid modeling of first-principles and machine learning: A step-by-step tutorial review for practical implementation”. In:Computers & Chemical Engineering194 (2025), p. 108926. ISSN: 0098-1354.URL:https : / / www . sciencedirect . com / science / article / pii /S0098135424003442

    2025

  62. [62]

    Integration of machine learning and first principles models

    Lokesh Rajulapati et al. “Integration of machine learning and first principles models”. In:AIChE Jour- nal68.6 (2022), e17715.URL:https://aiche.onlinelibrary.wiley.com/doi/abs /10.1002/aic.17715

    work page doi:10.1002/aic.17715 2022

  63. [63]

    AI-enhanced iterative solvers for accelerating the solution of large-scale parametrized systems

    Stefanos Nikolopoulos et al. “AI-enhanced iterative solvers for accelerating the solution of large-scale parametrized systems”. In:International Journal for Numerical Methods in Engineering125.2 (2024), e7372. eprint:https://onlinelibrary.wiley.com/doi/pdf/10.1002/nme.7372. URL:https://onlinelibrary.wiley.com/doi/abs/10.1002/nme.7372

    work page doi:10.1002/nme.7372 2024

  64. [64]

    Neural Ordinary Differential Equations

    Ricky T. Q. Chen et al. “Neural Ordinary Differential Equations”. In:Advances in Neural Information Processing Systems. Ed. by S. Bengio et al. V ol. 31. Curran Associates, Inc., 2018.URL:https : //proceedings.neurips.cc/paper_files/paper/2018/file/69386f6bb1dfed 68692a24c8686939b9-Paper.pdf

    2018

  65. [65]

    YongKyung Oh et al.Comprehensive Review of Neural Differential Equations for Time Series Analysis

  66. [66]

    arXiv:2502.09885.URL:https://arxiv.org/abs/2502.09885

    arXiv

  67. [67]

    Deep Implicit Layers - Neural ODEs, Deep Equili- birum Models, and Beyond

    Zico Kolter, David Duvenaud, and Matt Johnson. “Deep Implicit Layers - Neural ODEs, Deep Equili- birum Models, and Beyond”. NeurIPS 2020 tutorial.URL:https://implicit-layers-tutor ial.org/. 23

    2020

  68. [68]

    Belinda Tzen and Maxim Raginsky.Neural Stochastic Differential Equations: Deep Latent Gaussian Models in the Diffusion Limit. 2019. arXiv:1905 . 09883.URL:https : / / arxiv . org / abs /1905.09883

    arXiv 2019

  69. [69]

    Augmented Neural ODEs

    Emilien Dupont, Arnaud Doucet, and Yee Whye Teh. “Augmented Neural ODEs”. In:Advances in Neural Information Processing Systems32 (2019). Ed. by H. Wallach et al.URL:https://proce edings.neurips.cc/paper_files/paper/2019/file/21be9a4bd4f81549a9d1d 241981cec3c-Paper.pdf

    2019

  70. [70]

    Neural Controlled Differential Equations for Irregular Time Series

    Patrick Kidger et al. “Neural Controlled Differential Equations for Irregular Time Series”. In:Advances in Neural Information Processing Systems33 (2020). Ed. by H. Larochelle et al., pp. 6696–6707.URL: https://proceedings.neurips.cc/paper_files/paper/2020/file/4a5876b 450b45371f6cfe5047ac8cd45-Paper.pdf

    2020

  71. [71]

    Lagrangian Neural Networks

    Miles Cranmer et al. “Lagrangian Neural Networks”. In:ICLR 2020 Workshop on Integration of Deep Neural Models and Differential Equations. 2019.URL:https://openreview.net/forum?id =iE8tFa4Nq

    2020

  72. [72]

    Hamiltonian Neural Networks

    Samuel Greydanus, Misko Dzamba, and Jason Yosinski. “Hamiltonian Neural Networks”. In:Advances in Neural Information Processing Systems32 (2019). Ed. by H. Wallach et al.URL:https://pro ceedings.neurips.cc/paper_files/paper/2019/file/26cd8ecadce0d4efd6cc 8a8725cbd1f8-Paper.pdf

    2019

  73. [73]

    Generalized Framework for Liquid Neural Network upon Sequential and Non-Sequential Tasks

    Prakash Kumar Karn, Iman Ardekani, and Waleed H. Abdulla. “Generalized Framework for Liquid Neural Network upon Sequential and Non-Sequential Tasks”. In:Mathematics12.16 (2024).ISSN: 2227-7390.URL:https://www.mdpi.com/2227-7390/12/16/2525

    2024

  74. [74]

    On Second Order Behaviour in Augmented Neural ODEs

    Alexander Norcliffe et al. “On Second Order Behaviour in Augmented Neural ODEs”. In:Advances in Neural Information Processing Systems33 (2020). Ed. by H. Larochelle et al., pp. 5911–5921.URL: https://proceedings.neurips.cc/paper_files/paper/2020/file/418db2ea 5d227a9ea8db8e5357ca2084-Paper.pdf

    2020

  75. [75]

    Neural differential equations for learning to program neural nets through continuous learning rules

    Kazuki Irie, Francesco Faccio, and J ¨urgen Schmidhuber. “Neural differential equations for learning to program neural nets through continuous learning rules”. In:Advances in Neural Information Processing Systems35 (2022), pp. 38614–38628.URL:https://papers.nips.cc/paper_files/pape r/2022/file/fc09b26b85ab3abb2832bd555a2e4215-Paper-Conference.pdf

    2022

  76. [76]

    nmODE: neural memory ordinary differential equation

    Zhang Yi. “nmODE: neural memory ordinary differential equation”. In:Artificial Intelligence Review 56.12 (2023), pp. 14403–14438

    2023

  77. [77]

    Jonas Degrave et al.A Differentiable Physics Engine for Deep Learning in Robotics. 2018. arXiv: 1611.01652.URL:https://arxiv.org/abs/1611.01652

    Pith/arXiv arXiv 2018

  78. [78]

    Differentiable Physics Simulation

    Junbang Liang and Ming C. Lin. “Differentiable Physics Simulation”. In:ICLR 2020 Workshop on Integration of Deep Neural Models and Differential Equations. 2019.URL:https://openreview .net/forum?id=p-SG2KFY2

    2020

  79. [79]

    Learning to Control PDEs with Differentiable Physics

    Philipp Holl, Nils Thuerey, and Vladlen Koltun. “Learning to Control PDEs with Differentiable Physics”. In:International Conference on Learning Representations. 2020.URL:https://openreview.ne t/forum?id=HyeSin4FPB

    2020

  80. [80]

    Scalable Differentiable Physics for Learning and Control

    Yi-Ling Qiao et al. “Scalable Differentiable Physics for Learning and Control”. In:Proceedings of the 37th International Conference on Machine Learning. Ed. by Hal Daum´e III and Aarti Singh. V ol. 119. Proceedings of Machine Learning Research. PMLR, 2020, pp. 7847–7856.URL:https://procee dings.mlr.press/v119/qiao20a.html

    2020

Showing first 80 references.