Pith. sign in

REVIEW 4 major objections 6 minor 2 cited by

AttenGluco: Multimodal Transformer-Based Blood Glucose Forecasting on AI-READI Dataset

T0 review · 4 major / 6 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 AttenGluco forecasts blood glucose up to 60 minutes ahead with roughly 10% lower RMSE and 15% lower MAE than a multimodal LSTM baseline across all four AI-READI cohorts.

desk verdict Modest but real results on a new dataset, undermined by an internal contradiction in the per-horizon table. read the letter →

arxiv 2502.09919 v1 pith:DHWJI6JX submitted 2025-02-14 cs.LG cs.AI

Ebrahim Farahmand , Reza Rahimi Azghan , Nooshin Taheri Chatrudi , Eric Kim , Gautham Krishna Gudur , Edison Thomaz , Giulia Pedrielli , Pavan Turaga
show 1 more author
Hassan Ghasemzadeh
This is my paper
classification cs.LGcs.AI
keywords bloodglucoseforecastingtransformercross-attentionmulti-scaleattentioncontinuousmonitoringAI-READIdatasetmultimodaltimeseriestype2diabetes
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

AttenGluco is a transformer-based forecasting model that predicts blood glucose levels up to an hour ahead by combining continuous glucose monitor readings with two walking-activity signals: steps and intervals between walks. The paper's central claim is that its dual attention design, cross-attention to fuse differently sampled signals and multi-scale attention to catch long-range dependencies, reduces forecast error by about 10% in RMSE and 15% in MAE compared with a multimodal LSTM baseline across healthy, prediabetic, oral-medication, and insulin-treated cohorts in the AI-READI dataset. The paper also reports that the gain grows under cohort-wise fine-tuning, reaching about 12% RMSE reduction, and that AttenGluco degrades more gracefully when new cohorts are introduced. If the claim holds, the architecture supplies a concrete recipe for long-horizon glucose forecasting from wearable data.

What carries the argument

The identifying mechanism is a hybrid attention stack: cross-attention plus multi-scale attention. Cross-attention uses the glucose embedding as the query and each activity embedding as key and value, letting the model align irregularly sampled activity signals with glucose without explicit resampling. Multi-scale attention then repeats self-attention at temporal downsampling factors of 1, 2, and 4, upsampling and summing the branch outputs so that fine-grained details and long-range structure live in one representation. The transformer is encoder-only: it replaces the decoder with a linear head that outputs m future CGM values, each corresponding to one 5-minute interval.

What would settle it

Retrain AttenGluco with the input window set to 200, 400, and 600 minutes, and with heart-rate and stress-index streams added; if the RMSE advantage over the LSTM baseline shrinks, disappears, or reverses, the claimed benefit depends on window coverage and omitted activity channels rather than on the attention mechanism itself.

Watch

Extended reading notes

Core claim

On the paper's own terms, AttenGluco is the first blood-glucose forecasting model evaluated on the AI-READI dataset and is designed to solve two problems that hamper existing predictors: mismatched sampling rates between CGM and activity sensors, and the difficulty of retaining long-term temporal context. The model takes the last 400 minutes of CGM, walking steps, and walking intervals, embeds them, and runs a two-branch cross-attention in which glucose queries each activity stream; the fused representation then passes through three parallel self-attention branches operating at downsampling factors 1, 2, and 4, whose outputs are summed. Across isolated-subject, cohort-wise fine-tuning, and continual-learning scenarios, AttenGluco reports lower RMSE and MAE and higher correlation than a 1D-CNN plus LSTM baseline in every cohort, with RMSE reductions of roughly 8 to 13 percent depending on scenario and cohort. The model's most notable advantage appears at the 60-minute horizon, where the baseline's error grows sharply while AttenGluco's RMSE stays comparatively stable.

Load-bearing premise

The load-bearing premise is that two walking-related activity features (steps and walking intervals) plus a 400-minute input window capture enough of the physical-activity influence on glucose that the remaining accuracy gap over the LSTM baseline can be attributed to the attention architecture; if activity acts through other channels such as heart rate or stress, the reported gains may not transfer.

