pith. machine review for the scientific record. sign in

arxiv: 2603.20410 · v2 · submitted 2026-03-20 · 💻 cs.LG

Recognition: 2 theorem links

· Lean Theorem

SLE-FNO: Single-Layer Extensions for Task-Agnostic Continual Learning in Fourier Neural Operators

Mahmoud Elhadidy , Roshan M. D'Souza , Amirhossein Arzani
Authors on Pith no claims yet

Pith reviewed 2026-05-15 08:02 UTC · model grok-4.3

classification 💻 cs.LG
keywords continual learningFourier Neural Operatorsfluid dynamicssurrogate modelscatastrophic forgettingblood flowsingle-layer extensiontask-agnostic adaptation
0
0 comments X

The pith

Single-layer extensions added to Fourier Neural Operators enable continual learning across shifting fluid tasks with zero forgetting and few new parameters.

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

The paper develops SLE-FNO to let surrogate models for fluid flows update themselves when new conditions appear without needing old training data. It tests the approach on a sequence of four out-of-distribution blood-flow simulation tasks that map concentration fields to wall shear stress. The single-layer additions allow the base Fourier Neural Operator to acquire new task knowledge while preserving accuracy on earlier tasks. In direct comparisons with replay, regularization, and other architecture-based methods, SLE-FNO shows the best combination of adaptation speed and retention. This setup addresses the practical problem that real fluid simulations frequently encounter changing geometries or regimes after initial training.

Core claim

SLE-FNO achieves accurate predictions on new out-of-distribution fluid tasks while producing zero forgetting on prior tasks and adding only minimal parameters, delivering a stronger plasticity-stability balance than EWC, LwF, replay buffers, OGD, GEM, PiggyBack, or LoRA in the four-task blood-flow sequence.

What carries the argument

The Single-Layer Extension (SLE) that appends lightweight task-specific layers to a frozen Fourier Neural Operator backbone, enabling task-agnostic updates without full retraining or data replay.

If this is right

  • Surrogate models for pulsatile flows can be updated sequentially as new experimental conditions arise without storing prior simulation data.
  • Computational cost for adapting to new geometries stays low because only a small number of extra parameters are introduced per task.
  • Zero-forgetting performance holds across the tested sequence of distribution shifts in aneurysmal blood flow.
  • Architecture-based continual learning outperforms or matches replay and regularization baselines in this spatial regression setting.

Where Pith is reading between the lines

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

  • The same single-layer pattern could be tested on other neural operator families for time-dependent physics problems beyond blood flow.
  • If single-layer capacity proves insufficient on wider task sequences, hybrid combinations with light replay buffers might become necessary.
  • The approach implies that many scientific surrogate models could be maintained as living models rather than retrained from scratch when conditions change.

Load-bearing premise

Single-layer additions alone can capture all required adaptations for out-of-distribution fluid tasks without harming earlier performance.

What would settle it

Run SLE-FNO on a fifth blood-flow task whose geometry or regime lies further outside the training distribution and check whether accuracy on the original four tasks remains at the reported level with no measurable drop.

Figures

Figures reproduced from arXiv: 2603.20410 by Amirhossein Arzani, Mahmoud Elhadidy, Roshan M. D'Souza.

