pith. sign in

arxiv: 2606.24985 · v1 · pith:NY6ET5KGnew · submitted 2026-06-23 · 💻 cs.LG · cs.AI

Retrieval-Augmented Personalization with Foundation Models for Wearable Stress Detection

Louis Simon , Mohamed Chetouani This is my paper

Pith reviewed 2026-06-26 00:25 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords retrieval-augmented personalizationfoundation modelswearable stress detectionWESAD datasettransformerphysiological signalsEDA BVP
0
0 comments X

The pith

Frozen foundation models retrieve user history patterns to personalize wearable stress detection without labeled data.

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

The paper proposes using frozen out-of-domain foundation models for retrieval-augmented personalization in wearable stress detection. Similar patterns from a target user's history are retrieved and encoded into a compact embedding that modulates representations from a lightweight transformer network. On the WESAD dataset with 15 users and wrist-worn signals, this yields gains of 3.92 percent accuracy and 4.76 percent macro F1 over a non-personalized baseline. Performance approaches that of supervised fine-tuning while requiring no labeled user data. The method remains effective with only prior temporal samples and extends to cross-dataset retrieval from K-Emocon to WESAD.

Core claim

Our method leverages frozen, out-of-domain foundation models to retrieve similar patterns from a target user's history and encode them into a compact personalized embedding that modulates representations extracted by a lightweight transformer network. We evaluate our approach on the WESAD stress detection dataset with N=15 users, comprising wrist-worn physiological (EDA, BVP, temperature) and activity (accelerometer) signals, and report gains of +3.92% in accuracy and +4.76% in macro F1-score over a non-personalized transformer baseline, approaching supervised fine-tuning performance without requiring any labeled user data.

What carries the argument

Compact personalized embedding from frozen foundation model retrieval of user history patterns, which modulates the representations of a lightweight transformer network.

If this is right

  • Personalization succeeds without any labeled data from the target user.
  • Temporal retrieval using only prior user samples performs close to full intra-user retrieval.
  • Cross-dataset retrieval from embeddings of the K-Emocon dataset personalizes stress detection on WESAD.
  • The approach remains lightweight and avoids costly user-specific fine-tuning or large-scale pre-training.

Where Pith is reading between the lines

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

  • This approach may apply to other wearable tasks such as activity recognition or sleep staging where inter-user variability is high.
  • It implies that general foundation models can support physiological personalization through retrieval rather than direct adaptation.
  • Real-world deployment could benefit from reduced privacy concerns since no labeled target-user data is collected.
  • Scaling the retrieval corpus size or testing different foundation models could further improve gains.

Load-bearing premise

That retrieval of similar patterns using out-of-domain foundation models from user history provides effective personalization for the stress detection task on the WESAD dataset.

What would settle it

Observing no accuracy or F1 improvement, or a performance drop, when applying the retrieval-augmented method to a new set of users or a different wearable stress dataset would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.24985 by Louis Simon, Mohamed Chetouani.

Figure 1
Figure 1. Figure 1: CNN + Transformer Backbone: Wearable input signals are processed individually with Inception-like CNNs (left), fused and jointly process by a Transformer. The full architecture with max pooling and a prediction (right) serves as a baseline. In contrast, our method constructs a personalization embedding from unlabeled physiological recordings via frozen, out￾of-domain foundation models, requiring no affect … view at source ↗
Figure 2
Figure 2. Figure 2: The Custom SetTransformer pools the set c of K retrieved samples into a smaller set by attending to learned tokens I. 3.3 Retrieval-augmented personalization The set of retrieved pattern c is then used to condition representation from the main predictive transformer f(x) through a personalized decoder. While UserLLM learns interleaved cross-attention modules in a frozen LLM to condition prediction on user … view at source ↗
Figure 3
Figure 3. Figure 3: Our proposed personalized model conditions non-personalized representation f(x) with intra-user contex￾tual representations built via foundation model embedding retrieval. the K-EmoCon dataset utilizes the Empatica E4 2 . We build on the original preprocessing code 3 to remove baseline calibration and extract debate sections and the corresponding wearable signals. We then adopt the same filtering and stand… view at source ↗
Figure 4
Figure 4. Figure 4: The temporal evolution of the cross-user retrieval ratio in hybrid cold-start. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Personalization magnitude measures the contribution of personalized representation in the final prediction (the closer to zero, the weaker the personalization) [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Per-user accuracy and macro F1 (mean ± std across seeds). Cells highlighted in green (resp. red ) indicate improvements (resp. degradations) exceeding the median absolute difference with respect to the baseline. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: UMAP projections of raw foundation-model embeddings on WESAD. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
read the original abstract