Editorial extensions

If this is right

  • If the reported gains generalize, 60-minute-ahead glucose forecasts become usable for proactive hypo-and-hyperglycemia alerts, because AttenGluco's long-horizon RMSE degrades far less than the baseline's.
  • The cohort-wise fine-tuning results imply that adding more subjects from a cohort continuously reduces AttenGluco's test error, so larger training sets should push accuracy further.
  • The forgetting analysis shows that both models lose accuracy on earlier cohorts when fine-tuned on new ones, but AttenGluco maintains lower absolute error throughout, making it a better starting point for continual-learning fixes.
  • The architecture offers a template for fusing any irregularly sampled physiological streams, not just glucose and activity, because cross-attention handles differing sampling rates without explicit resampling.

Reading between the lines

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

  • The paper documents forgetting but does not test mitigation; a natural extension is to add replay or regularization during cohort transitions and see whether AttenGluco can retain earlier cohorts while gaining new ones.
  • Because AI-READI also contains heart-rate-derived stress scores, adding that stream to the two activity features is a direct test of whether the two-feature activity representation captures all the relevant physical-activity influence.
  • The comparison is made against a single LSTM baseline, so testing against other transformer forecasters would clarify whether the gain comes from the hybrid attention design or from transformer scale in general.
  • The choice of downsampling factors 1, 2, and 4 suggests a connection to dilated or hierarchical time-series models; an ablation varying these factors would quantify how much accuracy is attributable to the specific scales chosen.
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 / 6 minor

Summary. The paper proposes AttenGluco, a Transformer-based framework for blood glucose forecasting that combines CGM data with two activity features (walking steps and walking intervals) using cross-attention and multi-scale attention. The authors evaluate on the AI-READI dataset across four subject cohorts (healthy, pre-T2DM, oral medication, insulin) in three scenarios: isolated subject training, cohort-wise fine-tuning, and a forgetting analysis. The headline claim is that AttenGluco outperforms a multimodal CNN-LSTM baseline by about 10% in RMSE and 15% in MAE, with stronger gains in the cohort-wise fine-tuning scenario.

Significance. If the reported gains hold, AttenGluco would be a useful contribution to multimodal glucose forecasting, especially for long-horizon prediction. The paper is the first to report blood glucose forecasting results on the AI-READI dataset, and the authors provide a public codebase. The isolated-subject experiment uses a clean 85/15 train-test split per subject, which supports the main claim in a non-circular way. The inclusion of per-cohort results, a prediction-horizon breakdown, and a forgetting analysis are commendable. However, the significance is tempered by internal contradictions in the horizon-level results, a circular evaluation protocol in the cohort-wise fine-tuning scenario, and the absence of uncertainty quantification and ablations.

major comments (4)
  1. [Table III and Conclusion] Table III directly contradicts the Conclusion's claim that AttenGluco 'consistently outperformed the baseline model' across prediction horizons. At PH=5 min, AttenGluco has higher RMSE than the baseline for Healthy (7.63 vs. 7.35), Pre-T2DM (8.70 vs. 7.94), and Oral (9.33 vs. 9.15), and is only marginally better for Insulin (11.94 vs. 12.11). The paper never states which prediction horizon or pooling rule produces the headline ~10% RMSE improvement in Tables I and II, and no per-horizon breakdown is provided for the isolated-subject scenario. The authors must specify the pooling rule, report per-horizon results for all scenarios, and either correct the 'consistently outperformed' claim or provide a granular analysis that resolves the discrepancy.
  2. [Section IV-B2, Table II] The cohort-wise fine-tuning evaluation is circular: the text states that 'each subject serving as both training and testing data.' Reporting RMSE on subjects whose data were used for training inflates the apparent improvement and does not measure generalization. This affects Table II and Fig. 3, and the statement that 'performance progressively improves' as more subjects are added. The authors should either add a held-out test set for this scenario, or clearly reframe Table II and Fig. 3 as training-fit diagnostics rather than predictive performance. The current framing overstates the benefit of the fine-tuning protocol.
  3. [Section IV-B, Tables I-III] The paper states that each model undergoes five independent training runs, but no standard deviation, confidence interval, or statistical significance test is reported for any metric. Without error bars, the reported differences (e.g., 18.04 vs. 16.05 in Table I) could be within run-to-run variability. Reporting means over five runs is insufficient; the authors should report variance and, where possible, paired significance tests, especially for the Table III PH=5 results where AttenGluco's deficit may or may not be meaningful.
  4. [Section III, Section IV-B] The paper attributes the performance gains to the cross-attention and multi-scale attention mechanisms, but no ablation study is provided. Replacing each attention component with alternatives, removing the activity features, or varying window length and downsampling factors are necessary to verify that the proposed architecture components are responsible for the improvements. Additionally, the input representation uses only walking steps and walking intervals even though the AI-READI dataset provides heart rate and stress index signals; the paper should justify this choice and ideally report sensitivity to the input feature set and window length (400 minutes).
minor comments (6)
  1. [Section IV-B] The phrase 'a baseline model consisting if a 1D-CNN and LSTM' contains a typo ('if' should be 'of').
  2. [Algorithm 1] The algorithm description says 'Multi-scale atention block' — 'atention' should be 'attention'.
  3. [Section III-B, Eq. (3)] The text describing the second cross-attention branch says XWI is used as 'both the key and the query'; based on Eq. (3), it should be 'key and value' with XG as the query. Please clarify.
  4. [Section IV-A] The cohort naming is inconsistent: 'pre T2DM', 'pre-T2DM', and 'Pre-T2DM' are used interchangeably. Please standardize.
  5. [Figure 4] The forgetting analysis in Fig. 4 is described only briefly; the exact evaluation protocol (which subjects are tested, which cohorts are trained, and what the error bars represent, if any) should be stated in the caption or text.
  6. [Section IV-B] The paper does not state how the error metrics (RMSE, MAE, correlation) in Tables I and II are aggregated across subjects and across prediction horizons. Please specify the aggregation procedure explicitly.