Figure 1
Figure 1. Figure 1: (a) Geometry sketch with four spatial parameters [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall Fourier neural operator (FNO) architecture. (a) FNO lifts the input to a [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Regularization-based: (a) Elastic Weight Consolidation (EWC) constrains updates us [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Single-Layer Extension with Fourier Neural Operator (SLE-FNO). (a) During training, [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: TAWSS results after the first stage (Training on A) for two randomly selected samples [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: TAWSS results at the second stage (Fine-Tuning at task B). (a) Results for the worst [PITH_FULL_IMAGE:figures/full_fig_p024_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: TAWSS results at the third stage (Fine-Tuning on C). (a) Worst-case sample from task [PITH_FULL_IMAGE:figures/full_fig_p025_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: TAWSS results at the third stage (Fine-Tuning on C). A random sample from task A is [PITH_FULL_IMAGE:figures/full_fig_p026_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: TAWSS results at the fourth stage (Fine-Tuning on task D). (a) Worst-case sample from [PITH_FULL_IMAGE:figures/full_fig_p027_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: TAWSS results at the fourth stage (Fine-Tuning on task D). (a) Worst-case sample from [PITH_FULL_IMAGE:figures/full_fig_p028_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Task A routing accuracy (%) shown as the mean (solid line) with a shaded region indicating ±1 standard deviation, plotted as a function of the number of retained KPCA modes. The x-axis represents the number of retained modes, while the y-axis reports the routing accuracy on the Task A test set. We examine the number of KPCA modes required for accurate task identification. Since RFF are used to lift the da… view at source ↗
read the original abstract

Scientific machine learning is increasingly used to build surrogate models, yet most models are trained under a restrictive assumption in which future data follow the same distribution as the training set. In practice, new experimental conditions or simulation regimes may differ significantly, requiring extrapolation and model updates without re-access to prior data. This creates a need for continual learning (CL) frameworks that can adapt to distribution shifts while preventing catastrophic forgetting. Such challenges are pronounced in fluid dynamics, where changes in geometry, boundary conditions, or flow regimes induce non-trivial changes to the solution. Here, we introduce a new architecture-based approach (SLE-FNO) combining a Single-Layer Extension (SLE) with the Fourier Neural Operator (FNO) to support efficient CL. SLE-FNO was compared with a range of established CL methods, including Elastic Weight Consolidation (EWC), Learning without Forgetting (LwF), replay-based approaches, Orthogonal Gradient Descent (OGD), Gradient Episodic Memory (GEM), PiggyBack, and Low-Rank Adaptation (LoRA), within a spatial field-to-field regression setting. The models were trained to map transient concentration fields to time-averaged wall shear stress (TAWSS) in pulsatile aneurysmal blood flow. Tasks were derived from 230 computational fluid dynamics simulations grouped into four sequential and out-of-distribution configurations. Results show that replay-based methods and architecture-based approaches (PiggyBack, LoRA, and SLE-FNO) achieve the best retention, with SLE-FNO providing the strongest overall balance between plasticity and stability, achieving accuracy with zero forgetting and minimal additional parameters. Our findings highlight key differences between CL algorithms and introduce SLE-FNO as a promising strategy for adapting baseline models when extrapolation is required.

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 manuscript introduces SLE-FNO, an architecture-based continual learning method that augments the Fourier Neural Operator with a Single-Layer Extension (SLE) to enable adaptation to distribution shifts without catastrophic forgetting. It evaluates SLE-FNO against baselines including EWC, LwF, GEM, OGD, PiggyBack, and LoRA on a spatial field-to-field regression task: mapping transient concentration fields to time-averaged wall shear stress (TAWSS) using data from 230 CFD simulations grouped into four sequential out-of-distribution configurations derived from aneurysmal blood flow. The central claim is that SLE-FNO achieves accuracy with zero forgetting and minimal added parameters while providing the strongest overall balance between plasticity and stability.

Significance. If the results hold under broader testing, the work would be significant for providing an efficient, task-agnostic extension mechanism for FNOs in scientific machine learning applications involving non-stationary data, such as varying geometries or flow regimes in fluid dynamics. It could support the development of adaptive surrogate models that extrapolate without replay or heavy regularization, and the empirical comparison highlights practical differences among CL algorithms in this domain.