Personalization in wearable-based stress detection remains challenging due to substantial inter-individual variability in physiological and behavioral responses. While traditional approaches rely on user-specific fine-tuning or costly self-supervised pre-training on large datasets, we propose a lightweight alternative based on retrieval-augmented personalization. Our method leverages frozen, out-of-domain foundation models to retrieve similar patterns from a target user's history and encode them into a compact personalized embedding that modulates representations extracted by a lightweight transformer network. We evaluate our approach on the WESAD stress detection dataset with N=15 users, comprising wrist-worn physiological (EDA, BVP, temperature) and activity (accelerometer) signals, and report gains of +3.92\% in accuracy and +4.76\% in macro F1-score over a non-personalized transformer baseline, approaching supervised fine-tuning performance without requiring any labeled user data. We further show that temporal retrieval, where only prior user samples are available, achieves performance close to full intra-user retrieval, demonstrating robustness to limited user history. Finally, we explore personalization in a cross-dataset retrieval setting, leveraging embeddings from the K-Emocon dataset to personalize representations for stress detection on the WESAD dataset.

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

1 major / 0 minor

Summary. The manuscript proposes a retrieval-augmented personalization method for wearable stress detection that leverages frozen out-of-domain foundation models to retrieve similar patterns from a target user's history, encodes them into a compact personalized embedding, and uses this embedding to modulate representations from a lightweight transformer network. On the WESAD dataset (N=15 users, wrist-worn EDA/BVP/temperature/accelerometer signals), the approach reports +3.92% accuracy and +4.76% macro F1 gains over a non-personalized transformer baseline while approaching supervised fine-tuning performance without any labeled user data. Additional results cover temporal retrieval (prior samples only) and cross-dataset retrieval from K-Emocon to WESAD.

Significance. If the empirical claims hold under rigorous controls, the work offers a practical, label-efficient route to personalization in physiological signal tasks where inter-user variability is high and labeled data per user is scarce. The explicit use of frozen foundation models and the temporal-retrieval ablation are strengths that could translate to deployment settings with limited history. The cross-dataset experiment further tests generalization of the retrieval mechanism.

major comments (1)
  1. [Abstract / Evaluation] Abstract and Evaluation section: the reported gains (+3.92% accuracy, +4.76% macro F1) are presented without any description of the experimental protocol (train/test split, cross-validation scheme, number of retrieval candidates, foundation-model choice and embedding dimension, or statistical tests). Because these details are load-bearing for the central claim that retrieval substitutes for labeled fine-tuning, the support for the performance numbers cannot be assessed from the provided text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback and recommendation. We respond to the major comment below.

read point-by-point responses

  1. Referee: [Abstract / Evaluation] Abstract and Evaluation section: the reported gains (+3.92% accuracy, +4.76% macro F1) are presented without any description of the experimental protocol (train/test split, cross-validation scheme, number of retrieval candidates, foundation-model choice and embedding dimension, or statistical tests). Because these details are load-bearing for the central claim that retrieval substitutes for labeled fine-tuning, the support for the performance numbers cannot be assessed from the provided text.

    Authors: We agree that the abstract and Evaluation section as presented lack a complete description of the experimental protocol, which limits assessment of the reported gains. We will revise the abstract to include a concise statement of the protocol and expand the Evaluation section to explicitly detail the train/test split, cross-validation scheme, number of retrieval candidates, foundation-model choice and embedding dimension, and statistical tests. These revisions will be incorporated in the next manuscript version. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper is an empirical ML methods paper proposing retrieval-augmented personalization using frozen out-of-domain foundation models to generate embeddings for a lightweight transformer on wearable stress data. No equations, derivations, or first-principles claims appear in the provided text. The central claims rest on experimental comparisons (gains over non-personalized baseline on WESAD, temporal and cross-dataset results) rather than any reduction of a 'prediction' to fitted inputs or self-citation chains. Self-citations, if present, are not load-bearing for any uniqueness theorem or ansatz. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are mentioned in the abstract.

