pith. sign in

arxiv: 2605.20751 · v1 · pith:TPIBUTKLnew · submitted 2026-05-20 · 💻 cs.LG · cs.AI· cs.SY· eess.SY

PACD-Net: Pseudo-Augmented Contrastive Distillation for Glycemic Control Estimation from SMBG

Canyu Lei , David Repaske , Jianxin Xie This is my paper

Pith reviewed 2026-05-21 06:39 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.SYeess.SY
keywords glycemic controlSMBGcontrastive distillationsparse observationstime in rangediabetes managementself-supervised learningknowledge distillation
0
0 comments X

The pith

A contrastive distillation method estimates time in, above, and below range from sparse self-monitored blood glucose readings.

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

The paper develops PACD-Net to estimate standard diabetes control metrics such as time in range, time above range, and time below range when only sparse, irregular finger-prick measurements are available. It generates pseudo-SMBG sequences that contain more frequent samples and treats those richer sequences as teachers that supervise a student model trained on the actual sparse inputs. Multi-view contrastive learning is added to keep the learned representations stable across different possible sampling patterns, and a hybrid transformer-CNN backbone extracts temporal structure from the irregular sequences. A sympathetic reader would care because continuous glucose monitors remain costly or inaccessible for many patients, so any reliable way to extract the same clinical metrics from ordinary self-monitoring could widen access to good glycemic assessment.

Core claim

PACD-Net is a self-supervised contrastive knowledge distillation framework that uses pseudo-SMBG samples with richer temporal coverage as teacher signals to guide learning from sparse observations; multi-view contrastive learning enforces representation consistency across diverse sampling patterns; the model adopts a hybrid Swin Transformer-CNN backbone to capture temporal dependencies in sparse SMBG sequences; experimental results show consistent outperformance over existing methods in estimating TAR, TIR, and TBR from real-world SMBG data together with improved accuracy, stability, and generalization under extremely sparse observation settings.

What carries the argument

Pseudo-augmented contrastive distillation that transfers knowledge from richer pseudo-SMBG teacher sequences to sparse student inputs while enforcing multi-view representation consistency.

If this is right

  • The method produces more accurate estimates of TAR, TIR, and TBR than prior approaches when only real-world sparse SMBG data are supplied.
  • Stability and generalization improve markedly once observation density falls to extremely low levels.
  • The resulting estimates can serve as a practical clinical aid for interpreting routine self-monitored readings.
  • The same pseudo-augmentation and contrastive structure offers a template for other tasks that must learn from sparse or irregularly sampled sensor streams.

Where Pith is reading between the lines

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

  • If the pseudo-sample generator itself stays unbiased across different patient populations, the framework could materially lower the number of patients who need continuous monitors for routine control assessment.
  • The same teacher-student plus multi-view consistency pattern could be tested on other irregularly sampled physiological signals such as heart-rate variability or activity counts from consumer wearables.
  • A controlled ablation that replaces the pseudo-teacher with real but still-limited additional SMBG would isolate how much of the reported gain truly comes from the augmentation step.

Load-bearing premise

Pseudo-SMBG samples with richer temporal coverage can serve as reliable teacher signals without introducing systematic bias or inaccuracies into the distillation process.

What would settle it

On a dataset of patients who wear both a continuous glucose monitor and perform sparse SMBG at the same times, compare the model's predicted TAR, TIR, and TBR directly against the same metrics computed from the continuous recordings; large systematic deviations would falsify the accuracy claim.

Figures

Figures reproduced from arXiv: 2605.20751 by Canyu Lei, David Repaske, Jianxin Xie.

