pith. sign in

arxiv: 2605.26194 · v2 · pith:EEMTSOYCnew · submitted 2026-05-25 · 💻 cs.LG

On the Role of Inductive Bias in Time-Series Pretraining: A Case Study in Learning Generalizable Representations for Clinical Time Series

Sharmita Dey , Diego Paez-Granados This is my paper

Pith reviewed 2026-06-29 23:11 UTC · model grok-4.3

classification 💻 cs.LG
keywords time-series pretraininginductive biasclinical time seriesgait analysisspinal cord injurytransformerrepresentation learningtransfer learning
0
0 comments X

The pith

Dynamics-centric mixtures of pretraining objectives produce the most balanced transfer across classification and regression tasks for clinical gait time series.

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

The paper investigates which inductive biases in pretraining objectives allow representations to transfer across subjects and task types when clinical time-series data are small and heterogeneous. It evaluates an encoder-centric transformer called PathoFM on multivariate gait windows from spinal cord injury patients, comparing grouping, dynamics-based, and generative objectives. The central result is that mixtures built around dynamics objectives—local reconstruction of masked spans plus prediction of temporal continuations, with in-context conditioning when exemplars are available—support both discriminative margins and magnitude fidelity better than pure alternatives. This matters for applications that must handle both pathology diagnosis and temporal forecasting under protocol drift. The work concludes that these mixtures deliver robust subject-generalizing representations.

Core claim

PathoFM is pretrained on multivariate gait windows using Local Completion to reconstruct contiguous masked spans, Temporal Continuity to predict a masked mid-horizon continuation from an observed prefix, and Unsupervised In-Context Dynamics to support query reconstruction conditioned on subject exemplar windows. Empirical comparison shows dynamics-centric mixtures achieve the most balanced transfer: grouping objectives improve classification margins yet degrade magnitude accuracy needed for continuous targets, while reconstruction-only objectives preserve waveform structure yet underperform on classification tasks.

What carries the argument

The three complementary pretraining objectives—Local Completion, Temporal Continuity, and Unsupervised In-Context Dynamics—applied inside an encoder-centric transformer to enforce local structure, smoothness, and causal consistency.

If this is right

  • Combining local reconstruction with temporal continuity yields representations that support both classification and continuous regression without large trade-offs.
  • Adding unsupervised in-context conditioning improves subject generalization when subject exemplar windows are available at inference time.
  • Grouping objectives alone improve discriminative performance but reduce fidelity on magnitude-sensitive downstream targets.
  • Pure reconstruction objectives maintain waveform structure yet show weaker results on classification tasks.

Where Pith is reading between the lines

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

  • The same objective mixture could be tested on other small-cohort clinical series such as vital-sign monitoring to check whether the balanced-transfer pattern holds beyond gait.
  • Protocol-drift robustness could be probed by training on one hospital's data and evaluating on another's without retraining.
  • When exemplar access is unavailable the in-context component can be dropped without collapsing the remaining local-plus-continuity benefits.

Load-bearing premise

Results from the single SCI gait dataset are assumed to demonstrate inductive biases that transfer across task types and subjects in general.

What would settle it

A replication on a different clinical time-series dataset such as cardiac or EEG recordings where the dynamics-centric mixture no longer outperforms grouping or reconstruction-only objectives on balanced transfer metrics for both classification and regression.

Figures

Figures reproduced from arXiv: 2605.26194 by Diego Paez-Granados, Sharmita Dey.

