Pith. sign in

REVIEW 4 major objections 5 minor 146 references

Human Heterogeneity Invariant Stress Sensing

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Trained only on healthy people, HHISS reaches 75% mean balanced accuracy on opioid-use-disorder patients before and after daily medication, beating all tested OOD baselines; the mechanism is intersecting per-person pruning masks to keep…

desk verdict A genuinely novel pruning-intersection idea with impressive dataset breadth, but the headline OOD margin is a single-split point estimate with no error bars, so 'consistently outperformed' is not yet established. read the letter →

arxiv 2506.02256 v1 pith:27IPTYQZ submitted 2025-06-02 cs.LG cs.AI

Yi Xiao , Harshit Sharma , Sawinder Kaur , Dessa Bergen-Cico , Asif Salekin This is my paper
classification cs.LGcs.AI
keywords stressdetectiondomaingeneralizationout-of-distributionneuralnetworkpruningwearablephysiologicalsensingopioidusedisorderinvariantfeaturelearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that a stress detector trained only on healthy people can generalize to people with opioid use disorder, including the same person before and after their daily methadone dose, if it is built to keep only the physiological patterns shared across all individuals. The method, HHISS, combines four components: treating each person as a domain under invariant-risk-minimization training, pruning each person's network separately and intersecting the surviving parameter masks, re-growing that intersection through sparse-to-sparse training, and fine-tuning with the continuous labels of the over-parameterized model. On the paper's headline evaluation, HHISS reaches 75.0% mean balanced accuracy on out-of-distribution OUD data versus 69.96% for the best baseline, and it also leads on unseen-stressor and in-the-wild tests. If correct, the result would mean clinical stress monitoring does not require collecting labeled stress data from the very population it serves. The authors themselves note that the OUD dataset is small and that cross-device sensor differences are outside the method's scope.

What carries the argument

The load-bearing object is the person-wise pruning intersection: for each training subject, prune the IRM-trained network down to the top $(100-K)\%$ of weights ranked by $|\text{gradient} \times \text{weight}|$ (Eq. 3), producing one binary mask per person, and then take the element-wise intersection of all subjects' masks (Eq. 2) so a parameter survives only if it is important to every individual. That intersection is meant to isolate parameters that are consistently important across humans, while parameters outside it are treated as bias-inducing and zeroed out. Three supporting mechanisms make the intersection usable: IRM regularization with each subject as a domain supplies the initial invariant representation; sparse-to-sparse training re-learns pruned parameters between rounds so the intersection converges to a larger, stable sub-network; and a continuous-label loss, drawn from the over-parameterized model's logits without temperature scaling, pushes the final model toward a flat minimum in the loss landscape. The ablations show that all components are needed, with the pruning component contributing the largest single performance drop when removed.

What would settle it

Train HHISS on healthy controls exactly as described, then apply it to a fresh held-out clinical cohort (another OUD group, or a different condition such as chronic pain patients) and compute the per-subject pruning masks of the new cohort: if those masks barely overlap the training intersection and balanced accuracy on that cohort falls back toward ordinary ERM's roughly 67%, the shared-parameter assumption is not doing the work. The same test can be run without new data by checking whether the reported 75.0% OOD mean holds under alternative person-disjoint random splits of the existing control subjects into training and test sets.

Watch

Extended reading notes

Core claim

HHISS's central claim is that person-specific bias, not person-specific information, is what breaks stress-detection models when they move to a new population. The paper operationalizes this by training an over-parameterized network with an IRM penalty that treats each subject as a domain, then computing, for each subject separately, a binary mask that keeps the top $(100-K)\%$ of weights by importance score $|\nabla_\theta L||\theta|$, and intersecting these masks across subjects so that only the parameters every person's sub-network agrees on survive while the rest are zeroed out. Sparse-to-sparse training then lets the intersection grow from roughly 30% to 40% of parameters across rounds while remaining stable, and a final loss term enforces agreement with the continuous logits of the original over-parameterized model, which the authors associate with a flatter loss landscape. The evidence carrying the claim is the evaluation sweep: trained on healthy controls, the model transfers to pre- and post-dose OUD sessions, to WESAD and driving data with unseen stressors (about 74% on driving), and to in-the-wild nursing and daily OUD monitoring, beating all tested generalization baselines in and out of distribution.

Load-bearing premise

The method assumes that the network parameters every training subject would keep under pruning are exactly the ones that matter for new people, and that anything outside that shared intersection is person-specific bias safe to discard.

Editorial extensions

If this is right

  • A stress-detection model can be built entirely from healthy-control data and still serve a clinical population it never saw, including opioid-use-disorder patients both before and after their daily medication dose (Table 3: 75.0% OOD mean accuracy vs. 69.96% best baseline).
  • Disease-specific training data is not the bottleneck: training on pre-dose or post-dose OUD data alone is less effective than training on diverse control data, so scarce-population data can be reserved for evaluation instead of training (Tables 4-5).
  • The method transfers across stressors, environments, and daily-life conditions rather than only lab protocols: it outperforms baselines on WESAD, on driving data (AffectiveROAD), and on in-the-wild nursing and at-home OUD monitoring.
  • The framework is backbone-agnostic, improving out-of-distribution accuracy with plain DNNs, ResNets, and multi-head attention, and its inference cost (about one millisecond on a phone) is small enough for wearable deployment.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The intersection-mask idea is a general recipe for any human-sensing task where the training population is a convenience sample and the deployment population is not; one testable extension is applying HHISS-style masks to affect recognition or seizure detection across age groups.
  • The paper's own observation that more variable training groups (controls) yield better generalization than more homogeneous ones (pre/post-dose) suggests a data-collection policy: for scarce clinical populations, prioritize breadth of everyday variation over condition-specific collection.
  • A concrete check the authors do not run is whether the benefit scales with training-population diversity, for example comparing OOD accuracy as the number of training subjects grows and whether the stable intersection-mask size (30% to 40%) shifts with that diversity.
  • Because the continuous-label loss is intentionally not temperature-scaled knowledge distillation, an immediate experiment is to vary the softness of those labels and see whether the flat-minimum effect, and with it the OOD gain, survives.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents HHISS (Human Heterogeneity Invariant Stress Sensing), a domain-generalization method for wearable stress detection that trains on healthy controls and aims to generalize to out-of-distribution populations, including individuals with opioid use disorder (OUD). HHISS combines IRM-regularized training with a novel per-subject gradient-pruning procedure: for each training subject a binary pruning mask is computed, the intersection of these masks is taken as the 'invariant sub-network', and the resulting sparse network is refined through sparse-to-sparse retraining while being supervised with continuous logits from the original over-parameterized model. The authors evaluate HHISS on seven datasets (four self-collected, three public), reporting that HHISS 'consistently outperformed' state-of-the-art baseline methods on out-of-distribution stress detection, with the headline result in Table 3 of 75.0% mean OOD balanced accuracy versus 69.96% for the best baseline.