Figure 1
Figure 1. Figure 1: AGP times-in-range “thermometer” (TAR/TIR/TBR). [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed PACD-Net architecture. [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The Architecture of the Swin-CRB backbone in PACD-Net. [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Scatter plots comparing predicted and ground truth glycemic metrics for different [PITH_FULL_IMAGE:figures/full_fig_p031_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Predicted vs. ground-truth AGP glycemic metric scatter plots for baseline, DPA [PITH_FULL_IMAGE:figures/full_fig_p035_5.png] view at source ↗
read the original abstract

Effective diabetes management requires continuous monitoring of glycemic levels. Clinically, glycemic control is assessed using metrics such as Time in Range (TIR), Time Below Range (TBR), and Time Above Range (TAR), typically derived from continuous glucose monitoring (CGM). However, many patients rely on self-monitoring of blood glucose (SMBG) due to the high cost and limited accessibility of CGM. Unlike CGM, SMBG provides sparse and irregular measurements, making accurate estimation of these metrics challenging. Conventional supervised learning approaches struggle under such sparsity, leading to poor generalization and unstable performance. To address this, we propose PACD-Net, a self-supervised contrastive knowledge distillation framework for estimating glycemic control from SMBG. Pseudo-SMBG samples with richer temporal coverage are used as teacher signals to guide learning from sparse observations. In addition, multi-view contrastive learning enforces representation consistency across diverse sampling patterns. The model adopts a hybrid Swin Transformer-CNN backbone to capture temporal dependencies in sparse SMBG sequences. Experimental results demonstrate that PACD-Net consistently outperforms existing methods in estimating TAR, TIR, and TBR from real-world SMBG data, achieving improved accuracy as well as enhanced stability and generalization under extremely sparse observation settings. The proposed framework provides a practical tool for clinical SMBG interpretation and offers a generalizable approach for learning from sparse and irregularly sampled sensor data in broader applications.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces PACD-Net, a self-supervised contrastive knowledge distillation framework for estimating glycemic control metrics (TAR, TIR, TBR) from sparse and irregular SMBG sequences. Pseudo-SMBG samples with richer temporal coverage act as teacher signals, multi-view contrastive learning enforces representation consistency across sampling patterns, and a hybrid Swin Transformer-CNN backbone captures temporal dependencies. The central claim is that this yields consistent outperformance over existing methods in accuracy, stability, and generalization on real-world SMBG data under extremely sparse settings.

Significance. If the performance claims hold after proper validation, the work addresses a clinically important gap in diabetes management where CGM is inaccessible. The pseudo-augmented distillation plus multi-view contrastive objective offers a potentially generalizable strategy for sparse sensor data. The hybrid backbone choice is a reasonable fit for sequential data with irregular sampling.

major comments (2)
  1. [Abstract] Abstract: the claim of 'consistent outperformance' and 'improved accuracy as well as enhanced stability and generalization' is load-bearing for the paper's contribution, yet the abstract supplies no information on datasets, baseline implementations, statistical tests, error bars, or exclusion criteria. Without these, the central empirical claim cannot be evaluated.
  2. [Framework overview] Pseudo-SMBG teacher signal construction (described in the framework overview): the assumption that richer pseudo-samples serve as unbiased teachers for distillation from sparse observations is load-bearing for the stability and generalization claims. No held-out validation against CGM ground truth or ablation isolating the teacher-signal contribution from the Swin-CNN backbone and contrastive loss is referenced, leaving open the risk that distributional assumptions in pseudo-generation propagate rather than correct errors in TAR/TIR/TBR estimates.
minor comments (1)
  1. [Method] Notation for the multi-view contrastive objective could be clarified with an explicit equation showing how views are sampled and how the loss is aggregated.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address each major comment point by point below and have revised the manuscript to strengthen the presentation of our claims and experimental details.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'consistent outperformance' and 'improved accuracy as well as enhanced stability and generalization' is load-bearing for the paper's contribution, yet the abstract supplies no information on datasets, baseline implementations, statistical tests, error bars, or exclusion criteria. Without these, the central empirical claim cannot be evaluated.

    Authors: We agree that the abstract would benefit from additional context to allow readers to properly evaluate the empirical claims. In the revised manuscript we have expanded the abstract to explicitly reference the real-world SMBG datasets employed, the set of baseline methods against which PACD-Net was compared, the statistical tests (including paired t-tests with reported p-values) used to assess significance, and the presence of error bars along with the data exclusion criteria applied during preprocessing. revision: yes

  2. Referee: [Framework overview] Pseudo-SMBG teacher signal construction (described in the framework overview): the assumption that richer pseudo-samples serve as unbiased teachers for distillation from sparse observations is load-bearing for the stability and generalization claims. No held-out validation against CGM ground truth or ablation isolating the teacher-signal contribution from the Swin-CNN backbone and contrastive loss is referenced, leaving open the risk that distributional assumptions in pseudo-generation propagate rather than correct errors in TAR/TIR/TBR estimates.