Figure 1
Figure 1. Figure 1: PATHOFM pretraining objectives and architecture. A shared transformer encoder processes multivariate gait windows. (i) Local Completion reconstructs masked spans. (ii) Temporal Continuity predicts masked futures from observed past. (iii) Unsupervised In-Context Dynamics reconstructs a masked query window conditioned on same-subject support windows in a small table, enabling non-parametric adaptation via at… view at source ↗
Figure 2
Figure 2. Figure 2: Qualitative pretext sanity check on an unseen SCI subject. Ground truth (solid) vs. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
read the original abstract

Clinical time-series learning is routinely constrained by small, heterogeneous cohorts and protocol drift, while its downstream use spans both classification (e.g., pathology diagnosis) and regression (e.g., temporal forecasting). These constraints make foundation-model pretraining appealing, but raises an important question of which inductive biases should the pretraining objective impose so that representations transfer across task types and subjects. We study this question in pathological gait analysis for spinal cord injury (SCI) via PathoFM, an encoder-centric transformer pretrained on multivariate gait windows with three complementary objectives: Local Completion (reconstruct contiguous masked spans to enforce local structure), Temporal Continuity (predict a masked mid-horizon continuation from an observed prefix to enforce smoothness and causal consistency), and Unsupervised In-Context Dynamics (support-query reconstruction conditioned on subject exemplar windows via attention). Empirically comparing objective families (grouping/contrastive, dynamics-based, and generative reconstruction), we find that dynamics-centric mixtures produce the most balanced transfer: grouping objectives favor discriminative margins but can degrade magnitude fidelity needed for continuous targets, whereas reconstruction-only objectives preserve waveform structure but may underperform on classification. Overall, combining local reconstruction with temporal continuity, and adding in-context conditioning when exemplar access is realistic, yields robust subject-generalizing representations.

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 PathoFM, an encoder-centric transformer pretrained on multivariate gait windows from spinal cord injury (SCI) patients using three objectives: Local Completion (reconstruct masked spans), Temporal Continuity (predict mid-horizon continuation), and Unsupervised In-Context Dynamics (support-query reconstruction conditioned on subject exemplars). Through empirical comparison of objective families (grouping/contrastive, dynamics-based, generative reconstruction) on the SCI gait dataset, it claims that dynamics-centric mixtures produce the most balanced transfer across classification and regression tasks and across subjects, while grouping objectives favor discriminative margins at the cost of magnitude fidelity and pure reconstruction preserves waveform structure but underperforms on classification.

Significance. If the comparative results hold under proper statistical controls and external validation, the work supplies actionable guidance on selecting inductive biases for clinical time-series foundation models operating under small-cohort and protocol-drift constraints, explicitly contrasting the trade-offs between discriminative, continuity, and reconstruction objectives for mixed task types.

major comments (2)
  1. [Empirical evaluation] Empirical evaluation (abstract and results): no dataset sizes, number of subjects, baseline implementations, statistical significance tests, error bars, or exclusion criteria are reported, preventing assessment of whether the observed advantages of dynamics-centric mixtures are load-bearing or artifactual.
  2. [Abstract and discussion] Abstract and §5 (discussion of generalizability): the central recommendation for dynamics-centric mixtures rests on transfer metrics from a single SCI gait dataset whose periodic multivariate structure and subject-variability patterns may not extend to other clinical domains (e.g., ECG, EEG); without cross-domain experiments or explicit discussion of protocol drift, the claim that these mixtures yield subject-generalizing representations cannot be verified beyond the case study.
minor comments (1)
  1. [Abstract] Abstract: minor phrasing issue in 'raises an important question of which inductive biases should the pretraining objective impose'—rephrase for grammatical clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on empirical reporting and scope. We address each major comment below.

read point-by-point responses

  1. Referee: [Empirical evaluation] Empirical evaluation (abstract and results): no dataset sizes, number of subjects, baseline implementations, statistical significance tests, error bars, or exclusion criteria are reported, preventing assessment of whether the observed advantages of dynamics-centric mixtures are load-bearing or artifactual.

    Authors: We agree that these details are necessary for evaluating the results. In the revised manuscript we will add a dedicated experimental setup section reporting: the number of SCI subjects and total multivariate gait windows, data splits and exclusion criteria, full baseline implementations with hyperparameters and references, statistical significance tests (e.g., subject-wise paired tests with p-values), and error bars (standard deviation across subjects or folds). These additions will substantiate the reported advantages of the dynamics-centric mixtures. revision: yes

  2. Referee: [Abstract and discussion] Abstract and §5 (discussion of generalizability): the central recommendation for dynamics-centric mixtures rests on transfer metrics from a single SCI gait dataset whose periodic multivariate structure and subject-variability patterns may not extend to other clinical domains (e.g., ECG, EEG); without cross-domain experiments or explicit discussion of protocol drift, the claim that these mixtures yield subject-generalizing representations cannot be verified beyond the case study.