major comments (2)
  1. [Abstract] Abstract: The headline claims of 'accuracy with zero forgetting' and 'strongest overall balance' are presented without any quantitative metrics, error bars, statistical tests, or details on the magnitude of improvements over baselines, which is load-bearing for verifying the central performance assertions.
  2. [Abstract] Abstract (results paragraph): The evaluation is restricted to one fixed sequential ordering of four out-of-distribution tasks from the 230 CFD runs; the absence of ablations on task permutation, scaling beyond four tasks, or transfer to different geometries/Reynolds regimes undermines the claim that SLE-FNO is robustly task-agnostic.
minor comments (1)
  1. [Abstract] Abstract: The experimental setup description omits specifics on the four task configurations (e.g., exact changes in geometry or boundary conditions) and the precise implementation details for the baseline methods.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and detailed review. The comments highlight important aspects of how the abstract presents our results and the scope of our evaluation. We address each point below and outline revisions that strengthen the manuscript without overstating our contributions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The headline claims of 'accuracy with zero forgetting' and 'strongest overall balance' are presented without any quantitative metrics, error bars, statistical tests, or details on the magnitude of improvements over baselines, which is load-bearing for verifying the central performance assertions.

    Authors: We agree that the abstract would be strengthened by including concrete quantitative support for these claims. In the revised version, we will incorporate specific metrics drawn from our experiments, such as the relative L2 errors on each task (e.g., SLE-FNO maintains <1% error with zero forgetting while baselines show 5-15% degradation), the number of additional parameters (under 2% of the base FNO), and standard deviations across repeated runs. We will also briefly note the magnitude of improvement over the strongest baselines (PiggyBack and LoRA) to make the performance assertions directly verifiable from the abstract. revision: yes

  2. Referee: [Abstract] Abstract (results paragraph): The evaluation is restricted to one fixed sequential ordering of four out-of-distribution tasks from the 230 CFD runs; the absence of ablations on task permutation, scaling beyond four tasks, or transfer to different geometries/Reynolds regimes undermines the claim that SLE-FNO is robustly task-agnostic.

    Authors: The referee correctly identifies that our experiments use a single task ordering. This ordering was selected to emulate a realistic progression of distribution shifts in aneurysmal flow modeling. Because SLE-FNO is purely architecture-based and adds task-agnostic single-layer extensions without relying on task identity, replay buffers, or regularization that depends on ordering, the method itself does not encode assumptions about sequence. Nevertheless, we did not perform permutation ablations or scale to more than four tasks owing to the substantial cost of generating additional high-fidelity CFD data. We will revise the abstract to replace 'robustly task-agnostic' with 'demonstrates strong task-agnostic adaptation in the evaluated setting' and will add a dedicated limitations paragraph in the discussion that explicitly acknowledges the restricted experimental scope while outlining directions for future validation on varied geometries and Reynolds numbers. revision: partial

Circularity Check

0 steps flagged

No circularity detected; purely empirical comparison with no derivations

full rationale

The manuscript introduces SLE-FNO as an architectural extension to FNO and reports empirical performance on a fixed sequence of four out-of-distribution tasks derived from 230 CFD runs. No mathematical derivations, uniqueness theorems, or first-principles predictions are claimed; results consist of accuracy, forgetting, and parameter-count metrics obtained by training and evaluating the models on the described data. Consequently there are no steps that reduce by construction to fitted inputs, self-citations, or renamed ansatzes. The work is self-contained as an experimental benchmark.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Based on abstract only, the central claim rests on the empirical success of the SLE mechanism in the described fluid dynamics tasks; no explicit free parameters, axioms, or invented entities are detailed beyond the introduction of the SLE component itself.

invented entities (1)
  • Single-Layer Extension (SLE) no independent evidence
    purpose: Enable task-specific adaptation in FNO for continual learning without catastrophic forgetting
    Core innovation introduced to support efficient CL in the architecture-based approach.

pith-pipeline@v0.9.0 · 5631 in / 1144 out tokens · 44680 ms · 2026-05-15T08:02:38.983643+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Replay-Based Continual Learning for Physics-Informed Neural Operators

    cs.LG 2026-05 unverdicted novelty 4.0

    A replay-based continual learning strategy for physics-informed neural operators mitigates catastrophic forgetting on prior physical problems while enabling efficient adaptation to new data using only physical constraints.