Significance. If the empirical claims withstand scrutiny, this work has practical and methodological value: it addresses a clinically important setting (stress monitoring for OUD) where target-population data are scarce, and it proposes a concrete mechanism (pruning-mask intersection over subjects) aimed at learning person-invariant features. Strengths of the manuscript include the collection of a new OUD dataset across pre- and post-medication states, the breadth of evaluation across seven datasets with varied stressors and environments, the use of linear mixed-effects models to motivate the design, extensive ablations, a runtime/scalability study, and a promise of released code and model checkpoints. The central claim is an empirical comparison on held-out OOD domains, so the result is not forced by construction. However, the evaluation as presented is not yet sufficient to establish 'consistent' superiority, because the reported margins rely on single person-disjoint splits with no variance estimates and the hyperparameter selection procedure is not fully transparent.

major comments (4)
  1. [Section 7.1.1, Table 3] The headline claim that HHISS 'consistently outperformed state-of-the-art baseline methods' rests on a single person-disjoint split with one run per method and no error bars, confidence intervals, or significance tests. The OOD test sets are small (14 pre-dose sessions and 15 post-dose sessions), so the reported 5-percentage-point margin over the best baseline (75.0% vs. 69.96%) is within plausible split-to-split and seed-to-seed variability. Please report results from multiple subject-disjoint splits (e.g., repeated random splits or subject-level cross-validation) with mean and standard deviation, and perform paired significance tests (e.g., Wilcoxon or bootstrap) across methods. This issue also affects Tables 4–8, where all numbers are single point estimates.
  2. [Appendix A.1] The hyperparameters (K=50%, T=70%, R=50, lambda=0.5, beta=0.3) are said to be fixed via grid search, but the manuscript does not describe the validation split used for this selection. If any of the OOD test sets (Pre-dose, Post-dose, WESAD, AffectiveROAD, Nurse Stress, OUD Daily) or the held-out Control-test set was used during model selection, the reported OOD numbers are optimistically biased and the comparison is no longer a clean OOD evaluation. Please specify exactly which data were used for hyperparameter tuning and, if OOD data were involved, re-run the evaluation with a strictly separated validation protocol.
  3. [Section 5.4, Eq. (4)] The continuous-label loss is defined as L_CN(f_pruned(X_i), f_over(X_i)), i.e., cross-entropy between the pruned network's logits and the over-parameterized model's logits. This is a form of knowledge distillation with soft targets. The text claims that HHISS 'entirely avoids temperature scaling' and 'does not mimic the behavior of the parent network', which is contradicted by the equation itself: the student is explicitly trained to reproduce the teacher's output distribution. Please clarify how this differs from standard knowledge distillation and whether the distinction is substantive, or revise the novelty claim accordingly.
  4. [Section 5.2.2, Algorithm 1] The core assumption of HHISS is that parameters lying outside the intersection of per-subject pruning masks carry 'varying importance across subjects and hence may introduce bias' and can be safely zeroed out. This is a load-bearing design decision, yet the paper provides only empirical support (Figure 12) showing that the intersection mask grows from 30% to 40% over rounds on one dataset. No analysis is given for how the size of the intersection—and the resulting invariant sub-network—depends on the number of training subjects, the pruning ratio K, or the heterogeneity of the training population. If the intersection becomes very small or unstable under different subject compositions, the method's OOD gains may not replicate. Please add sensitivity analyses or at least discuss conditions under which this assumption could fail.
minor comments (5)
  1. [Table 8] The MLDG row in Table 8 contains malformed numbers: '0.0.5743' for Accuracy and '0.0.5741' for Macro F1. Please correct these typos.
  2. [Section 7.1.6] The text says 'As shown in Table 6' when referring to the morning/noon/evening balanced accuracies; these results are displayed in Figure 6, not Table 6. Please correct the cross-reference.
  3. [Equation (3)] The importance score is written as '|∇L / ∇θ| |θ|', which is ambiguous and dimensionally odd; it should be written as |∂L/∂θ| · |θ| with a clear definition of the gradient with respect to each weight.
  4. [Algorithm 1, line 12] The inner-loop termination condition 'until Model performance higher than T' does not specify what metric or which data split the threshold T is applied to (training accuracy? validation balanced accuracy?). This makes the algorithm under-specified; please define T precisely.
  5. [Section 6.2] The evaluation setup says '32 subjects were randomly selected for training', but no random seed or repeated selection procedure is described. Even apart from the need for error bars, please report the seed(s) used or otherwise describe how many different splits were considered.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HHISS's core OOD claim is tested on external held-out datasets and public benchmarks, not forced by construction.

full rationale

The derivation chain is not circular. HHISS's central claim—that training on healthy controls generalizes to out-of-distribution populations—is established by held-out person-disjoint evaluations on self-collected OUD pre-dose/post-dose sessions and on external public datasets (WESAD, AffectiveROAD, nurse work stress, OUD daily stress) that are not used for training or hyperparameter selection. The self-referential element is the continuous-label loss (Eq. 4), where the pruned model is trained against logits of the IRM-regularized over-parameterized model trained on the same source data; however, this is a regularization term, not the predicted quantity, and its effect is measured against ground-truth stress labels on external OOD test sets, so the reported accuracy is not forced by construction. The person-wise pruning intersection (Eqs. 2-3) operationally defines 'subject-invariant parameters,' but the paper does not equate that definition with the empirical OOD accuracy; instead, the assumption is tested through ablations, saliency analysis, and comparison with baselines. Minor self-citations ([110], [132]) support data-collection and normalization choices but are not load-bearing evidence for the generalization claim. Potential statistical concerns—single-split point estimates without error bars and grid-searched hyperparameters in Appendix A.1—are evaluation-validity issues, not circularity.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

HHISS introduces no new physical entities. Its reliance on hyperparameters and domain assumptions is typical for ML domain-generalization methods, but these are not derived from first principles.

free parameters (7)
  • pruning amount K = 50%
    Grid-searched; ablation shows best at 50% for HHISS.
  • sparse-to-sparse threshold T = 70%
    Set in Algorithm 1; model retrains until performance exceeds T.
  • training rounds R = 50
    Number of sparse-to-sparse iterations.
  • continuous-label weight lambda = 0.5
    Weight for L_Continuous in Eq. 4/5.
  • IRM penalty weight beta = 0.3
    Weight for IRM regularization term in Eq. 5.
  • learning rate = 0.0001
    Adam optimizer fixed learning rate.
  • hidden size = 128 (inconsistency: ablation says 256)
    Appendix A.1 sets 128 units; Section 7.2.3 states all methods best at 256.
assumptions (6)
  • domain assumption Invariant stress features exist across individuals and stressor types
    Core premise of domain generalization; invoked implicitly throughout Section 5.
  • domain assumption Each subject can be treated as a domain for IRM
    Section 5.1: 'we consider each subject to be a unique domain'.
  • domain assumption Gradient-times-weight importance score identifies parameters relevant for generalization
    Used in Eq. 3 and Algorithm 1; based on pruning literature, not proven for this setting.
  • domain assumption Continuous labels from the over-parameterized model yield a flatter loss landscape and better OOD generalization
    Section 5.4 cites [127, 141]. Empirical justification in Figure 13 is qualitative.
  • domain assumption EMA self-reports of stress are accurate labels
    Used to label the OUD Daily stress dataset (Section 3.3.1).
  • domain assumption First three windows of a calm session provide a valid baseline for change-score normalization
    Section 3.5, following [110].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Human Heterogeneity Invariant Stress Sensing." pith.science (2026). https://pith.science/paper/27IPTYQZ

@misc{pith2026250602256,
  author       = {Pith},
  title        = {Pith review of: Human Heterogeneity Invariant Stress Sensing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/27IPTYQZ}},
  note         = {Machine review of arXiv:2506.02256}
}
read the original abstract