    Authors: The manuscript is explicitly positioned as a case study on SCI gait (see title and abstract). We will revise the abstract and §5 to explicitly bound the claims to this domain, note the periodic structure of gait data, and add a limitations paragraph discussing protocol drift and the absence of cross-domain validation on modalities such as ECG or EEG. We will also clarify that subject generalization is demonstrated across held-out subjects within the SCI cohort and outline why the inductive biases may be relevant more broadly, while acknowledging that transfer to other domains remains future work. revision: partial

Circularity Check

0 steps flagged

No significant circularity; empirical claims rest on dataset comparisons without definitional reduction

full rationale

The paper is an empirical case study that compares families of pretraining objectives (grouping/contrastive, dynamics-based, generative reconstruction) on the SCI gait dataset and reports transfer performance for classification and regression tasks. The abstract and provided text contain no equations, no fitted parameters renamed as predictions, no self-citations invoked as uniqueness theorems, and no ansatzes smuggled through prior work. All load-bearing statements are presented as outcomes of the reported experiments rather than derivations that reduce to their own inputs by construction. This is the normal, non-circular outcome for an empirical methods paper.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 1 invented entities

The central claim rests on standard representation learning assumptions plus the domain assumption that the chosen objectives impose transferable inductive biases; the paper introduces PathoFM as a new named model but provides no independent evidence for its components beyond the empirical results.

free parameters (2)
  • mask span length for Local Completion
    Hyperparameter selected to enforce local structure in the reconstruction objective.
  • mid-horizon length for Temporal Continuity
    Hyperparameter selected to enforce smoothness and causal consistency in the continuation prediction.
axioms (1)
  • domain assumption Pretraining objectives that combine local reconstruction, temporal continuity, and in-context dynamics will produce representations that transfer across task types and subjects in clinical time series.
    This is the core hypothesis tested by the empirical comparison in the abstract.
invented entities (1)
  • PathoFM no independent evidence
    purpose: Encoder-centric transformer pretrained with the three complementary objectives for clinical gait data.
    New model name introduced for the proposed architecture; no independent evidence provided beyond the study itself.

pith-pipeline@v0.9.1-grok · 5764 in / 1565 out tokens · 39304 ms · 2026-06-29T23:11:38.071899+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

56 extracted references · 10 canonical work pages · 2 internal anchors

  1. [1]

    The evolution of clinical gait analysis part iii–kinetics and energy assessment.Gait & posture, 21(4):447–461, 2005

    David H Sutherland. The evolution of clinical gait analysis part iii–kinetics and energy assessment.Gait & posture, 21(4):447–461, 2005

    2005

  2. [2]

    Estimation of minimal data sets sizes for machine learning predictions in digital mental health interventions.npj Digital Medicine, 7(1):361, 2024

    Kirsten Zantvoort, Barbara Nacke, Dennis Görlich, Silvan Hornstein, Corinna Jacobi, and Burkhardt Funk. Estimation of minimal data sets sizes for machine learning predictions in digital mental health interventions.npj Digital Medicine, 7(1):361, 2024

    2024

  3. [3]

    Domain adaptation for medical image analysis: a survey.IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, 2021

    Hao Guan and Mingxia Liu. Domain adaptation for medical image analysis: a survey.IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, 2021

    2021

  4. [4]

    A survey on negative transfer.IEEE/CAA Journal of Automatica Sinica, 10(2):305–329, 2022

    Wen Zhang, Lingfei Deng, Lei Zhang, and Dongrui Wu. A survey on negative transfer.IEEE/CAA Journal of Automatica Sinica, 10(2):305–329, 2022

    2022

  5. [5]

