pith. sign in

arxiv: 2602.08916 · v2 · pith:MJOUGBBKnew · submitted 2026-02-09 · 💻 cs.SC · cs.ET· cs.LG

AMS-HD: Hyperdimensional Computing for Real-Time and Energy-Efficient Acute Mountain Sickness Detection

Abu Masum , Mehran Moghadam , M. Hassan Najafi , Bige Unluturk , Ulkuhan Guler , Beth A. Beidleman , Sercan Aygun This is my paper

Pith reviewed 2026-05-21 14:25 UTC · model grok-4.3

classification 💻 cs.SC cs.ETcs.LG
keywords hyperdimensional computingacute mountain sicknesswearable sensorsFPGA implementationenergy efficient monitoringSpO2heart ratereal-time detection
0
0 comments X

The pith

Hyperdimensional computing detects acute mountain sickness from wearable SpO2 and heart rate signals with accuracy matching SVM and MLP but far lower resource use.

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

The paper introduces AMS-HD, the first hyperdimensional computing framework for detecting acute mountain sickness using only SpO2 and heart rate data from standard wearables. It combines mutual information feature selection with hypervector encoding and positional projection to enable efficient classification in both bipolar form for mobiles and binary form for FPGAs. The method reaches up to 91 percent accuracy and 90 percent F1-score in binary tasks while matching or exceeding traditional baselines on external datasets. Its value lies in the drastic cuts to hardware needs and energy draw, needing just 60 bytes of memory and 2.5 milliseconds per inference on mobile devices. This supports continuous monitoring in settings where conventional machine learning would exceed battery or chip limits.

Core claim

AMS-HD is the first complete hyperdimensional computing framework for AMS detection that spans high-level bipolar computing for mobile platforms and low-level binary computing for FPGA and ASIC targets, integrating mutual information feature selection, hypervector encoding, and positional projection on wearable SpO2 and heart rate signals to deliver competitive accuracy with major reductions in resource consumption.

What carries the argument

Hyperdimensional computing that encodes selected physiological features into high-dimensional vectors for simple, efficient classification operations across mobile and hardware platforms.

If this is right

  • Real-time AMS monitoring becomes practical on low-power wearables and smartwatches without heavy computation.
  • FPGA implementations reduce LUT usage by 7.3 times and flip-flop usage by 5.8 times while using 3.9 times less power than MLP baselines.
  • Mobile deployments require only 1 percent battery per session and 2.50 milliseconds inference time, about 2 times lower energy than SVM.
  • Accuracy holds at up to 85 percent on external AMS-related datasets for both binary and multiclass tasks.
  • The approach offers a scalable hardware-aware alternative for continuous health monitoring in constrained environments.

Where Pith is reading between the lines

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

  • Similar encoding techniques could extend to other physiological monitoring tasks that rely on basic wearable signals.
  • The extreme memory efficiency suggests deployment on simpler microcontrollers for remote or long-duration expeditions.
  • Adding minimal extra signals might increase robustness without losing the efficiency gains.
  • Field trials during actual high-altitude ascents would test performance beyond lab-collected data.

Load-bearing premise

SpO2 and heart rate signals collected from standard wearables contain sufficient information to detect AMS reliably across varied individuals, ascent rates, and conditions without extra sensors or subject-specific calibration.

What would settle it

A study on a broad group of climbers showing accuracy falling below 70 percent under real ascent conditions with diverse rates and environments would indicate the signals lack enough discriminative power.

Figures

Figures reproduced from arXiv: 2602.08916 by Abu Masum, Beth A. Beidleman, Bige Unluturk, Mehran Moghadam, M. Hassan Najafi, Sercan Aygun, Ulkuhan Guler.