Reference graph

Works this paper leans on

121 extracted references · 121 canonical work pages · cited by 1 Pith paper · 9 internal anchors

  1. [1]

    Thiyagalingam, M

    J. Thiyagalingam, M. Shankar, G. Fox, and T. Hey. Scientific machine learning benchmarks. Nature Reviews Physics, 4(6):413–420, 2022

    work page 2022

  2. [2]

    G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, and L. Yang. Physics- informed machine learning.Nature Reviews Physics, 3(6):422–440, 2021

    work page 2021

  3. [3]

    S. L. Brunton and J. N. Kutz.Data-driven science and engineering: Machine learning, dynamical systems, and control. Cambridge University Press, 2022

    work page 2022

  4. [4]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anand- kumar. Fourier neural operator for parametric partial differential equations.arXiv preprint arXiv:2010.08895, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  5. [5]

    L. Lu, P. Jin, G. Pang, Z. Zhang, and G. E. Karniadakis. Learning nonlinear operators via deeponet based on the universal approximation theorem of operators.Nature Machine Intelligence, 3(3):218–229, 2021

    work page 2021

  6. [6]

    Kovachki, Z

    N. Kovachki, Z. Li, B. Liu, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, and A. Anand- kumar. Neural operator: Learning maps between function spaces with applications to PDEs. Journal of Machine Learning Research, 24(89):1–97, 2023

    work page 2023

  7. [7]

    Raissi, P

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

    work page 2019

  8. [8]

    Goswami, A

    S. Goswami, A. Bora, Y. Yu, and G. E. Karniadakis. Physics-informed deep neural operator networks. InMachine Learning in Modeling and Simulation: Methods and Applications, pages 219–254. Springer, 2023. 45

    work page 2023

  9. [9]

    S. Wang, Y. Teng, and P. Perdikaris. Understanding and mitigating gradient flow pathologies in physics-informed neural networks.SIAM Journal on Scientific Computing, 43(5):A3055– A3081, 2021

    work page 2021

  10. [10]

    Kiyani, K

    E. Kiyani, K. Shukla, J. F. Urbán, J. Darbon, and G. E. Karniadakis. Which optimizer works best for physics-informed neural networks and Kolmogorov-Arnold networks?arXiv preprint arXiv:2501.16371, 2025

    work page arXiv 2025

  11. [11]

    Y. Choi, S. W. Cheung, Y. Kim, P. Tsai, A. N. Diaz, I. Zanardi, S. W. Chung, D. M. Copeland, C. Kendrick, W. Anderson, et al. Defining foundation models for computational science: A call for clarity and rigor.arXiv preprint arXiv:2505.22904, 2025

    work page arXiv 2025

  12. [12]

    S. S. Menon, T. Mondal, S. Brahmachary, A. Panda, S. M. Joshi, K. Kalyanaraman, and A. D. Jagtap. On scientific foundation models: Rigorous definitions, key applications, and a survey.SSRN, 2025

    work page 2025

  13. [13]

    Quiñonero-Candela, M

    J. Quiñonero-Candela, M. Sugiyama, A. Schwaighofer, and N. D. Lawrence.Dataset shift in machine learning. MIT Press, 2008

    work page 2008

  14. [14]

    A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks

    D. Hendrycks and K. Gimpel. A baseline for detecting misclassified and out-of-distribution examples in neural networks.arXiv preprint arXiv:1610.02136, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  15. [15]

    J. Yang, K. Zhou, Y. Li, and Z. Liu. Generalized out-of-distribution detection: A survey. International Journal of Computer Vision, 132(12):5635–5662, 2024

    work page 2024

  16. [16]