Circularity Check

1 steps flagged · score 4.0 of 10

Cohort-wise fine-tuning evaluation is in-sample; the central isolated-subject result is non-circular.

  1. fitted input called prediction [Section IV-B-2 (Cohort-Wise Fine-Tuning), Table II]
    "the model is first trained on one subject and then fine-tuned sequentially across the other subjects in the same category, with each subject serving as both training and testing data."

    Table II and the associated ~12-13% RMSE improvements are computed in a scenario where each tested subject's data is also used to train or fine-tune the model. The reported 'prediction' errors therefore reduce by construction to in-sample reconstruction errors rather than out-of-sample forecasts, making the claimed cohort-wise improvement circular. This does not invalidate the abstract's ~10% headline, which comes from the isolated-subject scenario with an 85/15 train/test split (Table I); it only means the cohort-wise fine-tuning comparison is not evidence of generalization.

full rationale

The central performance claim (about 10% RMSE and 15% MAE improvement over a multimodal LSTM) is supported by Table I, where each subject is trained on 85% of their data and tested on the held-out 15%; that split is clean and the result is not circular. The secondary cohort-wise fine-tuning experiment, however, explicitly uses each subject as both training and testing data, making Table II and the related conclusions in-sample comparisons. The PH-wise Table III also shows that AttenGluco is worse than the baseline at 5 minutes in three of four cohorts, contradicting the text's claim of consistent outperformance, but that is a correctness/consistency issue rather than a circularity. No load-bearing self-citation chain or imported uniqueness theorem appears; self-citations are motivational or methodological and do not force the empirical results. Overall, one secondary evaluation is circular by construction, while the headline result rests on non-circular evidence, so the circularity score is modest.

Assumptions & free parameters 2 free parameters · 2 assumptions · 0 invented entities

The model introduces no new physical entities or functional forms beyond standard neural network components. The only hand-chosen settings are architectural hyperparameters, which are not fitted to data and are not the central claim.

free parameters (2)
  • Multi-scale downsampling factors
    Downsampling factors of 1, 2, and 4 in the multi-scale attention block are chosen by hand without ablation or sensitivity analysis.
  • Input window length = 400 minutes
    The sliding window of 6.66 hours is set without testing how performance varies with window length.
assumptions (2)
  • domain assumption CGM and walking step/interval data are sufficient predictors of future glucose
    The model uses only these inputs; other available signals from AI-READI (heart rate, stress index) are excluded, so predictive sufficiency is assumed but not demonstrated.
  • domain assumption The LSTM baseline is a representative state-of-the-art comparator
    The paper compares only against a 1D-CNN+LSTM and claims superiority, but does not benchmark against transformer-based glucose forecasters like Gluformer, which it cites.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AttenGluco: Multimodal Transformer-Based Blood Glucose Forecasting on AI-READI Dataset." pith.science (2026). https://pith.science/paper/DHWJI6JX