    Multitask learning and benchmarking with clinical time series data.Scientific data, 6(1):96, 2019

    Hrayr Harutyunyan, Hrant Khachatrian, David C Kale, Greg Ver Steeg, and Aram Galstyan. Multitask learning and benchmarking with clinical time series data.Scientific data, 6(1):96, 2019

    2019

  6. [6]

    Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Nee- lakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

    1901

  7. [7]

    On the Opportunities and Risks of Foundation Models

    Rishi Bommasani. On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  8. [8]

    Reducing the dimensionality of data with neural networks

    Geoffrey E Hinton and Ruslan R Salakhutdinov. Reducing the dimensionality of data with neural networks. science, 313(5786):504–507, 2006

    2006

  9. [9]

    Why does unsupervised pre-training help deep learning? InProceedings of the thirteenth international conference on artificial intelligence and statistics, pages 201–208

    Dumitru Erhan, Aaron Courville, Yoshua Bengio, and Pascal Vincent. Why does unsupervised pre-training help deep learning? InProceedings of the thirteenth international conference on artificial intelligence and statistics, pages 201–208. JMLR Workshop and Conference Proceedings, 2010

    2010

  10. [10]

    The need for biases in learning generalizations

    Tom M Mitchell. The need for biases in learning generalizations. 1980

    1980

  11. [11]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations. InInternational conference on machine learning, pages 1597–1607. PmLR, 2020

    2020

  12. [12]

    Representation Learning with Contrastive Predictive Coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  13. [13]

    Ti-mae: Self-supervised masked time series autoencoders.arXiv preprint arXiv:2301.08871, 2023

    Zhe Li, Zhongwen Rao, Lujia Pan, Pengyun Wang, and Zenglin Xu. Ti-mae: Self-supervised masked time series autoencoders.arXiv preprint arXiv:2301.08871, 2023

    work page arXiv 2023

  14. [14]

    CRC Press, 2024

    Jacquelin Perry and Judith Burnfield.Gait analysis: normal and pathological function. CRC Press, 2024. 10 Role of Inductive Bias in Time-Series PretrainingA PREPRINT

    2024

  15. [15]

    A survey of human gait-based artificial intelligence applications

    Elsa J Harris, I-Hung Khoo, and Emel Demircan. A survey of human gait-based artificial intelligence applications. Frontiers in Robotics and AI, 8:749274, 2022

    2022

  16. [16]

    John wiley & sons, 2009

    David A Winter.Biomechanics and motor control of human movement. John wiley & sons, 2009

    2009

  17. [17]

    A survey on self-supervised learning: Algorithms, applications, and future trends.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12):9052–9071, 2024

    Jie Gui, Tuo Chen, Jing Zhang, Qiong Cao, Zhenan Sun, Hao Luo, and Dacheng Tao. A survey on self-supervised learning: Algorithms, applications, and future trends.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12):9052–9071, 2024

    2024

  18. [18]

    Contrastive representation learning: A framework and review.Ieee Access, 8:193907–193934, 2020

    Phuc H Le-Khac, Graham Healy, and Alan F Smeaton. Contrastive representation learning: A framework and review.Ieee Access, 8:193907–193934, 2020

    2020

  19. [19]

    Prototypical contrastive learning of unsupervised representations,

    Junnan Li, Pan Zhou, Caiming Xiong, and Steven CH Hoi. Prototypical contrastive learning of unsupervised representations.arXiv preprint arXiv:2005.04966, 2020

    work page arXiv 2005

  20. [20]

    Momentum contrast for unsupervised visual representation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual representation learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9729–9738, 2020

    2020

  21. [21]

    Emerging properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerging properties in self-supervised vision transformers. InProceedings of the IEEE/CVF international conference on computer vision, pages 9650–9660, 2021

    2021

  22. [22]

    Self-supervised contrastive learning for medical time series: A systematic review.Sensors, 23(9):4221, 2023

    Ziyu Liu, Azadeh Alavi, Minyi Li, and Xiang Zhang. Self-supervised contrastive learning for medical time series: A systematic review.Sensors, 23(9):4221, 2023