    L. Yuan, H. S. Park, and E. Lejeune. Towards out of distribution generalization for problems in mechanics.Computer Methods in Applied Mechanics and Engineering, 400:115569, 2022

    work page 2022

  17. [17]

    Arzani, L

    A. Arzani, L. Yuan, P. Newell, and B. Wang. Interpreting and generalizing deep learn- ing in physics-based problems with functional linear models.Engineering with Computers, 41(1):135–157, 2025

    work page 2025

  18. [18]

    Y. M. K. W. Church, Z. Chen. Emerging trends: A gentle introduction to fine-tuning.Natural Language Engineering, 27(6):763–778, 2021

    work page 2021

  19. [19]

    R. M. French. Catastrophic forgetting in connectionist networks.Trends in Cognitive Sci- ences, 3(4):128–135, 1999

    work page 1999

  20. [20]

    J. Gama, I. Žliobait˙ e, A. Bifet, M. Pechenizkiy, and A. Bouchachia. A survey on concept drift adaptation.ACM computing surveys (CSUR), 46(4):1–37, 2014

    work page 2014

  21. [21]

    Liang, Y

    S. Liang, Y. Li, and R. Srikant. Enhancing the reliability of out-of-distribution image detec- tion in neural networks.arXiv preprint arXiv:1706.02690, 2017

    work page arXiv 2017

  22. [22]

    Y.Ovadia, E.Fertig, J.Ren, Z.Nado, D.Sculley, S.Nowozin, J.Dillon, B.Lakshminarayanan, and J. Snoek. Can you trust your model’s uncertainty? evaluating predictive uncertainty under dataset shift.Advances in Neural Information Processing Systems, 32, 2019

    work page 2019

  23. [23]

    P. W. Koh, S. Sagawa, H. Marklund, S. M. Xie, M. Zhang, A. Balsubramani, W. Hu, M. Ya- sunaga, R. L. Phillips, I. Gao, et al. Wilds: A benchmark of in-the-wild distribution shifts. InInternational Conference on Machine Learning, pages 5637–5664. PMLR, 2021

    work page 2021

  24. [24]

    Hadsell, D

    R. Hadsell, D. Rao, A. A. Rusu, and R. Pascanu. Embracing change: Continual learning in deep neural networks.Trends in Cognitive Sciences, 24(12):1028–1040, 2020. 46

    work page 2020

  25. [25]

    De Lange, R

    M. De Lange, R. Aljundi, M. Masana, S. Parisot, X. Jia, A. Leonardis, G. Slabaugh, and T. Tuytelaars. A continual learning survey: Defying forgetting in classification tasks.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7):3366–3385, 2021

    work page 2021

  26. [26]

    L. Wang, X. Zhang, H. Su, and J. Zhu. A comprehensive survey of continual learning: Theory, method and application.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(8):5362–5383, 2024

    work page 2024

  27. [27]

    G. M. Van de Ven and A. S. Tolias. Three scenarios for continual learning.arXiv preprint arXiv:1904.07734, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1904

  28. [28]

    Li and D

    Z. Li and D. Hoiem. Learning without forgetting.IEEE Transactions on Pattern Analysis and Machine Intelligence, 40(12):2935–2947, 2017

    work page 2017

  29. [29]

    Kirkpatrick, R

    J. Kirkpatrick, R. Pascanu, N. Rabinowitz, J. Veness, G. Desjardins, A. A. Rusu, K. Milan, J. Quan, T. Ramalho, A. Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks.Proceedings of the National Academy of Sciences, 114(13):3521–3526, 2017

    work page 2017

  30. [30]

    Rolnick, A

    D. Rolnick, A. Ahuja, J. Schwarz, T. Lillicrap, and G. Wayne. Experience replay for continual learning.Advances in Neural Information Processing Systems, 32, 2019

    work page 2019

  31. [31]

    H. Shin, J. K. Lee, J. Kim, and J. Kim. Continual learning with deep generative replay. Advances in Neural Information Processing Systems, 30, 2017

    work page 2017

  32. [32]