pith-pipeline@v0.9.1-grok · 5735 in / 1078 out tokens · 35773 ms · 2026-06-26T00:25:34.428548+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

294 extracted references · 242 canonical work pages · 22 internal anchors

  1. [1]

    FiLM: Visual reasoning with a general conditioning layer.Proceedings of the AAAI Conference on Artificial Intelligence, 32(1), 2018

    Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2018 , keywords =. doi:10.1609/aaai.v32i1.11671 , abstract =

    work page doi:10.1609/aaai.v32i1.11671 2018

  2. [2]

    Set Transformer: A Framework for Attention-based Permutation-Invariant Neural Networks

    Lee, Juho and Lee, Yoonho and Kim, Jungtaek and Kosiorek, Adam R. and Choi, Seungjin and Teh, Yee Whye , month = may, year =. Set. doi:10.48550/arXiv.1810.00825 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1810.00825

  3. [3]

    Perceiver:

    Jaegle, Andrew and Gimeno, Felix and Brock, Andy and Vinyals, Oriol and Zisserman, Andrew and Carreira, Joao , month = jul, year =. Perceiver:. Proceedings of the 38th

  4. [4]

    Douze, Matthijs and Guzhva, Alexandr and Deng, Chengqi and Johnson, Jeff and Szilvasy, Gergely and Mazaré, Pierre-Emmanuel and Lomeli, Maria and Hosseini, Lucas and Jégou, Hervé , month = oct, year =. The. doi:10.48550/arXiv.2401.08281 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2401.08281

  5. [5]

    Assessing

    Christenson, Matthias and Geary, Cove and Locke, Brian and Koirala, Pranav and Pettine, Warren Woodrich , month = dec, year =. Assessing. doi:10.48550/arXiv.2412.03427 , abstract =

    work page doi:10.48550/arxiv.2412.03427

  6. [6]

    Phukan, Orchid Chetia and Behera, Swarup Ranjan and Girish and Akhtar, Mohd Mujtaba and Buduru, Arun Balaji and Sharma, Rajesh , month = oct, year =. Beyond. doi:10.48550/arXiv.2410.12645 , abstract =

    work page doi:10.48550/arxiv.2410.12645

  7. [7]

    Chetia Phukan, Orchid and Behera, Swarup and. Beyond. 2025 , doi =

    2025

  8. [8]

    Ahmed Al Dossary, Alia and Chollet, Mathieu and Vinciarelli, Alessandro , month = oct, year =. From. Proceedings of the 27th. doi:10.1145/3716553.3750753 , language =

    work page doi:10.1145/3716553.3750753

  9. [9]

    Proceedings of the 41st

    Goswami, Mononito and Szafer, Konrad and Choudhry, Arjun and Cai, Yifu and Li, Shuo and Dubrawski, Artur , month = jul, year =. Proceedings of the 41st

  10. [10]

    npj Digital Medicine , author =

    Self-supervised learning for human activity recognition using 700,000 person-days of wearable data , volume =. npj Digital Medicine , author =. 2024 , note =. doi:10.1038/s41746-024-01062-3 , abstract =

    work page doi:10.1038/s41746-024-01062-3 2024

  11. [11]

    Pillai, Arvind and Spathis, Dimitris and Kawsar, Fahim and Malekzadeh, Mohammad , month = oct, year =

  12. [12]

    and Narayanan, Shrikanth , month = oct, year =

    Avramidis, Kleanthis and Feng, Tiantian and Jeong, Woojae and Lee, Jihwan and Cui, Wenhui and Leahy, Richard M. and Narayanan, Shrikanth , month = oct, year =. Neural. doi:10.48550/arXiv.2510.09095 , abstract =

    work page doi:10.48550/arxiv.2510.09095

  13. [13]

    and Mascolo, Cecilia , month = apr, year =

    Spathis, Dimitris and Perez-Pozuelo, Ignacio and Brage, Soren and Wareham, Nicholas J. and Mascolo, Cecilia , month = apr, year =. Self-supervised transfer learning of physiological representations from free-living wearable data , url =. Proceedings of the. doi:10.1145/3450439.3451863 , abstract =

    work page doi:10.1145/3450439.3451863

  14. [14]

    and Futoma, Joseph , month = oct, year =

    Erturk, Eray and Kamran, Fahad and Abbaspourazad, Salar and Jewell, Sean and Sharma, Harsh and Li, Yujie and Williamson, Sinead and Foti, Nicholas J. and Futoma, Joseph , month = oct, year =. Beyond. Proceedings of the 42nd