@misc{pith2026250209919,
  author       = {Pith},
  title        = {Pith review of: AttenGluco: Multimodal Transformer-Based Blood Glucose Forecasting on AI-READI Dataset},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DHWJI6JX}},
  note         = {Machine review of arXiv:2502.09919}
}
read the original abstract

Diabetes is a chronic metabolic disorder characterized by persistently high blood glucose levels (BGLs), leading to severe complications such as cardiovascular disease, neuropathy, and retinopathy. Predicting BGLs enables patients to maintain glucose levels within a safe range and allows caregivers to take proactive measures through lifestyle modifications. Continuous Glucose Monitoring (CGM) systems provide real-time tracking, offering a valuable tool for monitoring BGLs. However, accurately forecasting BGLs remains challenging due to fluctuations due to physical activity, diet, and other factors. Recent deep learning models show promise in improving BGL prediction. Nonetheless, forecasting BGLs accurately from multimodal, irregularly sampled data over long prediction horizons remains a challenging research problem. In this paper, we propose AttenGluco, a multimodal Transformer-based framework for long-term blood glucose prediction. AttenGluco employs cross-attention to effectively integrate CGM and activity data, addressing challenges in fusing data with different sampling rates. Moreover, it employs multi-scale attention to capture long-term dependencies in temporal data, enhancing forecasting accuracy. To evaluate the performance of AttenGluco, we conduct forecasting experiments on the recently released AIREADI dataset, analyzing its predictive accuracy across different subject cohorts including healthy individuals, people with prediabetes, and those with type 2 diabetes. Furthermore, we investigate its performance improvements and forgetting behavior as new cohorts are introduced. Our evaluations show that AttenGluco improves all error metrics, such as root mean square error (RMSE), mean absolute error (MAE), and correlation, compared to the multimodal LSTM model. AttenGluco outperforms this baseline model by about 10% and 15% in terms of RMSE and MAE, respectively.

Figures

Figures reproduced from arXiv: 2502.09919 by the authors.

Figure 2
Figure 2. The standard transformer architecture is typically made up of an encoder-decoder for data reconstruction. However, we modified this design for our forecasting model and framed it as a supervised learning task. Specifically, we eliminated the decoder and only utilized the encoder for data representation learning. Our customized transformer architecture incorpo￾rates two attention mechanisms: cross-attention and multi… view at source ↗
Figure 1
Figure 1. Overview of the AttenGluco framework including sensing module, data preparation, and forecasting model. + Output Embedding Linear Inputs Positional Encoding Output + US-4 + DS-n : Downsample with scale 1/n US-n : Upsample with scale n Multi-Head Attention1 Feed Forward DS-2 Q1 K1 V1 Add & Norm Multi-Scale Attention Feed Forward Add & Norm Cross Attention Multi-Head Attention2 Q2 K2 V2 Q5 K5 V5 Multi-Head Attention1 … view at source ↗
Figure 2
Figure 2. AttenGluco model architecture consists of cross-attention and multi-scale attention to forecast BGL. CA (XG, XWS, XWS) = [H1, . . . , HmH ]WCA H (2) Hh = Attention(XGWCA Q , XWSWCA K , XWSWCA V ) (3) Where WCA Q , WCA K , and WCA V are weight matrices specific to the attention head and belong to R dmodel×dmodel . Moreover, WCA H ∈ R (mH·dmodel)×dmodel is the final weight matrix that projects the concatenated attenti… view at source ↗
Figures from the paper (2 more)
Figure 3
Figure 3. Figure 3: demonstrates that as more subjects are added into each cohort, the model’s performance progressively improves in this scenario. This results in lower errors for newer subjects when used for testing. Notably, the reduction in test error is more significant in AttenGluco…
Figure 4
Figure 4. Figure 4: Fine-tuning the model on new cohorts leads to the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Gated Adaptation for Continual Learning in Human Activity Recognition

    cs.LG 2026-03 conditional novelty 5.0 of 10

    Channel-wise gating of frozen pretrained features reduces catastrophic forgetting in subject-incremental HAR, reaching ~78% final accuracy on PAMAP2 versus ~57% for full fine-tuning.