    Farajtabar, N

    M. Farajtabar, N. Azizan, A. Mott, and A. Li. Orthogonal gradient descent for continual learning.arXiv preprint arXiv:1910.07104, 2020

    work page arXiv 1910

  33. [33]

    Lopez-Paz and M

    D. Lopez-Paz and M. Ranzato. Gradient episodic memory for continual learning.Advances in Neural Information Processing Systems, 30, 2017

    work page 2017

  34. [34]

    G. Saha, I. Garg, and K. Roy. Gradient projection memory for continual learning.arXiv preprint arXiv:2103.09762, 2021

    work page arXiv 2021

  35. [35]

    Lazebnik A

    S. Lazebnik A. Mallya. Packnet: Adding multiple tasks to a single network by iterative pruning. InProceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 7765–7773, 2018

    work page 2018

  36. [36]

    Mallya, D

    A. Mallya, D. Davis, and S. Lazebnik. Piggyback: Adapting a single network to multiple tasks by learning to mask weights. InProceedings of the European Conference on Computer Vision (ECCV), pages 67–82, 2018

    work page 2018

  37. [37]

    A. A. Rusu, N. C. Rabinowitz, G. Desjardins, H. Soyer, J. Kirkpatrick, K. Kavukcuoglu, R. Pascanu, and R. Hadsell. Progressive neural networks.arXiv preprint arXiv:1606.04671, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  38. [38]

    D. Rao, F. Visin, A. Rusu, R. Pascanu, Y. W. Teh, and R. Hadsell. Continual unsupervised representation learning.Advances in Neural Information Processing Systems, 32, 2019

    work page 2019

  39. [39]

    H. Cha, J. Lee, and J. Shin. Co2l: Contrastive continual learning. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 9516–9525, 2021

    work page 2021

  40. [40]

    D. Zhou, Q. Wang, Z. Qi, H. Ye, D. Zhan, and Z. Liu. Class-incremental learning: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 47

    work page 2024

  41. [41]

    G. M. Van de Ven, T. Tuytelaars, and A. S. Tolias. Three types of incremental learning. Nature Machine Intelligence, 4(12):1185–1197, 2022

    work page 2022

  42. [42]

    He and B

    Y. He and B. Sick. CLeaR: An adaptive continual learning framework for regression tasks. AI Perspectives, 3(1):2, 2021

    work page 2021

  43. [43]

    Besnard and N

    Q. Besnard and N. Ragot. Continual learning for time series forecasting: A first survey. Engineering Proceedings, 68(1):49, 2024

    work page 2024

  44. [44]

    M. Ding, K. Ji, D. Wang, and J. Xu. Understanding forgetting in continual learning with linear regression.arXiv preprint arXiv:2405.17583, 2024

    work page arXiv 2024

  45. [45]

    Howard, Y

    A. Howard, Y. Fu, and P. Stinis. A multifidelity approach to continual learning for physical systems.Machine Learning: Science and Technology, 5(2):025042, 2024

    work page 2024

  46. [46]

    Tripura and S

    T. Tripura and S. Chakraborty. Neural combinatorial wavelet neural operator for catas- trophic forgetting free in-context operator learning of multiple partial differential equations. Computer Physics Communications, page 109882, 2025

    work page 2025

  47. [47]

    K. M. Samuel and F. Ahmed. Continual learning strategies for 3D engineering regression problems: A benchmarking study.Journal of Computing and Information Science in Engi- neering, 25(10):101003, 2025

    work page 2025

  48. [48]

    H. Kang, J. Yoon, S. J. Hwang, and C. D. Yoo. Continual learning: Forget-free winning subnetworks for video representations.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    work page 2024

  49. [49]