  15. [15]

    and Futoma, Joseph , month = jun, year =

    Erturk, Eray and Kamran, Fahad and Abbaspourazad, Salar and Jewell, Sean and Sharma, Harsh and Li, Yujie and Williamson, Sinead and Foti, Nicholas J. and Futoma, Joseph , month = jun, year =. Beyond. doi:10.48550/arXiv.2507.00191 , abstract =

    work page doi:10.48550/arxiv.2507.00191

  16. [16]

    Foundation

    Lee, Simon and Akamatsu, Kai , month = aug, year =. Foundation

  17. [17]

    Luo, Yunfei and Chen, Yuliang and Salekin, Asif and Rahman, Tauhidur , month = may, year =. Toward. doi:10.48550/arXiv.2412.09758 , abstract =

    work page doi:10.48550/arxiv.2412.09758

  18. [18]

    , month = aug, year =

    Agarwal, Rohit and Sinha, Aman and Vishwakarma, Ayan and Coubez, Xavier and Clausel, Marianne and Constant, Mathieu and Horsch, Alexander and Prasad, Dilip K. , month = aug, year =. No. doi:10.48550/arXiv.2309.08698 , abstract =

    work page doi:10.48550/arxiv.2309.08698

  19. [19]

    Huynh, Lam and Nguyen, Tri and Nguyen, Thu and Pirttikangas, Susanna and Siirtola, Pekka , month = sep, year =. Adjunct. doi:10.1145/3460418.3479320 , language =

    work page doi:10.1145/3460418.3479320

  20. [20]

    Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , author =

    Systematic. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , author =. 2024 , pages =. doi:10.1145/3699724 , abstract =

    work page doi:10.1145/3699724 2024

  21. [21]

    Sensors (Basel, Switzerland) , author =

    Can. Sensors (Basel, Switzerland) , author =. 2020 , pmid =. doi:10.3390/s20226535 , abstract =

    work page doi:10.3390/s20226535 2020

  22. [22]

    Sentinel:

    Villaboni, Davide and Castellini, Alberto and Danesi, Ivan Luciano and Farinelli, Alessandro , month = mar, year =. Sentinel:. doi:10.48550/arXiv.2503.17658 , abstract =

    work page doi:10.48550/arxiv.2503.17658

  23. [23]

    A Time Series is Worth 64 Words: Long-term Forecasting with Transformers

    Nie, Yuqi and Nguyen, Nam H. and Sinthong, Phanwadee and Kalagnanam, Jayant , month = mar, year =. A. doi:10.48550/arXiv.2211.14730 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2211.14730

  24. [25]

    WE- SAD: a multimodal dataset for wearable stress and affect detection

    Schmidt, Philip and Reiss, Attila and Duerichen, Robert and Marberger, Claus and Van Laerhoven, Kristof , month = oct, year =. Introducing. Proceedings of the 20th. doi:10.1145/3242969.3242985 , abstract =

    work page doi:10.1145/3242969.3242985

  25. [26]

    2025 , note =

    google-research/tuning\_playbook , url =. 2025 , note =

    2025

  26. [27]

    On the. SIAM J. Numer. Anal. , author =. 1979 , pages =. doi:10.1137/0716058 , abstract =

    work page doi:10.1137/0716058 1979

  27. [28]

    Uncertainty-

    Huang, Yingsong and Bai, Bing and Zhao, Shengwei and Bai, Kun and Wang, Fei , month = jul, year =. Uncertainty-. doi:10.48550/arXiv.2207.05471 , abstract =

    work page doi:10.48550/arxiv.2207.05471

  28. [29]

    and Shah, Mubarak , month = apr, year =

    Rizve, Mamshad Nayeem and Duarte, Kevin and Rawat, Yogesh S. and Shah, Mubarak , month = apr, year =. In. doi:10.48550/arXiv.2101.06329 , abstract =

    work page doi:10.48550/arxiv.2101.06329

  29. [30]