    2023

  23. [23]

    Self-supervised transformer model training for a sleep-eeg foundation model

    Mattson Ogg and William G Coon. Self-supervised transformer model training for a sleep-eeg foundation model. In2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 1–6. IEEE, 2024

    2024

  24. [24]

    Ts2vec: Towards universal representation of time series

    Zhihan Yue, Yujing Wang, Juanyong Duan, Tianmeng Yang, Congrui Huang, Yunhai Tong, and Bixiong Xu. Ts2vec: Towards universal representation of time series. InProceedings of the AAAI conference on artificial intelligence, volume 36, pages 8980–8987, 2022

    2022

  25. [25]

    Time series data augmentation for deep learning: A survey.arXiv preprint arXiv:2002.12478, 2020

    Qingsong Wen, Liang Sun, Fan Yang, Xiaomin Song, Jingkun Gao, Xue Wang, and Huan Xu. Time series data augmentation for deep learning: A survey.arXiv preprint arXiv:2002.12478, 2020

    work page arXiv 2002

  26. [26]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16000–16009, 2022

    2022

  27. [27]

    Temporal fusion transformers for interpretable multi-horizon time series forecasting.International journal of forecasting, 37(4):1748–1764, 2021

    Bryan Lim, Sercan Ö Arık, Nicolas Loeff, and Tomas Pfister. Temporal fusion transformers for interpretable multi-horizon time series forecasting.International journal of forecasting, 37(4):1748–1764, 2021

    2021

  28. [28]

    Continual learning from simulated interactions via multitask prospective rehearsal for bionic limb behavior modeling.arXiv preprint arXiv:2405.01114, 2024

    Sharmita Dey, Benjamin Paassen, Sarath Ravindran Nair, Sabri Boughorbel, and Arndt F Schilling. Continual learning from simulated interactions via multitask prospective rehearsal for bionic limb behavior modeling.arXiv preprint arXiv:2405.01114, 2024

    work page arXiv 2024

  29. [29]

    Self-supervised learning for time series analysis: Taxonomy, progress, and prospects.IEEE transactions on pattern analysis and machine intelligence, 46(10):6775–6794, 2024

    Kexin Zhang, Qingsong Wen, Chaoli Zhang, Rongyao Cai, Ming Jin, Yong Liu, James Y Zhang, Yuxuan Liang, Guansong Pang, Dongjin Song, et al. Self-supervised learning for time series analysis: Taxonomy, progress, and prospects.IEEE transactions on pattern analysis and machine intelligence, 46(10):6775–6794, 2024

    2024

  30. [30]

    Scheduled sampling for sequence prediction with recurrent neural networks.Advances in neural information processing systems, 28, 2015

    Samy Bengio, Oriol Vinyals, Navdeep Jaitly, and Noam Shazeer. Scheduled sampling for sequence prediction with recurrent neural networks.Advances in neural information processing systems, 28, 2015

    2015

  31. [31]

    Improving multi-step prediction of learned time series models

    Arun Venkatraman, Martial Hebert, and J Bagnell. Improving multi-step prediction of learned time series models. InProceedings of the AAAI Conference on Artificial Intelligence, volume 29, 2015

    2015

  32. [32]

    Csdi: Conditional score-based diffusion models for probabilistic time series imputation.Advances in neural information processing systems, 34:24804–24816, 2021

    Yusuke Tashiro, Jiaming Song, Yang Song, and Stefano Ermon. Csdi: Conditional score-based diffusion models for probabilistic time series imputation.Advances in neural information processing systems, 34:24804–24816, 2021

    2021

  33. [33]

    Cross-modal diffusion for biomechanical dynamical systems through local manifold alignment.arXiv preprint arXiv:2503.12214, 2025

    Sharmita Dey and Sarath Ravindran Nair. Cross-modal diffusion for biomechanical dynamical systems through local manifold alignment.arXiv preprint arXiv:2503.12214, 2025

    work page arXiv 2025

  34. [34]

    Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  35. [35]

    Improved denoising diffusion probabilistic models

    Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. InInterna- tional conference on machine learning, pages 8162–8171. PMLR, 2021

    2021

  36. [36]

    Joint diffusion models in continual learning

    Paweł Skier´s and Kamil Deja. Joint diffusion models in continual learning. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4380–4390, 2025. 11 Role of Inductive Bias in Time-Series PretrainingA PREPRINT

    2025

  37. [37]

    Gruver, N., Finzi, M., Qiu, S., and Wilson, A

    Abhimanyu Das, Matthew Faw, Rajat Sen, and Yichen Zhou. In-context fine-tuning for time-series foundation models.arXiv preprint arXiv:2410.24087, 2024

    work page arXiv 2024

  38. [38]

    arXiv preprint arXiv:2405.14982 , year=

    Jiecheng Lu, Yan Sun, and Shihao Yang. In-context time series predictor.arXiv preprint arXiv:2405.14982, 2024

    work page arXiv 2024

  39. [39]

    Matching networks for one shot learning

    Oriol Vinyals, Charles Blundell, Timothy Lillicrap, Daan Wierstra, et al. Matching networks for one shot learning. Advances in neural information processing systems, 29, 2016

    2016

  40. [40]

    Cased-based reasoning for medical knowledge-based systems.International Journal of Medical Informatics, 64(2-3):355–367, 2001

    Rainer Schmidt, Stefania Montani, Riccardo Bellazzi, Luigi Portinale, and Lothar Gierl. Cased-based reasoning for medical knowledge-based systems.International Journal of Medical Informatics, 64(2-3):355–367, 2001

    2001

  41. [41]

    Sharmita Dey and Sarath Ravindran Nair. Remap: Neural model reprogramming with network inversion and retrieval-augmented mapping for adaptive motion forecasting.Advances in Neural Information Processing Systems, 37:25195–25227, 2024

    2024

  42. [42]

    Sharmita Dey and Arndt F Schilling. A function approximator model for robust online foot angle trajectory prediction using a single imu sensor: Implication for controlling active prosthetic feet.IEEE Transactions on Industrial Informatics, 19(2):1467–1475, 2022

    2022

  43. [43]

    A hybrid approach for dynamically training a torque prediction model for devising a human- machine interface control strategy.arXiv preprint arXiv:2110.03085, 2021

    Sharmita Dey, Takashi Yoshida, Robert H Foerster, Michael Ernst, Thomas Schmalz, Rodrigo M Carnier, and Arndt F Schilling. A hybrid approach for dynamically training a torque prediction model for devising a human- machine interface control strategy.arXiv preprint arXiv:2110.03085, 2021

    work page arXiv 2021

  44. [44]

    Sharmita Dey, Mahdy Eslamy, Takashi Yoshida, Michael Ernst, Thomas Schmalz, and ArndtF Schilling. A support vector regression approach for continuous prediction of ankle angle and moment during walking: An implication for developing a control strategy for active ankle prostheses. In2019 IEEE 16th International Conference on Rehabilitation Robotics (ICORR)...

    2019

  45. [45]

    Sharmita Dey, Takashi Yoshida, Robert H Foerster, Michael Ernst, Thomas Schmalz, and Arndt F Schilling. Continuous prediction of joint angular positions and moments: A potential control strategy for active knee-ankle prostheses.IEEE Transactions on Medical Robotics and Bionics, 2(3):347–355, 2020

    2020

  46. [46]

    A random forest approach for continuous prediction of joint angles and moments during walking: An implication for controlling active knee-ankle prostheses/orthoses

    Sharmita Dey, Takashi Yoshida, Michael Ernst, Thomas Schmalz, and Arndt F Schilling. A random forest approach for continuous prediction of joint angles and moments during walking: An implication for controlling active knee-ankle prostheses/orthoses. In2019 IEEE International conference on Cyborg and bionic systems (CBS), pages 66–71. IEEE, 2019

    2019

  47. [47]