Stress affects physical and mental health, and wearable devices have been widely used to detect daily stress through physiological signals. However, these signals vary due to factors such as individual differences and health conditions, making generalizing machine learning models difficult. To address these challenges, we present Human Heterogeneity Invariant Stress Sensing (HHISS), a domain generalization approach designed to find consistent patterns in stress signals by removing person-specific differences. This helps the model perform more accurately across new people, environments, and stress types not seen during training. Its novelty lies in proposing a novel technique called person-wise sub-network pruning intersection to focus on shared features across individuals, alongside preventing overfitting by leveraging continuous labels while training. The study focuses especially on people with opioid use disorder (OUD)-a group where stress responses can change dramatically depending on their time of daily medication taking. Since stress often triggers cravings, a model that can adapt well to these changes could support better OUD rehabilitation and recovery. We tested HHISS on seven different stress datasets-four of which we collected ourselves and three public ones. Four are from lab setups, one from a controlled real-world setting, driving, and two are from real-world in-the-wild field datasets without any constraints. This is the first study to evaluate how well a stress detection model works across such a wide range of data. Results show HHISS consistently outperformed state-of-the-art baseline methods, proving both effective and practical for real-world use. Ablation studies, empirical justifications, and runtime evaluations confirm HHISS's feasibility and scalability for mobile stress sensing in sensitive real-world applications.

Figures

Figures reproduced from arXiv: 2506.02256 by the authors.

Figure 1
Figure 1. Distribution of self-reported stress labels across different times of day during in-the-wild field data collection for two OUD participants of the OUD Daily stress dataset. 3.3.2 Nurses work Stress Dataset. [53]: This dataset investigates stress in a workplace setting influenced by various social, cultural, and psychological factors. It contains 1,250 hours of physiological data collected from 15 nurses. After each … view at source ↗
Figure 2
Figure 2. Approach Workflow 5.1 Causally Invariant Features Extraction through Initial IRM Regularization: 5.1.1 Challenge: Human physiological responses to stressors are heterogeneous across inter- [43, 54, 65] and intra- individuals [86]. To develop a robust model applicable across diverse scenarios, it is crucial to ensure the model focuses on invariant and consistent features or embedding generation [PITH_FULL_IMAGE:figu… view at source ↗
Figure 3
Figure 3. Subject wise importance score on the non-subject wise gradient based pruning. [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: 5-fold Subject Independent Cross-Validation [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Evaluate on off-the-shelf dataset when train on Control dataset [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: Balanced accuracy of HHISS model across different survey time windows [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Accuracy variations relative to HHISS when different components are removed. [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: Performance comparison of HHISS at different prune [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 10
Figure 10. Figure 10: Accuracy variations relative to HHISS when different input modality are removed. [PITH_FULL_IMAGE:figures/full_fig_p024_10.png]
Figure 12
Figure 12. Figure 12: Intersectional ‘Mask’ size across iterations. [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]
Figure 13
Figure 13. Figure 13: Losslandscape Comparison [PITH_FULL_IMAGE:figures/full_fig_p025_13.png]
Figure 14
Figure 14. Figure 14: Saliency Maps of Selected Features Across Datasets for different approaches. [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]
Figure 15
Figure 15. Figure 15: Embedding distribution of intermediate layers when trained with Control dataset. [PITH_FULL_IMAGE:figures/full_fig_p027_15.png]
Figure 16
Figure 16. Figure 16: Subject wise PCA visualization of raw features [PITH_FULL_IMAGE:figures/full_fig_p027_16.png]
Figure 17
Figure 17. Figure 17: Balanced accuracy on ERM, HHISS, and feature select. [PITH_FULL_IMAGE:figures/full_fig_p029_17.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

146 extracted references · 68 canonical work pages

  1. [1]

    Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, and Mario Marchand. 2014. Domain-adversarial neural networks. arXiv preprint arXiv:1412.4446 (2014)

    arXiv 2014

  2. [2]

    Mohamed Akrout, Amal Feriani, Faouzi Bellili, Amine Mezghani, and Ekram Hossain. 2023. Domain Generalization in Machine Learning Models for Wireless Communications: Concepts, State-of-the-Art, and Open Issues. IEEE Communications Surveys & Tutorials (2023)

    2023

  3. [3]

    Julaine Allan. 2019. Ethics and practice of research with people who use drugs. (2019)

    2019

  4. [4]

    Andrew P Allen, Paul J Kennedy, Samantha Dockray, John F Cryan, Timothy G Dinan, and Gerard Clarke. 2017. The trier social stress test: principles and practice. Neurobiology of stress 6 (2017), 113–126

    2017

  5. [5]

    Mohammed A Almazrouei, Ruth M Morgan, and Itiel E Dror. 2023. A method to induce stress in human subjects in online research environments. Behavior research methods 55, 5 (2023), 2575–2582

    2023

  6. [6]

    Emily Anderson and Lindsay McNair. 2018. Ethical issues in research involving participants with opioid use disorder. Therapeutic innovation & regulatory science 52, 3 (2018), 280–284

    2018

  7. [7]

    Martin Arjovsky, Léon Bottou, Ishaan Gulrajani, and David Lopez-Paz. 2019. Invariant risk minimization.arXiv preprint arXiv:1907.02893 (2019)

    arXiv 2019

  8. [8]

    Shannon Avery-Desmarais, Monika Schuler, Jennifer Viveiros, Mary McCurry, Mirinda Tyo, and Brianna Kauranen. 2024. Barriers and Facilitators to Recruitment of a Multifaceted Population of Opioid Use Disorder Community Stakeholders. Journal of Addictions Nursing 35, 3 (2024), 137–141

    2024

Show all 146 references
  1. [9]

    Sudie E Back, Daniel F Gros, Matthew Price, Steve LaRowe, Julianne Flanagan, Kathleen T Brady, Charles Davis, Maryanne Jaconis, and Jenna L McCauley. 2015. Laboratory-induced stress and craving among individuals with prescription opioid dependence. Drug and alcohol dependence ...

    2015

  2. [10]

    Wenxuan Bao, Tianxin Wei, Haohan Wang, and Jingrui He. 2024. Adaptive test-time personalization for federated learning. Advances in Neural Information Processing Systems 36 (2024)

    2024

  3. [11]

    Brian Bartoldson, Ari Morcos, Adrian Barbu, and Gordon Erlebacher. 2020. The generalization-stability tradeoff in neural network pruning. Advances in Neural Information Processing Systems 33 (2020), 20852–20864

    2020

  4. [12]

    Felix Beierle and Rüdiger Pryss. 2023. Automating the Development of Stress Detection Systems. In 2023 Congress in Computer Science, Computer Engineering, & Applied Computing (CSCE) . IEEE, 2694–2696

    2023

  5. [13]

    Atif Belal, Madhu Kiran, Jose Dolz, Louis-Antoine Blais-Morin, Eric Granger, et al. 2021. Knowledge distillation methods for efficient unsupervised adaptation across multiple domains. Image and Vision Computing 108 (2021), 104096

    2021

  6. [14]

    Brinnae Bent, Benjamin A Goldstein, Warren A Kibbe, and Jessilyn P Dunn. 2020. Investigating sources of inaccuracy in wearable optical heart rate sensors. NPJ digital medicine 3, 1 (2020), 18

    2020

  7. [15]