  2. Trustworthy AI in Digital Health: A Comprehensive Review of Robustness and Explainability

    cs.AI 2026-08 conditional novelty 3.0 of 10

    A survey of robustness and explainability methods for digital health AI, proposing a taxonomy and illustrating known XAI tools, without new empirical or theoretical results.

Reference graph

Works this paper leans on

34 extracted references · 27 canonical work pages · cited by 2 Pith papers

  1. [1]

    Epidemiology of type 2 dia- betes—global burden of disease and forecasted trends,

    M. Abdul Basith Khan, M. J. Hashim, J. K. King, R. D. Goven- der, H. Mustafa, and J. Al Kaabi, “Epidemiology of type 2 dia- betes—global burden of disease and forecasted trends,” Journal of epidemiology and global health , vol. 10, no. 1, pp. 107–111, 2020

    work page 2020

  2. [2]

    Urgent action needed as global diabetes cases increase four-fold over past decades,

    World Health Organization, “Urgent action needed as global diabetes cases increase four-fold over past decades,” World Health Organization, News release, November 2024. [Online]. Available: https://www.who.int/

    work page 2024

  3. [3]

    Sedentary behavior and the biological hallmarks of aging,

    J. Raffin, P. de Souto Barreto, A. P. Le Traon, B. Vellas, M. Aubertin- Leheudre, and Y . Rolland, “Sedentary behavior and the biological hallmarks of aging,” Ageing Research Reviews , vol. 83, p. 101807, 2023

    work page 2023

  4. [4]

    Deep multitask learning by stacked long short-term memory for predicting personalized blood glucose concentration,

    M. M. H. Shuvo and S. K. Islam, “Deep multitask learning by stacked long short-term memory for predicting personalized blood glucose concentration,” IEEE Journal of Biomedical and Health Informatics , vol. 27, no. 3, pp. 1612–1623, 2023

    work page 2023

  5. [5]

    Empirical evaluation of gated recurrent neural networks on sequence modeling,

    J. Chung, C. Gulcehre, K. Cho, and Y . Bengio, “Empirical evaluation of gated recurrent neural networks on sequence modeling,” arXiv preprint arXiv:1412.3555, 2014

    arXiv 2014

  6. [6]

    Deep gated recurrent and convolutional network hybrid model for univariate time series classification,

    N. Elsayed, A. S, and M. Bayoumi, “Deep gated recurrent and convolutional network hybrid model for univariate time series classification,” International Journal of Advanced Computer Science and Applications , vol. 10, no. 5, 2019. [Online]. Available: http://dx.doi.org/10.14569/IJACSA.2019.0100582

    work page arXiv 2019

  7. [7]

    Optimizing Time Series Forecasting: A Comparative Study of Adam and Nesterov Accelerated Gradient on LSTM and GRU networks Using Stock Market data

    A. Makinde, “Optimizing time series forecasting: A comparative study of adam and nesterov accelerated gradient on lstm and gru networks using stock market data,” 2024. [Online]. Available: https://arxiv.org/abs/2410.01843

    work page Pith review arXiv 2024

  8. [8]

    Modeling type 1 diabetes progression using machine learning and single-cell transcriptomic measurements in human islets,

    A. R. Patil, J. Schug, C. Liu, D. Lahori, H. C. Descamps, A. Naji, K. H. Kaestner, R. B. Faryabi, and G. Vahedi, “Modeling type 1 diabetes progression using machine learning and single-cell transcriptomic measurements in human islets,” Cell Reports Medicine , vol. 5, no. 5, 2024

    work page 2024

Show all 34 references
  1. [9]

    A comprehensive survey of time series forecasting: Architectural diversity and open challenges,

    J. Kim, H. Kim, H. Kim, D. Lee, and S. Yoon, “A comprehensive survey of time series forecasting: Architectural diversity and open challenges,” arXiv preprint arXiv:2411.05793 , 2024

    2024 arXiv

  2. [10]

    Unlocking the power of lstm for long term time series forecasting,

    Y . Kong, Z. Wang, Y . Nie, T. Zhou, S. Zohren, Y . Liang, P. Sun, and Q. Wen, “Unlocking the power of lstm for long term time series forecasting,” arXiv preprint arXiv:2408.10006 , 2024

    2024 arXiv

  3. [11]

    Attention is all you need,

    A. Vaswani, “Attention is all you need,” Advances in Neural Informa- tion Processing Systems , 2017

    2017

  4. [12]