    Continuous-phase control of a powered knee–ankle prosthesis: Amputee experiments across speeds and inclines.IEEE Transactions on Robotics, 34(3):686–701, 2018

    David Quintero, Dario J Villarreal, Daniel J Lambert, Susan Kapp, and Robert D Gregg. Continuous-phase control of a powered knee–ankle prosthesis: Amputee experiments across speeds and inclines.IEEE Transactions on Robotics, 34(3):686–701, 2018

    2018

  48. [48]

    A survey on in-context learning

    Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Jingyuan Ma, Rui Li, Heming Xia, Jingjing Xu, Zhiyong Wu, Baobao Chang, et al. A survey on in-context learning. InProceedings of the 2024 conference on empirical methods in natural language processing, pages 1107–1128, 2024

    2024

  49. [49]

    Variadic meta-learning by bayesian nonparametric deep embedding.Advances in Neural Information Processing Systems (NeurIPS), 2018

    Kelsey R Allen, Hanul Shin, Evan Shelhamer, and Joshua B Tenenbaum. Variadic meta-learning by bayesian nonparametric deep embedding.Advances in Neural Information Processing Systems (NeurIPS), 2018

    2018

  50. [50]

    Explaining machine learning models for clinical gait analysis.ACM Transactions on Computing for Healthcare (HEALTH), 3(2):1–27, 2021

    Djordje Slijepcevic, Fabian Horst, Sebastian Lapuschkin, Brian Horsak, Anna-Maria Raberger, Andreas Kranzl, Wojciech Samek, Christian Breiteneder, Wolfgang Immanuel Schöllhorn, and Matthias Zeppelzauer. Explaining machine learning models for clinical gait analysis.ACM Transactions on Computing for Healthcare (HEALTH), 3(2):1–27, 2021

    2021

  51. [51]

    Automatic classification of pathological gait patterns using ground reaction forces and machine learning algorithms

    Murad Alaqtash, Thompson Sarkodie-Gyan, Huiying Yu, Olac Fuentes, Richard Brower, and Amr Abdelgawad. Automatic classification of pathological gait patterns using ground reaction forces and machine learning algorithms. In2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pages 453–

  52. [52]

    Case-based reasoning for medical decision support tasks: The inreca approach.Artificial Intelligence in Medicine, 12(1):25–41, 1998

    Klaus-Dieter Althoff, Ralph Bergmann, Stefan Wess, Michel Manago, Eric Auriol, Oleg I Larichev, Alexander Bolotov, Yurii I Zhuravlev, and Serge I Gurov. Case-based reasoning for medical decision support tasks: The inreca approach.Artificial Intelligence in Medicine, 12(1):25–41, 1998

    1998

  53. [53]

    Deep learning based ground reaction force estimation for stair walking using kinematic data.Measurement, 198:111344, 2022

    Dongwei Liu, Ming He, Meijin Hou, and Ye Ma. Deep learning based ground reaction force estimation for stair walking using kinematic data.Measurement, 198:111344, 2022

    2022

  54. [54]

    Clinical time series prediction: Toward a hierarchical dynamical system framework.Artificial intelligence in medicine, 65(1):5–18, 2015

    Zitao Liu and Milos Hauskrecht. Clinical time series prediction: Toward a hierarchical dynamical system framework.Artificial intelligence in medicine, 65(1):5–18, 2015

    2015

  55. [55]

    Multi-task learning as multi-objective optimization.Advances in neural information processing systems, 31, 2018

    Ozan Sener and Vladlen Koltun. Multi-task learning as multi-objective optimization.Advances in neural information processing systems, 31, 2018

    2018

  56. [56]

    Scim–spinal cord independence measure: a new disability scale for patients with spinal cord lesions.Spinal cord, 35(12):850–856, 1997

    Amiram Catz, M Itzkovich, E Agranov, H Ring, and A Tamir. Scim–spinal cord independence measure: a new disability scale for patients with spinal cord lesions.Spinal cord, 35(12):850–856, 1997. 12

    1997