    André Bernardes, Ricardo Couceiro, Júlio Medeiros, Jorge Henriques, César Teixeira, Marco Simões, João Durães, Raul Barbosa, Henrique Madeira, and Paulo Carvalho. 2022. How reliable are ultra-short-term HRV measurements during cognitively demanding tasks? Sensors 22, 17 (2022), 6528

    2022

  8. [16]

    Marco Bolpagni, Susanna Pardini, Marco Dianti, and Silvia Gabrielli. 2024. Personalized Stress Detection Using Biosignals from Wearables: A Scoping Review. Sensors 24, 10 (2024), 3221. Human Heterogeneity Invariant Stress Sensing • 31

    2024

  9. [17]

    Wolfram Boucsein. 2012. Electrodermal activity. Springer Science & Business Media

    2012

  10. [18]

    Ruud M Buijs and Corbert G Van Eden. 2000. The integration of stress by the hypothalamus, amygdala and prefrontal cortex: balance between the autonomic nervous system and the neuroendocrine system. In Progress in brain research. Vol. 126. Elsevier, 117–132

    2000

  11. [19]

    Sara Campanella, Ayham Altaleb, Alberto Belli, Paola Pierleoni, and Lorenzo Palma. 2023. A method for stress detection using empatica E4 bracelet and machine-learning techniques. Sensors 23, 7 (2023), 3565

    2023

  12. [20]

    Joshua M Carlson, Gülce N Dikecligil, Tsafrir Greenberg, and Lilianne R Mujica-Parodi. 2012. Trait reappraisal is associated with resilience to acute psychological stress. Journal of research in personality 46, 5 (2012), 609–613

    2012

  13. [21]

    Stephanie Carreiro, Hua Fang, Jianying Zhang, Kelley Wittbold, Shicheng Weng, Rachel Mullins, David Smelson, and Edward W Boyer

  14. [22]

    Stephanie Carreiro, Kelley Wittbold, Premananda Indic, Hua Fang, Jianying Zhang, and Edward W Boyer. 2016. Wearable biosensors to detect physiologic change during opioid use. Journal of medical toxicology 12 (2016), 255–262

    2016

  15. [23]

    Junbum Cha, Sanghyuk Chun, Kyungjae Lee, Han-Cheol Cho, Seunghyun Park, Yunsung Lee, and Sungrae Park. 2021. Swad: Domain generalization by seeking flat minima. Advances in Neural Information Processing Systems 34 (2021), 22405–22418

    2021

  16. [24]

    Junyi Chai, Taeuk Jang, and Xiaoqian Wang. 2022. Fairness without demographics through knowledge distillation. Advances in Neural Information Processing Systems 35 (2022), 19152–19164

    2022

  17. [25]

    Varun Chandra, Ankit Priyarup, and Divyashikha Sethia. 2021. Comparative study of physiological signals from empatica e4 wristband for stress classification. In Advances in Computing and Data Sciences: 5th International Conference, ICACDS 2021, Nashik, India, April 23–24, 2021...

    2021

  18. [26]

    Brittany P Chapman, Bhanu Teja Gullapalli, Tauhidur Rahman, David Smelson, Edward W Boyer, and Stephanie Carreiro. 2022. Impact of individual and treatment characteristics on wearable sensor-based digital biomarkers of opioid use. NPJ Digital Medicine 5, 1 (2022), 123

    2022

  19. [27]

    Hongrong Cheng, Miao Zhang, and Javen Qinfeng Shi. 2024. A Survey on Deep Neural Network Pruning: Taxonomy, Comparison, Analysis, and Recommendations. IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    2024

  20. [28]

    Keerthi Kumar Chintha, Premananda Indic, Brittany Chapman, Edward W Boyer, and Stephanie Carreiro. 2018. Wearable biosensors to evaluate recurrent opioid toxicity after naloxone administration: a Hilbert transform approach. In Proceedings of the... Annual Hawaii International ...

    2018

  21. [29]

    Jiho Choi, Jun Seong Lee, Moonwook Ryu, Gyutae Hwang, Gyeongyeon Hwang, and Sang Jun Lee. 2022. Attention-LRCN: long-term recurrent convolutional network for stress detection from photoplethysmography. In 2022 IEEE International Symposium on Medical Measurements and Applicatio...

    2022

  22. [30]

    Samantha L Connolly and Lauren B Alloy. 2018. Negative event recall as a vulnerability for depression: Relationship between momentary stress-reactive rumination and memory for daily life stress. Clinical Psychological Science 6, 1 (2018), 32–47

    2018

  23. [31]

    Loes Crielaard, Mary Nicolaou, Alexia Sawyer, Rick Quax, and Karien Stronks. 2021. Understanding the impact of exposure to adverse socioeconomic conditions on chronic stress from a complexity science perspective. BMC medicine 19 (2021), 1–20

    2021

  24. [32]

    Xiaoliang Dai, Hongxu Yin, and Niraj K Jha. 2019. NeST: A neural network synthesis tool based on a grow-and-prune paradigm. IEEE Trans. Comput. 68, 10 (2019), 1487–1497

    2019

  25. [33]

    Matthew A Davis, Lewei A Lin, Haiyin Liu, and Brian D Sites. 2017. Prescription opioid use among adults with mental health disorders in the United States. The Journal of the American Board of Family Medicine 30, 4 (2017), 407–417

    2017

  26. [34]

    Mario De La Rosa, Rosa Babino, Adelaida Rosario, Natalia Valiente Martinez, and Lubna Aijaz. 2012. Challenges and strategies in recruiting, interviewing, and retaining recent Latino immigrants in substance abuse and HIV epidemiologic studies. The American Journal on Addictions...

    2012

  27. [35]

    Maciej Dzieżyc, Martin Gjoreski, Przemysław Kazienko, Stanisław Saganowski, and Matjaž Gams. 2020. Can we ditch feature engineering? end-to-end deep learning for affect recognition from physiological sensor data. Sensors 20, 22 (2020), 6535

    2020

  28. [36]

    Scott M Fishman, Barth Wilsey, Jane Yang, Gary M Reisfield, Tara B Bandman, and David Borsook. 2000. Adherence monitoring and drug surveillance in chronic opioid therapy. Journal of pain and symptom management 20, 4 (2000), 293–307

    2000

  29. [37]

    Andreas Foltyn and Jessica Deuschel. 2021. Towards reliable multimodal stress detection under distribution shift. In Companion Publication of the 2021 International Conference on Multimodal Interaction . 329–333

    2021

  30. [38]

    Simon Föll, Martin Maritsch, Federica Spinola, Varun Mishra, Filipe Barata, Tobias Kowatsch, Elgar Fleisch, and Felix Wortmann

  31. [39]

    Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, Mario March, and Victor Lempitsky. 2016. Domain-adversarial training of neural networks. Journal of machine learning research 17, 59 (2016), 1–35

    2016

  32. [40]

    Shruti Gedam and Sanchita Paul. 2021. A review on mental stress detection using wearable sensors and machine learning techniques. IEEE Access 9 (2021), 84045–84066. 32 • Yi Xiao, Harshit Sharma, Sawinder Kaur, Dessa Bergen-Cico, and Asif Salekin

    2021

  33. [41]

    Shadi Ghiasi, Alberto Greco, Riccardo Barbieri, Enzo Pasquale Scilingo, and Gaetano Valenza. 2020. Assessing autonomic function from electrodermal activity and heart rate variability during cold-pressor test and emotional challenge. Scientific reports 10, 1 (2020), 5406