    Learning From Noisy Labels By Regularized Estimation Of Annotator Confusion

    Tanno, Ryutaro and Saeedi, Ardavan and Sankaranarayanan, Swami and Alexander, Daniel C. and Silberman, Nathan , month = jun, year =. Learning. doi:10.48550/arXiv.1902.03680 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1902.03680 1902

  30. [31]

    doi:10.48550/arXiv.2011.07734 , abstract =

    Wang, Can and Chen, Jiawei and Zhou, Sheng and Shi, Qihao and Feng, Yan and Chen, Chun , month = jul, year =. doi:10.48550/arXiv.2011.07734 , abstract =

    work page doi:10.48550/arxiv.2011.07734 2011

  31. [32]

    ACM Transactions on Intelligent Systems and Technology , author =

    Labeling. ACM Transactions on Intelligent Systems and Technology , author =. 2024 , pages =. doi:10.1145/3626242 , abstract =

    work page doi:10.1145/3626242 2024

  32. [33]

    Yu, Wenhui and Qin, Zheng , month = jul, year =. Sampler. Proceedings of the 43rd. doi:10.1145/3397271.3401155 , language =

    work page doi:10.1145/3397271.3401155

  33. [34]

    Efficient

    Zhang, Wenzhen and Cheng, Debo and Lu, Guangquan and Zhou, Bo and Li, Jiaye and Zhang, Shichao , month = feb, year =. Efficient. doi:10.48550/arXiv.2502.00386 , abstract =

    work page doi:10.48550/arxiv.2502.00386

  34. [35]

    Neural Networks , author =

    Bayesian. Neural Networks , author =. 2024 , keywords =. doi:10.1016/j.neunet.2024.106122 , abstract =

    work page doi:10.1016/j.neunet.2024.106122 2024

  35. [36]

    Agreement or

    Liu, Jiarun and Jiang, Daguang and Yang, Yukun and Li, Ruirui , month = aug, year =. Agreement or. doi:10.48550/arXiv.2203.15317 , abstract =

    work page doi:10.48550/arxiv.2203.15317

  36. [37]

    Learning from

    Song, Hwanjun and Kim, Minseok and Park, Dongmin and Shin, Yooju and Lee, Jae-Gil , month = mar, year =. Learning from. doi:10.48550/arXiv.2007.08199 , abstract =

    work page doi:10.48550/arxiv.2007.08199 2007

  37. [38]

    Han, Jiangfan and Luo, Ping and Wang, Xiaogang , month = aug, year =. Deep. doi:10.48550/arXiv.1908.02160 , abstract =

    work page doi:10.48550/arxiv.1908.02160 1908

  38. [39]

    Pairwise

    Bai, Sihan , month = apr, year =. Pairwise. doi:10.48550/arXiv.2404.01853 , abstract =

    work page doi:10.48550/arxiv.2404.01853

  39. [40]

    doi:10.48550/arXiv.2402.17589 , abstract =

    Liu, Xiaoyu and Zhou, Beitong and Yue, Zuogong and Cheng, Cheng , month = nov, year =. doi:10.48550/arXiv.2402.17589 , abstract =

    work page doi:10.48550/arxiv.2402.17589

  40. [41]

    Pseudo-labelling meets

    Saravanan, Darshana and Manwani, Naresh and Gandhi, Vineet , month = jun, year =. Pseudo-labelling meets. doi:10.48550/arXiv.2402.04835 , abstract =

    work page doi:10.48550/arxiv.2402.04835

  41. [42]

    Patrini, Giorgio and Rozza, Alessandro and Menon, Aditya and Nock, Richard and Qu, Lizhen , month = mar, year =. Making. doi:10.48550/arXiv.1609.03683 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1609.03683

  42. [43]

    Mixed Precision , author =

    Tinkering. Mixed Precision , author =

  43. [44]