    Pathformer: Multi-scale transformers with adaptive pathways for time series forecasting,

    P. Chen, Y . Zhang, Y . Cheng, Y . Shu, Y . Wang, Q. Wen, B. Yang, and C. Guo, “Pathformer: Multi-scale transformers with adaptive pathways for time series forecasting,” arXiv preprint arXiv:2402.05956 , 2024

    2024 arXiv

  5. [13]

    Multi-resolution time-series transformer for long-term forecasting,

    Y . Zhang, L. Ma, S. Pal, Y . Zhang, and M. Coates, “Multi-resolution time-series transformer for long-term forecasting,” in International Conference on Artificial Intelligence and Statistics . PMLR, 2024, pp. 4222–4230

    2024

  6. [14]

    AI-READI: rethinking data collection, preparation and sharing for propelling AI-based discoveries in diabetes research and beyond,

    AI-READI Consortium (2024), “AI-READI: rethinking data collection, preparation and sharing for propelling AI-based discoveries in diabetes research and beyond,” Nature Metabolism, 2024. [Online]. Available: https://doi.org/10.1038/s42255-024-01165-x

    2024 doi

  7. [15]

    Flagship Dataset of Type 2 Diabetes from the AI-READI Project (2.0.0) [Data set],

    AI-READI Consortium, “Flagship Dataset of Type 2 Diabetes from the AI-READI Project (2.0.0) [Data set],” FAIRhub., 2024. [Online]. Available: https://doi.org/10.60775/fairhub.2

    2024 doi

  8. [16]

    Management of diabetes and hyperglycaemia in the hospital,

    F. J. Pasquel, M. C. Lansang, K. Dhatariya, and G. E. Umpierrez, “Management of diabetes and hyperglycaemia in the hospital,” The lancet Diabetes & endocrinology , vol. 9, no. 3, pp. 174–188, 2021

    2021

  9. [17]

    Artificial intelligence in diabetes management: advancements, opportunities, and challenges,

    Z. Guan, H. Li, R. Liu, C. Cai, Y . Liu, J. Li, X. Wang, S. Huang, L. Wu, D. Liu et al., “Artificial intelligence in diabetes management: advancements, opportunities, and challenges,” Cell Reports Medicine , 2023

    2023

  10. [18]

    The growing epidemic of diabetes mellitus,

    D. Lovic, A. Piperidou, I. Zografou, H. Grassos, A. Pittaras, and A. Manolis, “The growing epidemic of diabetes mellitus,” Current vascular pharmacology, vol. 18, no. 2, pp. 104–109, 2020

    2020

  11. [19]

    Benchmarking machine learning algorithms on blood glucose prediction for type i diabetes in comparison with classical time-series models,

    J. Xie and Q. Wang, “Benchmarking machine learning algorithms on blood glucose prediction for type i diabetes in comparison with classical time-series models,” IEEE Transactions on Biomedical En- gineering, vol. 67, no. 11, pp. 3101–3124, 2020

    2020

  12. [20]

    Machine learning techniques for hypoglycemia prediction: trends and challenges,

    O. Mujahid, I. Contreras, and J. Vehi, “Machine learning techniques for hypoglycemia prediction: trends and challenges,” Sensors, vol. 21, no. 2, p. 546, 2021

    2021

  13. [21]

    Personalized modeling and detection of moments of cannabis use in free-living environments,

    R. R. Azghan, N. C. Glodosky, R. K. Sah, C. Cuttler, R. McLaughlin, M. J. Cleveland, and H. Ghasemzadeh, “Personalized modeling and detection of moments of cannabis use in free-living environments,” in 2023 IEEE 19th International Conference on Body Sensor Networks (BSN). IEEE...