    Mahmoudi, A

    M. Mahmoudi, A. Farghadan, D. R. McConnell, A. J. Barker, J. J. Wentzel, M. J. Budoff, and A. Arzani. The story of wall shear stress in coronary artery atherosclerosis: biochemical transport and mechanotransduction.Journal of Biomechanical Engineering, 143(4):041002, 2021

    work page 2021

  50. [50]

    M. L. Raghavan, B. Ma, and R. E. Harbaugh. Quantified aneurysm shape and rupture risk. Journal of Neurosurgery, 102(2):355–362, 2005

    work page 2005

  51. [51]

    J. R. Womersley. Method for the calculation of velocity, rate of flow and viscous drag in arteries when the pressure gradient is known.The Journal of Physiology, 127(3):553, 1955

    work page 1955

  52. [52]

    D. N. Ku, D. P. Giddens, C. K. Zarins, and S. Glagov. Pulsatile flow and atherosclerosis in the human carotid bifurcation. positive correlation between plaque location and low oscillating shear stress.Arteriosclerosis: An Official Journal of the American Heart Association, Inc., 5(3):293–302, 1985

    work page 1985

  53. [53]

    A. M. Malek, S. L. Alper, and S. Izumo. Hemodynamic shear stress and its role in atheroscle- rosis.Jama, 282(21):2035–2042, 1999

    work page 2035

  54. [54]

    A. Arzani. Accounting for residence-time in blood rheology models: do we really need non- newtonian blood flow modelling in large arteries?Journal of The Royal Society Interface, 15(146):20180486, 2018

    work page 2018

  55. [55]

    Staarmann, M

    B. Staarmann, M. Smith, and C. J. Prestigiacomo. Shear stress and aneurysms: a review. Neurosurgical Focus, 47(1):E2, 2019. 48

    work page 2019

  56. [56]

    Ayachit.The paraview guide: a parallel visualization application

    U. Ayachit.The paraview guide: a parallel visualization application. Kitware, Inc., 2015

    work page 2015

  57. [57]

    W. M. Czarnecki, S. Osindero, M. Jaderberg, G. Swirszcz, and R. Pascanu. Sobolev training for neural networks.Advances in Neural Information Processing Systems, 30, 2017

    work page 2017

  58. [58]

    I. J. Goodfellow, M. Mirza, D. Xiao, A. Courville, and Y. Bengio. An empirical investigation of catastrophic forgetting in gradient-based neural networks.arXiv preprint arXiv:1312.6211, 2013

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  59. [59]

    Y. Luo, Z. Yang, F. Meng, Y. Li, J. Zhou, and Y. Zhang. An empirical study of catas- trophic forgetting in large language models during continual fine-tuning.arXiv preprint arXiv:2308.08747, 2023

    work page arXiv 2023

  60. [60]

    L. Y. Pratt. Discriminability-based transfer between neural networks.Advances in Neural Information Processing Systems, 5, 1992

    work page 1992

  61. [61]

    Buzzega, M

    P. Buzzega, M. Boschini, A. Porrello, D. Abati, and S. Calderara. Dark experience for general continual learning: a strong, simple baseline.Advances in Neural Information Processing Systems, 33:15920–15930, 2020

    work page 2020

  62. [62]

    Prabhu, P

    A. Prabhu, P. H. Torr, and P. K. Dokania. Gdumb: A simple approach that questions our progress in continual learning. InEuropean Conference on Computer Vision, pages 524–540. Springer, 2020

    work page 2020

  63. [63]

    Efficient Lifelong Learning with A-GEM

    A. Chaudhry, M. Ranzato, M. Rohrbach, and M. Elhoseiny. Efficient lifelong learning with A-GEM.arXiv preprint arXiv:1812.00420, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  64. [64]

    E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen, et al. LoRA: Low-rank adaptation of large language models.ICLR, 1(2):3, 2022

    work page 2022

  65. [65]

    J. O. Zhang, A. Sax, A. Zamir, L. Guibas, and J. Malik. Side-tuning: a baseline for network adaptation via additive side networks. InEuropean Conference on Computer Vision, pages 698–714. Springer, 2020

    work page 2020

  66. [66]