    Authors: We appreciate the referee drawing attention to the importance of validating the pseudo-SMBG teacher signal. The pseudo-samples are constructed by interpolating within the observed temporal distribution of each patient’s SMBG records rather than introducing external assumptions; we have now added an explicit ablation (new Table in Section 4.3) that isolates the distillation component by comparing the full PACD-Net against variants that remove the teacher signal while retaining the Swin-CNN backbone and contrastive loss. Regarding held-out CGM validation, the primary datasets consist of sparse SMBG recordings without paired CGM; however, we have included a limited sensitivity analysis on a small subset of patients who later received CGM, showing that the pseudo-teacher improves rather than degrades alignment with the available CGM-derived metrics. We have clarified this limitation and the corresponding analysis in the revised text. revision: partial

Circularity Check

0 steps flagged

No significant circularity in PACD-Net derivation or claims

full rationale

The paper proposes a self-supervised contrastive knowledge distillation framework (PACD-Net) that introduces independent components including pseudo-SMBG teacher signals, multi-view contrastive objectives, and a hybrid Swin Transformer-CNN backbone. These elements are defined as novel contributions for handling sparse SMBG data and do not reduce by construction to fitted parameters, target metrics (TAR/TIR/TBR), or self-citations. The central claims rest on experimental results from real-world data rather than any self-definitional loop or imported uniqueness theorem. No load-bearing step equates the output estimates to the inputs via definition or prior author work alone; the method remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; full paper may contain additional model-specific hyperparameters or training assumptions not visible here.

axioms (1)
  • domain assumption Pseudo-SMBG samples with richer temporal coverage provide reliable teacher signals for distillation from sparse observations
    Invoked to justify the self-supervised guidance mechanism in the framework description.

pith-pipeline@v0.9.0 · 5794 in / 1343 out tokens · 45781 ms · 2026-05-21T06:39:08.850826+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

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

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

Reference graph

Works this paper leans on

67 extracted references · 67 canonical work pages

  1. [1]

    Diabetes Care , volume=

    Diabetes and multiple long-term conditions: a review of our current global health challenge , author=. Diabetes Care , volume=. 2023 , publisher=

    work page 2023

  2. [2]

    Current diabetes reviews , volume=

    Chronic complications of diabetes mellitus: a mini review , author=. Current diabetes reviews , volume=. 2017 , publisher=

    work page 2017

  3. [3]

    Nature Reviews Endocrinology , volume=

    The burden and risks of emerging complications of diabetes mellitus , author=. Nature Reviews Endocrinology , volume=. 2022 , publisher=

    work page 2022

  4. [4]

    Physical therapy , volume=

    Epidemiology of diabetes and diabetes-related complications , author=. Physical therapy , volume=. 2008 , publisher=

    work page 2008

  5. [5]

    , author=

    Complications of diabetes mellitus: A review. , author=. Drug Invention Today , volume=

    work page

  6. [6]

    Handbook of global health , pages=

    Global burden of diabetes mellitus , author=. Handbook of global health , pages=. 2021 , publisher=

    work page 2021

  7. [7]

    European Journal of Cardiovascular Prevention & Rehabilitation , volume=

    The global burden of diabetes and its complications: an emerging pandemic , author=. European Journal of Cardiovascular Prevention & Rehabilitation , volume=. 2010 , publisher=

    work page 2010

  8. [8]

    The American journal of medicine , volume=

    Glycemic control and complications in type 2 diabetes mellitus , author=. The American journal of medicine , volume=. 2010 , publisher=

    work page 2010

  9. [9]

    Diabetes Care , volume=

    Relation of glycemic control to diabetic complications and health outcomes , author=. Diabetes Care , volume=. 1998 , publisher=

    work page 1998

  10. [10]

    Clinical diabetes , volume=

    Effects of glycemic control on diabetes complications and on the prevention of diabetes , author=. Clinical diabetes , volume=. 2004 , publisher=

    work page 2004

  11. [11]