    2020

  34. [42]

    Grace E Giles, Caroline R Mahoney, Tad T Brunyé, Holly A Taylor, and Robin B Kanarek. 2014. Stress effects on mood, HPA axis, and autonomic response: comparison of three psychosocial stress paradigms. PloS one 9, 12 (2014), e113618

    2014

  35. [43]

    Charles F Gillespie, Justine Phifer, Bekh Bradley, and Kerry J Ressler. 2009. Risk and resilience: genetic and environmental influences on development of the stress response. Depression and anxiety 26, 11 (2009), 984–992

    2009

  36. [44]

    Maria Luisa González Ramírez, Juan Pablo García Vázquez, Marcela D Rodríguez, Luis Alfredo Padilla-López, Gilberto Manuel Galindo- Aldana, and Daniel Cuevas-González. 2023. Wearables for Stress Management: A Scoping Review. In Healthcare, Vol. 11. MDPI, 2369

    2023

  37. [45]

    Eric Granger, Madhu Kiran, Marco Pedersoli, Louis-Antoine Blais-Morin, et al. 2021. Progressive gradient pruning for classification, detection and domain adaptation. In 2020 25th International Conference on Pattern Recognition (ICPR) . IEEE, 2795–2802

    2021

  38. [46]

    Bhanu Teja Gullapalli, Stephanie Carreiro, Brittany P Chapman, Deepak Ganesan, Jan Sjoquist, and Tauhidur Rahman. 2021. Opitrack: a wearable-based clinical opioid use tracker with temporal convolutional attention networks. Proceedings of the ACM on interactive, mobile, wearabl...

    2021

  39. [47]

    Bhanu Teja Gullapalli, Annamalai Natarajan, Gustavo A Angarita, Robert T Malison, Deepak Ganesan, and Tauhidur Rahman. 2019. On-body sensing of cocaine craving, euphoria and drug-seeking behavior using cardiac and respiratory signals. Proceedings of the ACM on Interactive, Mob...

    2019

  40. [48]

    Marcia C Haigis and Bruce A Yankner. 2010. The aging stress response. Molecular cell 40, 2 (2010), 333–344

    2010

  41. [49]

    Abid Haleem, Mohd Javaid, Ravi Pratap Singh, Rajiv Suman, and Shanay Rab. 2021. Biosensors applications in medical field: A brief review. Sensors International 2 (2021), 100100

    2021

  42. [50]

    Yunjo Han, Panyu Zhang, Minseo Park, and Uichin Lee. 2024. Systematic Evaluation of Personalized Deep Learning Models for Affect Recognition. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 8, 4 (2024), 1–35

    2024

  43. [51]

    Neska El Haouij, Jean-Michel Poggi, Sylvie Sevestre-Ghalila, Raja Ghozi, and Mériem Jaïdane. 2018. AffectiveROAD system and database to assess driver’s attention. In Proceedings of the 33rd Annual ACM Symposium on Applied Computing . 800–803

    2018

  44. [52]

    Judy Hoffman, Erik Rodner, Jeff Donahue, Brian Kulis, and Kate Saenko. 2014. Asymmetric and category invariant feature transforma- tions for domain adaptation. International journal of computer vision 109 (2014), 28–41

    2014

  45. [53]

    Seyedmajid Hosseini, Raju Gottumukkala, Satya Katragadda, Ravi Teja Bhupatiraju, Ziad Ashkar, Christoph W Borst, and Kenneth Cochran. 2022. A multimodal sensor dataset for continuous stress detection of nurses in a hospital. Scientific Data 9, 1 (2022), 255

    2022

  46. [54]

    Rebecca Jacoby, Keren Greenfeld Barsky, Tal Porat, Stav Harel, Tsipi Hanalis Miller, and Gil Goldzweig. 2021. Individual stress response patterns: Preliminary findings and possible implications. PLoS One 16, 8 (2021), e0255889

    2021

  47. [55]

    Dibyanshu Jaiswal, Shalini Mukhopadhyay, and Varsha Sharma. 2024. Tinystressnet: On-device stress assessment with wearable sensors on edge devices. In 2024 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Wor...

    2024

  48. [56]

    Junguang Jiang, Yang Shu, Jianmin Wang, and Mingsheng Long. 2022. Transferability in Deep Learning: A Survey. arXiv:2201.05867 [cs.LG]

    2022 arXiv

  49. [57]

    Tian Jin. 2022. On neural network pruning’s effect on generalization . Ph. D. Dissertation. Massachusetts Institute of Technology

    2022

  50. [58]

    Bo Fu Mingsheng Long Junguang Jiang, Baixu Chen. 2020. Transfer-Learning-library. https://github.com/thuml/Transfer-Learning- Library

    2020

  51. [59]

    Mintong Kang, Linyi Li, and Bo Li. 2023. Fashapley: Fast and approximated shapley based model pruning towards certifiably robust dnns. In 2023 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) . IEEE, 575–592

    2023

  52. [60]

    Clemens Kirschbaum, Karl-Martin Pirke, and Dirk H Hellhammer. 1993. The ‘Trier Social Stress Test’–a tool for investigating psychobiological stress responses in a laboratory setting. Neuropsychobiology 28, 1-2 (1993), 76–81

    1993

  53. [61]

    Gregory Koch, Richard Zemel, Ruslan Salakhutdinov, et al. 2015. Siamese neural networks for one-shot image recognition. In ICML deep learning workshop, Vol. 2. Lille, 1–30

    2015

  54. [62]

    Asking why

    Ethan Kross, David Gard, Patricia Deldin, Jessica Clifton, and Ozlem Ayduk. 2012. “Asking why” from a distance: Its cognitive and emotional consequences for people with major depressive disorder. Journal of abnormal psychology 121, 3 (2012), 559

    2012

  55. [63]

    David Krueger, Ethan Caballero, Joern-Henrik Jacobsen, Amy Zhang, Jonathan Binas, Dinghuai Zhang, Remi Le Priol, and Aaron Courville. 2021. Out-of-distribution generalization via risk extrapolation (rex). In International conference on machine learning . PMLR, 5815–5826

    2021

  56. [64]

    Tamara P Lambert, Michael Chan, Jesus Antonio Sanchez-Perez, Mohammad Nikbakht, David J Lin, Afra Nawar, Syed Khairul Bashar, Jacob P Kimball, Jonathan S Zia, Asim H Gazi, et al. 2024. A Comparison of Normalization Techniques for Individual Baseline-Free Estimation of Absolute...