    2023

  14. [22]

    Multimodal time-series activity forecasting for adaptive lifestyle intervention design,

    A. Mamun, K. S. Leonard, M. P. Buman, and H. Ghasemzadeh, “Multimodal time-series activity forecasting for adaptive lifestyle intervention design,” in 2022 IEEE-EMBS International Conference on Wearable and Implantable Body Sensor Networks (BSN) . IEEE, 2022, pp. 1–4

    2022

  15. [23]

    Cudle: Learning under label scarcity to detect cannabis use in uncontrolled environments using wearables,

    R. R. Azghan, N. C. Glodosky, R. K. Sah, C. Cuttler, R. McLaughlin, M. J. Cleveland, and H. Ghasemzadeh, “Cudle: Learning under label scarcity to detect cannabis use in uncontrolled environments using wearables,” IEEE Sensors Journal , 2025

    2025

  16. [24]

    Wavelet-augmented self-supervised learning for accurate classifica- tion of cognitive workload,

    N. T. Chatrudi, W. Clegern, R. Hager, L. Nelson, and H. Ghasemzadeh, “Wavelet-augmented self-supervised learning for accurate classifica- tion of cognitive workload,” in 2024 IEEE 20th International Confer- ence on Body Sensor Networks (BSN) , 2024, pp. 1–4

    2024

  17. [25]

    Expediting rein- forcement learning by incorporating temporal causal information,

    J. Corazza, H. P. Aria, D. Neider, and Z. Xu, “Expediting rein- forcement learning by incorporating temporal causal information,” in Causal Representation Learning Workshop at NeurIPS 2023 , 2023

    2023

  18. [26]

    Distributed reinforcement learning for swarm systems with reward machines,

    S. M. Alsadat, N. Baharisangari, Y . Paliwal, and Z. Xu, “Distributed reinforcement learning for swarm systems with reward machines,” in 2024 American Control Conference (ACC) , 2024, pp. 33–38

    2024

  19. [27]

    Hybrid attention model using feature decomposition and knowledge distillation for glucose forecasting,

    E. Farahmand, S. B. Soumma, N. T. Chatrudi, and H. Ghasemzadeh, “Hybrid attention model using feature decomposition and knowledge distillation for glucose forecasting,” arXiv preprint arXiv:2411.10703, 2024

    2024 arXiv

  20. [28]

    Analyzing the performance of transformers for the prediction of the blood glucose level consider- ing imputation and smoothing,

    E. Acuna, R. Aparicio, and V . Palomino, “Analyzing the performance of transformers for the prediction of the blood glucose level consider- ing imputation and smoothing,” Big Data and Cognitive Computing , vol. 7, no. 1, p. 41, 2023

    2023

  21. [29]

    Gluformer: Transformer-based personalized glucose forecasting with uncertainty quantification,

    R. Sergazinov, M. Armandpour, and I. Gaynanova, “Gluformer: Transformer-based personalized glucose forecasting with uncertainty quantification,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

    2023

  22. [30]

    Edge- based temporal fusion transformer for multi-horizon blood glucose prediction,

    T. Zhu, T. Chen, L. Kuangt, J. Zeng, K. Li, and P. Georgiou, “Edge- based temporal fusion transformer for multi-horizon blood glucose prediction,” in 2023 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2023, pp. 1–5

    2023

  23. [31]

    Scaleformer: Iterative multi-scale refining transformers for time series forecasting,

    A. Shabani, A. Abdi, L. Meng, and T. Sylvain, “Scaleformer: Iterative multi-scale refining transformers for time series forecasting,” arXiv preprint arXiv:2206.04038, 2022

    2022 arXiv

  24. [32]

    Glysim: Modeling and simulating glycemic response for behavioral lifestyle interventions,

    A. Arefeen and H. Ghasemzadeh, “Glysim: Modeling and simulating glycemic response for behavioral lifestyle interventions,” in2023 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI). IEEE, 2023, pp. 1–5

    2023

  25. [33]

    Joint embedding of food photographs and blood glucose for improved calorie estimation,

    L. Zhang, S. Huang, A. Das, E. Do, N. Glantz, W. Bevier, R. Santiago, D. Kerr, R. Gutierrez-Osuna, and B. J. Mortazavi, “Joint embedding of food photographs and blood glucose for improved calorie estimation,” in 2023 IEEE EMBS International Conference on Biomedical and Health ...

    2023

  26. [34]

    Overcoming catastrophic forgetting in neural networks,

    J. Kirkpatrick, R. Pascanu, N. Rabinowitz, J. Veness, G. Desjardins, A. A. Rusu, K. Milan, J. Quan, T. Ramalho, A. Grabska-Barwinska et al., “Overcoming catastrophic forgetting in neural networks,” Pro- ceedings of the national academy of sciences , vol. 114, no. 13, pp. 3521–...

    2017

Pith tools

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