    2020 , publisher=

    The role of blood glucose monitoring in diabetes management , author=. 2020 , publisher=

    work page 2020

  12. [12]

    Journal of Diabetes and its Complications , volume=

    Role of continuous glucose monitoring for type 2 in diabetes management and research , author=. Journal of Diabetes and its Complications , volume=. 2017 , publisher=

    work page 2017

  13. [13]

    Diabetes & metabolism journal , volume=

    Continuous glucose monitoring sensors for diabetes management: a review of technologies and applications , author=. Diabetes & metabolism journal , volume=

    work page

  14. [14]

    Critical Care , volume=

    Clinical review: consensus recommendations on measurement of blood glucose and reporting glycemic control in critically ill adults , author=. Critical Care , volume=. 2013 , publisher=

    work page 2013

  15. [15]

    Diabetes Spectrum , volume=

    Clinical application of time in range and other metrics , author=. Diabetes Spectrum , volume=. 2021 , publisher=

    work page 2021

  16. [16]

    Systematic reviews , volume=

    Comparing effects of continuous glucose monitoring systems (CGMs) and self-monitoring of blood glucose (SMBG) amongst adults with type 2 diabetes mellitus: a systematic review protocol , author=. Systematic reviews , volume=. 2020 , publisher=

    work page 2020

  17. [17]

    Journal of Diabetes Science and Technology , volume=

    The use of continuous glucose monitoring in comparison to self-monitoring of blood glucose in gestational diabetes: a systematic review , author=. Journal of Diabetes Science and Technology , volume=. 2026 , publisher=

    work page 2026

  18. [18]

    Diabetes Therapy , volume=

    Ambulatory glucose profile (AGP) report in daily care of patients with diabetes: practical tips and recommendations , author=. Diabetes Therapy , volume=. 2022 , publisher=

    work page 2022

  19. [19]

    Children , volume=

    Exploring the continuous glucose monitoring in pediatric diabetes: current practices, innovative metrics, and future implications , author=. Children , volume=. 2024 , publisher=

    work page 2024

  20. [20]

    Diabetes care , volume=

    Effects of continuous glucose monitoring on metrics of glycemic control in diabetes: a systematic review with meta-analysis of randomized controlled trials , author=. Diabetes care , volume=. 2020 , publisher=

    work page 2020

  21. [21]

    Diabetes technology & therapeutics , volume=

    Continuous glucose monitoring: a review of successes, challenges, and opportunities , author=. Diabetes technology & therapeutics , volume=. 2016 , publisher=

    work page 2016

  22. [22]

    Diabetes care , volume=

    The cost-effectiveness of continuous glucose monitoring in type 1 diabetes , author=. Diabetes care , volume=. 2010 , publisher=

    work page 2010

  23. [23]

    BMJ Public Health , volume=

    Availability, prices and affordability of self-monitoring blood glucose devices: Surveys in six low-income and middle-income countries , author=. BMJ Public Health , volume=. 2025 , publisher=

    work page 2025

  24. [24]

    Current diabetes reports , volume=

    Advances, challenges, and cost associated with continuous glucose monitor use in adolescents and young adults with type 1 diabetes , author=. Current diabetes reports , volume=. 2021 , publisher=

    work page 2021

  25. [25]

    Clinical diabetes , volume=

    Self-monitoring of blood glucose: the basics , author=. Clinical diabetes , volume=. 2002 , publisher=

    work page 2002

  26. [26]

    Journal of diabetes Science and Technology , volume=

    Interferences and limitations in blood glucose self-testing: an overview of the current knowledge , author=. Journal of diabetes Science and Technology , volume=. 2016 , publisher=

    work page 2016

  27. [27]

    Diabetes technology & therapeutics , volume=

    Self-monitoring of blood glucose (SMBG) in insulin-and non--insulin-using adults with diabetes: consensus recommendations for improving SMBG accuracy, utilization, and research , author=. Diabetes technology & therapeutics , volume=. 2008 , publisher=

    work page 2008

  28. [28]

    Diabetes Technology & Therapeutics , volume=

    Differences for percentage times in glycemic range between continuous glucose monitoring and capillary blood glucose monitoring in adults with type 1 diabetes: analysis of the REPLACE-BG dataset , author=. Diabetes Technology & Therapeutics , volume=. 2020 , publisher=

    work page 2020

  29. [29]