    Rahaman, A

    N. Rahaman, A. Baratin, D. Arpit, F. Draxler, M. Lin, F. Hamprecht, Y. Bengio, and A.Courville. On thespectralbiasof neuralnetworks. InInternational Conference on Machine Learning, pages 5301–5310. PMLR, 2019

    work page 2019

  67. [67]

    Serra, D

    J. Serra, D. Suris, M. Miron, and A. Karatzoglou. Overcoming catastrophic forgetting with hard attention to the task. InInternational Conference on Machine Learning, pages 4548–

    work page

  68. [68]

    D. Ha, A. Dai, and Q. V. Le. Hypernetworks.arXiv preprint arXiv:1609.09106, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  69. [69]

    Aljundi, P

    R. Aljundi, P. Chakravarty, and T. Tuytelaars. Expert gate: Lifelong learning with a net- work of experts. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3366–3375, 2017

    work page 2017

  70. [70]

    H. Zhu, M. Majzoubi, A. Jain, and A. Choromanska. TAME: Task agnostic continual learning using multiple experts. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4139–4148, 2024. 49

    work page 2024

  71. [71]

    G. Kim, S. Esmaeilpour, C. Xiao, and B. Liu. Continual learning based on ood detection and task masking. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3856–3866, 2022

    work page 2022

  72. [72]

    Wortsman, V

    M. Wortsman, V. Ramanujan, R. Liu, A. Kembhavi, M. Rastegari, J. Yosinski, and A. Farhadi. Supermasks in superposition.Advances in Neural Information Processing Sys- tems, 33:15173–15184, 2020

    work page 2020

  73. [73]

    Nonlinearcomponentanalysisasakerneleigenvalue problem.Neural Computation, 10(5):1299–1319, 1998

    B.Schölkopf, A.Smola, andK.R.Müller. Nonlinearcomponentanalysisasakerneleigenvalue problem.Neural Computation, 10(5):1299–1319, 1998

    work page 1998

  74. [74]

    Schölkopf, R

    B. Schölkopf, R. C. Williamson, A. Smola, J. Shawe-Taylor, and J. Platt. Support vector method for novelty detection.Advances in Neural Information Processing Systems, 12, 1999

    work page 1999

  75. [75]

    K. Lee, K. Lee, H. Lee, and J. Shin. A simple unified framework for detecting out-of- distribution samples and adversarial attacks.Advances in Neural Information Processing Systems, 31, 2018

    work page 2018

  76. [76]

    W. Liu, X. Wang, J. Owens, and Y. Li. Energy-based out-of-distribution detection.Advances in Neural Information Processing Systems, 33:21464–21475, 2020

    work page 2020

  77. [77]

    K. Fang, Q. Tao, K. Lv, M. He, X. Huang, and J. Yang. Kernel PCA for out-of-distribution detection.Advances in Neural Information Processing Systems, 37:134317–134344, 2024

    work page 2024

  78. [78]

    Rahimi and B

    A. Rahimi and B. Recht. Random features for large-scale kernel machines.Advances in Neural Information Processing Systems, 20, 2007

    work page 2007

  79. [79]

    Takamoto, T

    M. Takamoto, T. Praditia, R. Leiteritz, D. MacKinlay, F. Alesiani, D. Pflüger, and M. Niepert. Pdebench: An extensive benchmark for scientific machine learning.Advances in Neural Information Processing Systems, 35:1596–1611, 2022

    work page 2022

  80. [80]

    Semmelrock, T

    H. Semmelrock, T. Ross-Hellauer, S. Kopeinik, D. Theiler, A. Haberl, S. Thalmann, and D. Kowald. Reproducibility in machine-learning-based research: Overview, barriers, and drivers.AI Magazine, 46(2):e70002, 2025

    work page 2025

Showing first 80 references.