    2024

  57. [65]

    D Lecic-Tosevski, O Vukovic, and J Stepanovic. 2011. Stress and personality. Psychiatriki 22, 4 (2011), 290–297

    2011

  58. [66]

    Alan I Leshner and Victor J Dzau. 2019. Medication-based treatment to address opioid use disorder. Jama 321, 21 (2019), 2071–2072. Human Heterogeneity Invariant Stress Sensing • 33

    2019

  59. [67]

    Da Li, Yongxin Yang, Yi-Zhe Song, and Timothy Hospedales. 2018. Learning to generalize: Meta-learning for domain generalization. In Proceedings of the AAAI conference on artificial intelligence , Vol. 32

    2018

  60. [68]

    Hao Li, Zheng Xu, Gavin Taylor, Christoph Studer, and Tom Goldstein. 2018. Visualizing the loss landscape of neural nets. Advances in neural information processing systems 31 (2018)

    2018

  61. [69]

    Russell Li and Zhandong Liu. 2020. Stress detection using deep neural networks. BMC Medical Informatics and Decision Making 20 (2020), 1–10

    2020

  62. [70]

    Tailin Liang, John Glossner, Lei Wang, Shaobo Shi, and Xiaotong Zhang. 2021. Pruning and quantization for deep neural network acceleration: A survey. Neurocomputing 461 (2021), 370–403

    2021

  63. [71]

    Jun Liu and Xi Liu. 2022. Energy-efficient allocation for multiple tasks in mobile edge computing. Journal of Cloud Computing 11, 1 (2022), 71

    2022

  64. [72]

    Shiwei Liu, Yuesong Tian, Tianlong Chen, and Li Shen. 2023. Don’t be so dense: Sparse-to-sparse gan training without sacrificing performance. International Journal of Computer Vision 131, 10 (2023), 2635–2648

    2023

  65. [73]

    Xue Liu, Weijie Xia, and Zhimiao Fan. 2020. A Deep Neural Network Pruning Method Based on Gradient L1-norm. In 2020 IEEE 6th International Conference on Computer and Communications (ICCC) . 2070–2074. https://doi.org/10.1109/ICCC51575.2020.9345039

    2020

  66. [74]

    Wang Lu, Jindong Wang, Yiqiang Chen, Sinno Jialin Pan, Chunyu Hu, and Xin Qin. 2022. Semantic-discriminative mixup for generalizable sensor-based cross-domain activity recognition. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, 2 (2022), 1–19

    2022

  67. [75]

    Jian-Hao Luo, Jianxin Wu, and Weiyao Lin. 2017. ThiNet: A Filter Level Pruning Method for Deep Neural Network Compression. In ICCV. 5058–5066

    2017

  68. [76]

    R Ross MacLean, Jessica L Armstrong, and Mehmet Sofuoglu. 2019. Stress and opioid use disorder: A systematic review. Addictive behaviors 98 (2019), 106010

    2019

  69. [77]

    Dominique Makowski, Tam Pham, Zen J Lau, Jan C Brammer, François Lespinasse, Hung Pham, Christopher Schölzel, and SH Annabel Chen. 2021. NeuroKit2: A Python toolbox for neurophysiological signal processing. Behavior research methods (2021), 1–8

    2021

  70. [78]

    Akhil Mathur, Anton Isopoussu, Nadia Berthouze, Nicholas D Lane, and Fahim Kawsar. 2019. Unsupervised domain adaptation for robust sensory systems. In Adjunct Proceedings of the 2019 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of th...

    2019

  71. [79]

    Maurizio Mauri, Valentina Magagnin, Pietro Cipresso, Luca Mainardi, Emery N Brown, Sergio Cerutti, Marco Villamira, and Riccardo Barbieri. 2010. Psychophysiological signals associated with affective states. In 2010 Annual International Conference of the IEEE Engineering in Med...

    2010

  72. [80]

    Alban Maxhuni. 2017. Managing the Scarcity of Monitoring Data through Machine Learning in Healthcare Domain . Ph. D. Dissertation. University of Trento

    2017

  73. [81]

    Cameron McCarthy, Nikhilesh Pradhan, Calum Redpath, and Andy Adler. 2016. Validation of the Empatica E4 wristband. In 2016 IEEE EMBS international student conference (ISC) . IEEE, 1–4

    2016

  74. [82]

    Mary E McCaul, George E Bigelow, Maxine L Stitzer, and Ira Liebson. 1982. Short-term effects of oral methadone in methadone maintenance subjects. Clinical Pharmacology & Therapeutics 31, 6 (1982), 753–761

    1982

  75. [83]

    Bruce S McEwen. 2008. Central effects of stress hormones in health and disease: Understanding the protective and damaging effects of stress and stress mediators. European journal of pharmacology 583, 2-3 (2008), 174–185

    2008

  76. [84]

    Lakmal Meegahapola, Hamza Hassoune, and Daniel Gatica-Perez. 2024. M3BAT: Unsupervised Domain Adaptation for Multimodal Mobile Sensing with Multi-Branch Adversarial Training. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 8, 2 (2024), 1–30

    2024

  77. [85]

    Luca Menghini, Evelyn Gianfranchi, Nicola Cellini, Elisabetta Patron, Mariaelena Tagliabue, and Michela Sarlo. 2019. Stressing the accuracy: Wrist-worn wearable sensor validation over different conditions. Psychophysiology 56, 11 (2019), e13441

    2019

  78. [86]

    in the wild

    Sujay Nagaraj, Sarah Goodday, Thomas Hartvigsen, Adrien Boch, Kopal Garg, Sindhu Gowda, Luca Foschini, Marzyeh Ghassemi, Stephen Friend, and Anna Goldenberg. 2023. Dissecting the heterogeneity of “in the wild” stress from multimodal sensor data. NPJ Digital Medicine 6, 1 (2023), 237

    2023

  79. [87]

    Engineering National Academies of Sciences, Medicine, et al. 2018. Medication-Assisted Treatment for Opioid Use Disorder. National Academies Press (US) (2018)

    2018

  80. [88]

    Wing WY Ng, Daniel S Yeung, Michael Firth, Eric CC Tsang, and Xi-Zhao Wang. 2008. Feature selection using localized generalization error for supervised classification problems using RBFNN. Pattern Recognition 41, 12 (2008), 3706–3719

    2008

  81. [89]

    Anallely Nguyen, Jing Wang, Kristin M Holland, Daniel C Ehlman, Laura E Welder, Kimberly D Miller, and Deborah M Stone. 2024. Trends in Drug Overdose Deaths by Intent and Drug Categories, United States, 1999–2022. American Journal of Public Health 114, 10 (2024), 1081–1085

    2024

  82. [90]

    Joshua Niemeijer and Jörg Peter Schäfer. 2022. Domain adaptation and generalization: A low-complexity approach. In 6th Annual Conference on Robot Learning . 34 • Yi Xiao, Harshit Sharma, Sawinder Kaur, Dessa Bergen-Cico, and Asif Salekin

    2022

  83. [91]

    Simon Ollander, Christelle Godin, Aurélie Campagne, and Sylvie Charbonnier. 2016. A comparison of wearable and stationary sensors for stress detection. In 2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC) . IEEE, 004362–004366

    2016

  84. [92]

    Marcello Passarelli, Laura Casetta, Luca Rizzi, and Raffaella Perrella. 2021. Responses to stress: investigating the role of gender, social relationships, and touch avoidance in Italy. International Journal of Environmental Research and Public Health 18, 2 (2021), 600

    2021

  85. [93]

    J Iván Pérez and M Pilar Matud. 2022. Gender, stress, and well-being in adulthood. Journal of Clinical Medicine 12, 1 (2022), 110

    2022

  86. [94]

    Joseph V Pergolizzi Jr, Robert B Raffa, and Melanie H Rosenblatt. 2020. Opioid withdrawal symptoms, a consequence of chronic opioid use and opioid use disorder: Current understanding and approaches to management. Journal of clinical pharmacy and therapeutics 45, 5 (2020), 892–903

    2020

  87. [95]