    Diabetes care , volume=

    REPLACE-BG: a randomized trial comparing continuous glucose monitoring with and without routine blood glucose monitoring in adults with well-controlled type 1 diabetes , author=. Diabetes care , volume=. 2017 , publisher=

    work page 2017

  30. [30]

    Diabetes Care , volume=

    Clinical utility of SMBG: recommendations on the use and reporting of SMBG in clinical research , author=. Diabetes Care , volume=. 2015 , publisher=

    work page 2015

  31. [31]

    Journal of Pharmacy Practice , volume=

    Self-monitoring blood glucose (SMBG): now and the future , author=. Journal of Pharmacy Practice , volume=. 2004 , publisher=

    work page 2004

  32. [32]

    Diabetes , volume=

    Mean amplitude of glycemic excursions, a measure of diabetic instability , author=. Diabetes , volume=. 1970 , publisher=

    work page 1970

  33. [33]

    N Engl J Med , volume=

    The Diabetes Control and Complications Trial Research Group , author=. N Engl J Med , volume=

    work page

  34. [34]

    Diabetes research and clinical practice , volume=

    Intensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study , author=. Diabetes research and clinical practice , volume=. 1995 , publisher=

    work page 1995

  35. [35]

    A new proposition of the assessment of current glucose control in diabetic patients , author=

    “J”-index. A new proposition of the assessment of current glucose control in diabetic patients , author=. Hormone and metabolic research , volume=. 1995 , publisher=

    work page 1995

  36. [36]

    Diabetes care , volume=

    Symmetrization of the blood glucose measurement scale and its applications , author=. Diabetes care , volume=. 1997 , publisher=

    work page 1997

  37. [37]

    Is it important? How to measure it? , author=

    Glycemic variability: the third component of the dysglycemia in diabetes. Is it important? How to measure it? , author=. Journal of diabetes science and technology , volume=. 2008 , publisher=

    work page 2008

  38. [38]

    Diabetologia , volume=

    Day-to-day variation of continuously monitored glycaemia: a further measure of diabetic instability , author=. Diabetologia , volume=. 1972 , publisher=

    work page 1972

  39. [39]

    Diabetes technology & therapeutics , volume=

    A novel approach to continuous glucose analysis utilizing glycemic variation , author=. Diabetes technology & therapeutics , volume=. 2005 , publisher=

    work page 2005

  40. [40]

    Diabetes care , volume=

    Standards of medical care in diabetes , author=. Diabetes care , volume=. 2005 , publisher=

    work page 2005

  41. [41]

    Diabetes care , volume=

    Standards of medical care in diabetes-2006 , author=. Diabetes care , volume=. 2006 , publisher=

    work page 2006

  42. [42]

    Procedia Computer Science , volume=

    Diabetes prediction using machine learning algorithms , author=. Procedia Computer Science , volume=. 2019 , publisher=

    work page 2019

  43. [43]

    Journal of Next-Gen Engineering Systems , year=

    Machine Learning And Artificial Intelligence in Diabetes Prediction And Management: A Comprehensive Review of Models , author=. Journal of Next-Gen Engineering Systems , year=

    work page

  44. [44]

    Current Diabetes Reports , volume=

    Artificial intelligence in current diabetes management and prediction , author=. Current Diabetes Reports , volume=. 2021 , publisher=

    work page 2021

  45. [45]

    Diabetology & Metabolic Syndrome , volume=

    Recent applications of machine learning and deep learning models in the prediction, diagnosis, and management of diabetes: a comprehensive review , author=. Diabetology & Metabolic Syndrome , volume=. 2022 , publisher=

    work page 2022

  46. [46]

    Journal of diabetes science and technology , volume=

    Hypoglycemia prediction using machine learning models for patients with type 2 diabetes , author=. Journal of diabetes science and technology , volume=. 2014 , publisher=

    work page 2014

  47. [47]

    Computer methods and programs in biomedicine , volume=

    Minimizing postprandial hypoglycemia in Type 1 diabetes patients using multiple insulin injections and capillary blood glucose self-monitoring with machine learning techniques , author=. Computer methods and programs in biomedicine , volume=. 2019 , publisher=

    work page 2019

  48. [48]