    Sequence

    Spathis, Dimitris and Servia-Rodriguez, Sandra and Farrahi, Katayoun and Mascolo, Cecilia and Rentfrow, Jason , year =. Sequence. Proceedings of the 25th. doi:10.1145/3292500.3330730 , abstract =

    work page doi:10.1145/3292500.3330730

  44. [45]

    Proceedings of the 46th

    Wu, Zhenchao and Zhou, Xiao , month = jul, year =. Proceedings of the 46th. doi:10.1145/3539618.3591719 , language =

    work page doi:10.1145/3539618.3591719

  45. [46]

    Personalized

    Yu, Runsheng and Gong, Yu and He, Xu and An, Bo and Zhu, Yu and Liu, Qingwen and Ou, Wenwu , month = dec, year =. Personalized. doi:10.48550/arXiv.2012.11842 , abstract =

    work page doi:10.48550/arxiv.2012.11842 2012

  46. [47]

    and Jain, Shomik and Chima, Kourtney and Rudovic, Ognjen and Matarić, Maja J

    Shi, Zhonghao and Groechel, Thomas R. and Jain, Shomik and Chima, Kourtney and Rudovic, Ognjen and Matarić, Maja J. , month = jan, year =. Toward. doi:10.48550/arXiv.2101.10580 , abstract =

    work page doi:10.48550/arxiv.2101.10580

  47. [48]

    Science Robotics , author =

    Personalized machine learning for robot perception of affect and engagement in autism therapy , volume =. Science Robotics , author =. 2018 , pmid =. doi:10.1126/scirobotics.aao6760 , abstract =

    work page doi:10.1126/scirobotics.aao6760 2018

  48. [49]

    , month = may, year =

    Huber, Bernd and Fazelnia, Ghazal and Damianou, Andreas and Peleato, Sebastian and Lefarov, Max and Ravichandran, Praveen and Nadai, Marco De and Lalmas-Roellke, Mounia and Bennett, Paul N. , month = may, year =. Embedding-to-. doi:10.48550/arXiv.2505.17051 , abstract =

    work page doi:10.48550/arxiv.2505.17051

  49. [50]

    Quantizable

    Bondarenko, Yelysei and Nagel, Markus and Blankevoort, Tijmen , month = nov, year =. Quantizable

  50. [51]

    Qiu, Yilun and Shi, Tianhao and Zhao, Xiaoyan and Zhu, Fengbin and Zhang, Yang and Feng, Fuli , month = jul, year =. Latent. doi:10.48550/arXiv.2507.20849 , abstract =

    work page doi:10.48550/arxiv.2507.20849

  51. [52]

    Bahdanau, Dzmitry and Cho, Kyunghyun and Bengio, Yoshua , editor =. Neural. 3rd

  52. [53]

    CoRR , author =

    Neural. CoRR , author =

  53. [54]

    Horn, Max and Moor, Michael and Bock, Christian and Rieck, Bastian and Borgwardt, Karsten , month = nov, year =. Set. Proceedings of the 37th

  54. [55]

    Advances in

    Choi, Edward and Bahadori, Mohammad Taha and Sun, Jimeng and Kulas, Joshua and Schuetz, Andy and Stewart, Walter , year =. Advances in

  55. [56]

    4369-4375, Macao, China

    In Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI-2019), pp. 4369-4375, Macao, China. , author =

    2019

  56. [57]

    Chen, Ricky T. Q. and Rubanova, Yulia and Bettencourt, Jesse and Duvenaud, David K , year =. Neural. Advances in

  57. [58]

    Rubanova, Yulia and Chen, Ricky T. Q. and Duvenaud, David K , year =. Latent. Advances in

  58. [59]

    Monitoring

    Harit, Anoushka and Sun, Zhongtian and Yu, Jongmin and Moubayed, Noura Al , month = oct, year =. Monitoring

  59. [61]

    Li, Yang and Du, Nan and Bengio, Samy , month = feb, year =. Time-

  60. [62]

    Towards Deep Learning Models for Psychological State Prediction using Smartphone Data: Challenges and Opportunities

    (. doi:10.48550/arXiv.1711.06350 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1711.06350

  61. [63]

    , year =

    Bucur, Ana-Maria and Cosma, Adrian and Rosso, Paolo and Dinu, Liviu P. , year =. It’s. Advances in. doi:10.1007/978-3-031-28244-7_13 , abstract =

    work page doi:10.1007/978-3-031-28244-7_13

  62. [65]