    Ming-Zher Poh, Tobias Loddenkemper, Claus Reinsberger, Nicholas C Swenson, Shubhi Goyal, Mangwe C Sabtala, Joseph R Madsen, and Rosalind W Picard. 2012. Convulsive seizure detection using a wrist-worn electrodermal activity and accelerometry biosensor. Epilepsia 53, 5 (2012), e93–e97

    2012

  88. [96]

    Laavanya Rachakonda, Saraju P Mohanty, Elias Kougianos, and Prabha Sundaravadivel. 2019. Stress-Lysis: A DNN-integrated edge device for stress level detection in the IoMT. IEEE Transactions on Consumer Electronics 65, 4 (2019), 474–483

    2019

  89. [97]

    Alan Ramponi and Barbara Plank. 2020. Neural unsupervised domain adaptation in NLP—a survey. arXiv preprint arXiv:2006.00632 (2020)

    2020 arXiv

  90. [98]

    Akshay Rangamani et al. 2020. Loss landscapes and generalization in neural networks: Theory and applications . Ph. D. Dissertation. Johns Hopkins University

    2020

  91. [99]

    Nikhil Reddy, Mahsa Baktashmotlagh, and Chetan Arora. 2024. Towards Domain-Aware Knowledge Distillation for Continual Model Generalization. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision . 696–707

    2024

  92. [100]

    Martha S Rosenthal and Bruce McKay. 2022. Drugs: mind, body, and society . Oxford University Press

    2022

  93. [101]

    Sebastian Ruder, Parsa Ghaffari, and John G Breslin. 2017. Knowledge adaptation: Teaching to adapt. arXiv preprint arXiv:1702.02052 (2017)

    2017 arXiv

  94. [102]

    Isma-Ilou Sadou, Seyed Morteza Nabavinejad, Zhonghai Lu, and Masoumeh Ebrahimi. 2022. Inference time reduction of deep neural networks on embedded devices: A case study. In 2022 25th Euromicro Conference on Digital System Design (DSD) . IEEE, 205–213

    2022

  95. [103]

    Aaqib Saeed, Tanir Ozcelebi, Johan Lukkien, Jan BF van Erp, and Stojan Trajanovski. 2018. Model adaptation and personalization for physiological stress detection. In 2018 IEEE 5th international conference on data science and advanced analytics (DSAA) . IEEE, 209–216

    2018

  96. [104]

    Shiori Sagawa, Pang Wei Koh, Tatsunori B Hashimoto, and Percy Liang. 2019. Distributionally robust neural networks for group shifts: On the importance of regularization for worst-case generalization. arXiv preprint arXiv:1911.08731 (2019)

    2019 arXiv

  97. [105]

    Antoine Sanner, Camila Gonzalez, and Anirban Mukhopadhyay. 2021. How reliable are out-of-distribution generalization methods for medical image segmentation?. In DAGM German Conference on Pattern Recognition . Springer, 604–617

    2021

  98. [106]

    Alexandre Schaefer, Frédéric Nils, Xavier Sanchez, and Pierre Philippot. 2010. Assessing the effectiveness of a large database of emotion-eliciting films: A new tool for emotion researchers. Cognition and emotion 24, 7 (2010), 1153–1172

    2010

  99. [107]

    Regina Franziska Schmid, Joachim Thomas, and Katrin Rentzsch. 2024. Individual differences in parasympathetic nervous system reactivity in response to everyday stress are associated with momentary emotional exhaustion. Scientific Reports 14, 1 (2024), 26662

    2024

  100. [108]

    Philip Schmidt, Attila Reiss, Robert Duerichen, Claus Marberger, and Kristof Van Laerhoven. 2018. Introducing wesad, a multimodal dataset for wearable stress and affect detection. In Proceedings of the 20th ACM international conference on multimodal interaction . 400–408

    2018

  101. [109]

    Andreas R Schwerdtfeger and Katharina Dick. 2019. Episodes of momentary resilience in daily life are associated with HRV reductions to stressful operations in firefighters: an ambulatory assessment approach using bayesian multilevel modeling. The Journal of Positive Psychology...

    2019

  102. [110]

    Harshit Sharma, Yi Xiao, Victoria Tumanova, and Asif Salekin. 2022. Psychophysiological Arousal in Young Children Who Stutter: An Interpretable AI Approach. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, 3 (2022), 1–32

    2022

  103. [111]

    Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2013. Deep inside convolutional networks: Visualising image classification models and saliency maps. arXiv preprint arXiv:1312.6034 (2013)

    2013 arXiv

  104. [112]

    Sidak Pal Singh and Dan Alistarh. 2020. Woodfisher: Efficient second-order approximation for neural network compression. Advances in Neural Information Processing Systems 33 (2020), 18098–18109

    2020

  105. [113]

    Saurabh Sonkusare, David Ahmedt-Aristizabal, Matthew J Aburn, Vinh Thai Nguyen, Tianji Pang, Sascha Frydman, Simon Denman, Clinton Fookes, Michael Breakspear, and Christine C Guo. 2019. Detecting changes in facial temperature induced by a sudden auditory stimulus based on deep...

    2019

  106. [114]

    Žiga Stržinar, Araceli Sanchis, Agapito Ledezma, Oscar Sipele, Boštjan Pregelj, and Igor Škrjanc. 2023. Stress detection using frequency spectrum analysis of wrist-measured electrodermal activity. Sensors 23, 2 (2023), 963

    2023

  107. [115]

    Substance Abuse and Mental Health Services Administration. 2023. Key Substance Use and Mental Health Indicators in the United States: Results from the 2022 National Survey on Drug Use and Health . HHS Publication No. PEP23-07-01-006, NSDUH Series H-58. Center for Behavioral He...

    2023

  108. [116]

    Sungho Suh, Vitor Fortes Rey, and Paul Lukowicz. 2022. Adversarial deep feature extraction network for user independent human activity recognition. In 2022 IEEE International Conference on Pervasive Computing and Communications (PerCom) . IEEE, 217–226

    2022

  109. [117]

    Benedek Szakonyi, István Vassányi, Edit Schumacher, and István Kósa. 2021. Efficient methods for acute stress detection using heart rate variability data from Ambient Assisted Living sensors. Biomedical engineering online 20 (2021), 1–19

    2021

  110. [118]

    Ali Tazarv, Sina Labbaf, Stephanie M Reich, Nikil Dutt, Amir M Rahmani, and Marco Levorato. 2021. Personalized stress monitoring using wearable sensors in everyday settings. In 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMB...