    JMIR mHealth and uHealth , volume=

    Development of a deep learning model for dynamic forecasting of blood glucose level for type 2 diabetes mellitus: secondary analysis of a randomized controlled trial , author=. JMIR mHealth and uHealth , volume=. 2019 , publisher=

    work page 2019

  49. [49]

    Journal of medical Internet research , volume=

    Data-driven blood glucose pattern classification and anomalies detection: machine-learning applications in type 1 diabetes , author=. Journal of medical Internet research , volume=. 2019 , publisher=

    work page 2019

  50. [50]

    Sensors , volume=

    Self-supervised contrastive learning for medical time series: A systematic review , author=. Sensors , volume=. 2023 , publisher=

    work page 2023

  51. [51]

    BMC medical informatics and decision making , volume=

    Representation learning for clinical time series prediction tasks in electronic health records , author=. BMC medical informatics and decision making , volume=. 2019 , publisher=

    work page 2019

  52. [52]

    ACM Transactions on Management Information Systems , volume=

    Time series prediction using deep learning methods in healthcare , author=. ACM Transactions on Management Information Systems , volume=. 2023 , publisher=

    work page 2023

  53. [53]

    IEEE transactions on biomedical circuits and systems , volume=

    Population-specific glucose prediction in diabetes care with transformer-based deep learning on the edge , author=. IEEE transactions on biomedical circuits and systems , volume=. 2024 , publisher=

    work page 2024

  54. [54]

    Machine Learning for Health (ML4H) , pages=

    TransEHR: Self-supervised transformer for clinical time series data , author=. Machine Learning for Health (ML4H) , pages=. 2023 , organization=

    work page 2023

  55. [55]

    ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=

    Self-supervised transformer for sparse and irregularly sampled multivariate clinical time-series , author=. ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=. 2022 , publisher=

    work page 2022

  56. [56]

    IEEE Journal of Biomedical and Health Informatics , volume=

    Multi-view integrative attention-based deep representation learning for irregular clinical time-series data , author=. IEEE Journal of Biomedical and Health Informatics , volume=. 2022 , publisher=

    work page 2022

  57. [57]

    International conference on machine learning , pages=

    Clocs: Contrastive learning of cardiac signals across space, time, and patients , author=. International conference on machine learning , pages=. 2021 , organization=

    work page 2021

  58. [58]

    Diabetes technology & therapeutics , volume=

    Utilizing the ambulatory glucose profile to standardize and implement continuous glucose monitoring in clinical practice , author=. Diabetes technology & therapeutics , volume=. 2019 , publisher=

    work page 2019

  59. [59]

    Diabetes Technology & Therapeutics , volume=

    The ambulatory glucose profile: opportunities for enhancement , author=. Diabetes Technology & Therapeutics , volume=. 2021 , publisher=

    work page 2021

  60. [60]

    Advances in neural information processing systems , volume=

    Attention is all you need , author=. Advances in neural information processing systems , volume=

    work page

  61. [61]

    Journal of Diabetes Science and Technology , volume=

    Optimizing display, analysis, interpretation and utility of self-monitoring of blood glucose (SMBG) data for management of patients with diabetes , author=. Journal of Diabetes Science and Technology , volume=. 2007 , publisher=

    work page 2007

  62. [62]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Swin transformer: Hierarchical vision transformer using shifted windows , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  63. [63]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Emerging properties in self-supervised vision transformers , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  64. [64]

    International conference on machine learning , pages=

    A simple framework for contrastive learning of visual representations , author=. International conference on machine learning , pages=. 2020 , organization=

    work page 2020

  65. [65]

    arXiv preprint arXiv:2510.06623 , year=

    DPA-Net: A Dual-Path Attention Neural Network for Inferring Glycemic Control Metrics from Self-Monitored Blood Glucose Data , author=. arXiv preprint arXiv:2510.06623 , year=

    work page arXiv

  66. [66]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Momentum contrast for unsupervised visual representation learning , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  67. [67]

    Advances in neural information processing systems , volume=

    Bootstrap your own latent-a new approach to self-supervised learning , author=. Advances in neural information processing systems , volume=

    work page