    Shaw, Peter and Uszkoreit, Jakob and Vaswani, Ashish , editor =. Self-. Proceedings of the 2018. 2018 , pages =. doi:10.18653/v1/N18-2074 , abstract =

    work page doi:10.18653/v1/n18-2074 2018

  63. [66]

    Unsupervised

    Tsitsulin, Anton and Munkhoeva, Marina and Perozzi, Bryan , month = sep, year =. Unsupervised. Proceedings of 2nd

  64. [67]

    Reparameterizable

    Xie, Sang Michael and Ermon, Stefano , month = feb, year =. Reparameterizable. doi:10.48550/arXiv.1901.10517 , abstract =

    work page doi:10.48550/arxiv.1901.10517 1901

  65. [68]

    Attention is

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N and Kaiser, Ł ukasz and Polosukhin, Illia , year =. Attention is. Advances in

  66. [69]

    Personalized

    Shaw, Abhinav and Simsiri, Natcha and Deznabi, Iman and Fiterau, Madalina and Rahaman, Tauhidur , month = may, year =. Personalized

  67. [70]

    Personalized

    Anonymous , month = may, year =. Personalized

  68. [71]

    ResearchGate , doi =

    (. ResearchGate , doi =

  69. [72]

    Kazemi, Seyed Mehran and Goel, Rishab and Eghbali, Sepehr and Ramanan, Janahan and Sahota, Jaspreet and Thakur, Sanjay and Wu, Stella and Smyth, Cathal and Poupart, Pascal and Brubaker, Marcus , month = sep, year =

  70. [73]

    Shukla, Satya Narayan and Marlin, Benjamin , month = oct, year =. Multi-

  71. [74]

    Prototypical networks for few-shot learning , isbn =

    Snell, Jake and Swersky, Kevin and Zemel, Richard , year =. Prototypical networks for few-shot learning , isbn =. Proceedings of the 31st

  72. [75]

    Bardes, Adrien and Ponce, Jean and Lecun, Yann , month = apr, year =

  73. [76]

    Advances in Neural Information Processing Systems , author =

    Achieving. Advances in Neural Information Processing Systems , author =. 2023 , pages =

    2023

  74. [77]

    Jiang, Wei-Bang and Fu, Xi and Ding, Yi and Guan, Cuntai , month = jun, year =. Towards. doi:10.48550/arXiv.2504.19596 , abstract =

    work page doi:10.48550/arxiv.2504.19596

  75. [78]

    ACM Comput

    Self-supervised. ACM Comput. Surv. , author =. 2025 , note =. doi:10.1145/3736574 , abstract =

    work page doi:10.1145/3736574 2025

  76. [79]

    IEEE Access , author =

    Applications of. IEEE Access , author =. 2023 , keywords =. doi:10.1109/ACCESS.2023.3344531 , abstract =

    work page doi:10.1109/access.2023.3344531 2023

  77. [80]

    Frontiers in Human Neuroscience , author =

    Self supervised learning based emotion recognition using physiological signals , volume =. Frontiers in Human Neuroscience , author =. 2024 , note =. doi:10.3389/fnhum.2024.1334721 , abstract =

    work page doi:10.3389/fnhum.2024.1334721 2024

  78. [81]

    Scientific Data , author =

    K-. Scientific Data , author =. 2020 , note =. doi:10.1038/s41597-020-00630-y , abstract =

    work page doi:10.1038/s41597-020-00630-y 2020

  79. [82]

    and Trajanovski, Stojan , month = oct, year =

    Saeed, Aaqib and Ozcelebi, Tanir and Lukkien, Johan and van Erp, Jan B.F. and Trajanovski, Stojan , month = oct, year =. Model. 2018. doi:10.1109/DSAA.2018.00031 , abstract =

    work page doi:10.1109/dsaa.2018.00031 2018

  80. [83]

    IEEE transactions on affective computing , author =

    Personalized. IEEE transactions on affective computing , author =. 2020 , pmid =. doi:10.1109/TAFFC.2017.2784832 , abstract =

    work page doi:10.1109/taffc.2017.2784832 2020

Showing first 80 references.