    2021

  111. [119]

    JHM Tulen, P Moleman, HG Van Steenis, and F Boomsma. 1989. Characterization of stress reactions to the Stroop Color Word Test. Pharmacology Biochemistry and Behavior 32, 1 (1989), 9–15

    1989

  112. [120]

    Victoria Tumanova and Nicole Backes. 2019. Autonomic nervous system response to speech production in stuttering and normally fluent preschool-age children. Journal of Speech, Language, and Hearing Research 62, 11 (2019), 4030–4044

    2019

  113. [121]

    Yiannis Varnava, Kiran Jakate, Richard Garnett, Dimitrios Androutsos, Pascal N Tyrrell, and April Khademi. 2025. Out-of-distribution generalization for segmentation of lymph node metastasis in breast cancer. Scientific Reports 15, 1 (2025), 1127

    2025

  114. [122]

    Gideon Vos, Kelly Trinh, Zoltan Sarnyai, and Mostafa Rahimi Azghadi. 2023. Generalizable machine learning for stress monitoring from wearable devices: A systematic literature review. International Journal of Medical Informatics 173 (2023), 105026

    2023

  115. [123]

    Haoxiang Wang, Haozhe Si, Bo Li, and Han Zhao. 2022. Provable domain generalization via invariant-feature subspace recovery. In International Conference on Machine Learning . PMLR, 23018–23033

    2022

  116. [124]

    Jindong Wang, Cuiling Lan, Chang Liu, Yidong Ouyang, Tao Qin, Wang Lu, Yiqiang Chen, Wenjun Zeng, and S Yu Philip. 2022. Generalizing to unseen domains: A survey on domain generalization. IEEE transactions on knowledge and data engineering 35, 8 (2022), 8052–8072

    2022

  117. [125]

    Lindsey Wang, Nora D Volkow, Nathan A Berger, Pamela B Davis, David C Kaelber, and Rong Xu. 2023. Cardiac and mortality outcome differences between methadone, buprenorphine and naltrexone prescriptions in patients with an opioid use disorder.Journal of Clinical Psychology 79, ...

    2023

  118. [126]

    Qianru Wang, Bin Guo, Lu Cheng, and Zhiwen Yu. 2023. sUrban: Stable Prediction for Unseen Urban Data from Location-based Sensors. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 7, 3 (2023), 1–20

    2023

  119. [127]

    Yufei Wang, Haoliang Li, Lap-pui Chau, and Alex C Kot. 2021. Embracing the dark knowledge: Domain generalization using regularized knowledge distillation. In Proceedings of the 29th ACM International Conference on Multimedia . 2595–2604

    2021

  120. [128]

    Zhiyuan Wang, Maria A Larrazabal, Mark Rucker, Emma R Toner, Katharine E Daniel, Shashwat Kumar, Mehdi Boukhechba, Bethany A Teachman, and Laura E Barnes. 2023. Detecting social contexts from mobile sensing indicators in virtual interactions with socially anxious individuals. ...

    2023

  121. [129]

    Zhao-Xin Wang, John X Zhang, Qiu-Lin Wu, Ning Liu, Xiao-Ping Hu, Raymond CK Chan, and Zhuang-Wei Xiao. 2010. Alterations in the processing of non-drug-related affective stimuli in abstinent heroin addicts. Neuroimage 49, 1 (2010), 971–976

    2010

  122. [130]

    Jeannette Weber, Peter Angerer, and Jennifer Apolinário-Hagen. 2022. Physiological reactions to acute stressors and subjective stress during daily life: A systematic review on ecological momentary assessment (EMA) studies. PloS one 17, 7 (2022), e0271996

    2022

  123. [131]

    Qianhui Wu, Huiqiang Jiang, Haonan Yin, Börje F Karlsson, and Chin-Yew Lin. 2022. Multi-level knowledge distillation for out-of- distribution detection in text. arXiv preprint arXiv:2211.11300 (2022)

    2022 arXiv

  124. [132]

    Yi Xiao, Harshit Sharma, Zhongyang Zhang, Dessa Bergen-Cico, Tauhidur Rahman, and Asif Salekin. 2024. Reading between the heat: Co-teaching body thermal signatures for non-intrusive stress detection. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Techno...

    2024

  125. [133]

    Huatao Xu, Pengfei Zhou, Rui Tan, and Mo Li. 2023. Practically Adopting Human Activity Recognition. In Proceedings of the 29th Annual International Conference on Mobile Computing and Networking . 1–15

    2023

  126. [134]

    Jian Xu and Shao-Lun Huang. 2023. A Joint Training-Calibration Framework for Test-Time Personalization with Label Shift in Federated Learning. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management . 4370–4374

    2023

  127. [135]

    Xuhai Xu, Xin Liu, Han Zhang, Weichen Wang, Subigya Nepal, Yasaman Sefidgar, Woosuk Seo, Kevin S Kuehn, Jeremy F Huckins, Margaret E Morris, et al. 2023. Globem: Cross-dataset generalization of longitudinal human behavior modeling. Proceedings of the ACM on Interactive, Mobile...

    2023

  128. [136]

    Yikai Yang, Nhan Duy Truong, Jason K Eshraghian, Christina Maher, Armin Nikpour, and Omid Kavehei. 2022. A multimodal AI system for out-of-distribution generalization of seizure identification. IEEE Journal of Biomedical and Health Informatics 26, 7 (2022), 3529–3538

    2022

  129. [137]

    Sofia Yfantidou, Marios Constantinides, Dimitris Spathis, Athena Vakali, Daniele Quercia, and Fahim Kawsar. 2023. Beyond accuracy: a critical review of fairness in machine learning for mobile and wearable computing. arXiv preprint arXiv:2303.15585 (2023)

    2023

  130. [138]

    Han Yu and Akane Sano. 2022. Semi-supervised learning and data augmentation in wearable-based momentary stress detection in the wild. arXiv preprint arXiv:2202.12935 (2022)

    2022 arXiv

  131. [139]

    Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. 2021. Understanding deep learning (still) requires rethinking generalization. Commun. ACM 64, 3 (2021), 107–115. 36 • Yi Xiao, Harshit Sharma, Sawinder Kaur, Dessa Bergen-Cico, and Asif Salekin

    2021

  132. [140]

    Panyu Zhang, Gyuwon Jung, Jumabek Alikhanov, Uzair Ahmed, and Uichin Lee. 2024. A Reproducible Stress Prediction Pipeline with Mobile Sensor Data. Proceedings of the ACM on interactive, mobile, wearable and ubiquitous technologies 8, 3 (2024), 1–35

    2024

  133. [141]

    Tianli Zhang, Mengqi Xue, Jiangtao Zhang, Haofei Zhang, Yu Wang, Lechao Cheng, Jie Song, and Mingli Song. 2023. Generalization Matters: Loss Minima Flattening via Parameter Hybridization for Efficient Online Knowledge Distillation. InProceedings of the IEEE/CVF Conference on C...

    2023

  134. [142]

    Yiran Zhao, Yujie Tao, Grace Le, Rui Maki, Alexander Adams, Pedro Lopes, and Tanzeem Choudhury. 2023. Affective Touch as Immediate and Passive Wearable Intervention. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, 4 (2023), 1–23

    2023

  135. [143]

    Xiao Zhou, Yong Lin, Weizhong Zhang, and Tong Zhang. 2022. Sparse invariant risk minimization. In International Conference on Machine Learning. PMLR, 27222–27244

    2022

  136. [144]

    Michael Zhu and Suyog Gupta. 2017. To prune, or not to prune: exploring the efficacy of pruning for model compression. arXiv:1710.01878 [stat.ML] A METHODOLOGICAL TRANSPARENCY & REPRODUCIBILITY APPENDIX To ensure reproducibility, we provide detail about the backbone model used...

    2017 arXiv

  137. [2015]

    Journal of medical systems 39 (2015), 1–8

    iMStrong: deployment of a biosensor system to detect cocaine use. Journal of medical systems 39 (2015), 1–8

    2015

  138. [2021]

    Computer Methods and Programs in Biomedicine (2021)

    FLIRT: A Feature Generation Toolkit for Wearable Data. Computer Methods and Programs in Biomedicine (2021). https: //doi.org/10.1016/j.cmpb.2021.106461

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.