Figure 1
Figure 1. Figure 1: Overview of the proposed AMS-HD framework. The system provides a complete design from high-level bipolar computing (−1/+1) on mobile and embedded processors to low-level binary computing (logic 0/1) on hardware platforms such as ASICs and FPGAs. By integrating physiological signals with lightweight hyperdimensional operations, the framework enables always￾on, resource-efficient detection of acute mountain … view at source ↗
Figure 2
Figure 2. Figure 2: Overview of an HDC model: encoding, training, and classification [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Neuro-symbolic learning architectures and their corresponding encoding: (a) language text processing using n-gram encoding, and (b) image processing [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Proposed AMS-HD training pipeline: MI feature selection, positional encoding, hyperdimensional projection, and ⃝a pseudo- vs. ⃝b quasi-random binarization. classifier (Step-I in [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Feature importance scores calculated via MI, highlighting SpO [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: A real application on the Mobile Phone & Smart Watch, including the development framework. AMS: Acute Mountain Sickness case study. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Hardware design strategy for HV generation. (a) Feature HVs based on thermometer (unary) encoding, (b) Position HVs based on a pseudo-LFSR structure and MISR, (c) Hamming distance similarity of feature HVs, and (d) Hamming distance similarity of position HVs. the HDC classifier with binary computing on an ARM Cortex￾A9 SoC, where feature and position HVs are generated via FPGA design and encoding is comple… view at source ↗
Figure 8
Figure 8. Figure 8: Binary computing implementation of AMS-HD on PYNQ-Z2 (Zynq￾7020). The ARM Cortex-A9 executes data read/write, sensor operations, and feature preprocessing. The FPGA fabric implements on-chip feature/position HV generation, binding/bundling with accumulative single/few-shot learning to form class HVs, a pipelined popcount–threshold (POP++) stage, and similarity search, yielding the final AMS/No-AMS decision… view at source ↗
Figure 9
Figure 9. Figure 9: Ground truth and predicted AMS total scores across individuals using [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of Power Consumption, Inference Time, and Energy [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Comparison of Model Memory Usage, Execution Time, and Battery [PITH_FULL_IMAGE:figures/full_fig_p009_11.png] view at source ↗
read the original abstract

Objective: Acute mountain sickness (AMS) is the most prevalent altitude illness, affecting unacclimatized individuals ascending above 2,500 m and potentially escalating to life threatening cerebral or pulmonary edema. Conventional machine learning (ML) methods for AMS detection from wearable physiological signals often fail to meet real-time hardware efficiency requirements of continuous monitoring. Methods: We present AMS-HD, the first hyperdimensional computing (HDC)-based framework for real-time AMS detection, spanning high- level bipolar (-1/+1) computing for mobile platforms and low-level binary (0/1) computing for FPGA and ASIC targets. The framework integrates mutual information feature selection, hypervector encoding, and positional projection to enhance classification efficiency. Validation spans ARM, FPGA, and smartwatch-smartphone platforms using wearable-accessible SpO2 and heart rate signals. Results: AMS-HD matches or outperforms SVM and MLP baselines in both binary and multiclass classification, achieving up to 91% accuracy and 90% F1-score in binary classification, and up to 85% accuracy on external AMS-related datasets. On FPGA, AMS-HD reduces LUT and flip-flop usage by 7.3x and 5.8x, while consuming 3.9x less power than MLP. On mobile platforms, AMS-HD requires only 1% battery per session, 60 Bytes of memory, and 2.50 ms inference time--approximately 2x and more than 3x lower energy consumption than SVM and MLP. Conclusion: AMS-HD provides a scalable, hardware-aware alternative to conventional ML for real-time AMS monitoring, achieving competitive performance with substantially lower resource consumption. Significance: This work presents the first complete HDC framework for altitude sickness detection, bridging wearable inference and low-level hardware deployment for resource-constrained health monitoring.

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

3 major / 2 minor

Summary. The paper introduces AMS-HD, the first hyperdimensional computing (HDC) framework for real-time acute mountain sickness (AMS) detection from wearable SpO2 and heart rate signals. It combines mutual information feature selection, hypervector encoding, and positional projection, with implementations for high-level bipolar computing on mobile platforms and low-level binary computing on FPGA/ASIC. Validation on ARM, FPGA, and smartwatch-smartphone platforms claims that AMS-HD matches or exceeds SVM and MLP baselines, reaching up to 91% accuracy and 90% F1-score in binary classification, up to 85% accuracy on external datasets, while delivering 7.3x LUT and 5.8x flip-flop reductions and 3.9x lower power versus MLP on FPGA, plus 1% battery use, 60 Bytes memory, and 2.50 ms inference on mobile devices.

Significance. If the performance and efficiency claims prove robust under proper validation, the work would be significant as the first complete HDC pipeline for AMS monitoring, demonstrating practical hardware advantages for continuous wearable inference. The quantified resource savings and cross-platform deployment provide concrete evidence of HDC's suitability for resource-constrained health applications. However, the absence of dataset and validation details substantially weakens the ability to evaluate generalizability and the premise that SpO2+HR signals alone carry reliable discriminative information.

major comments (3)
  1. The manuscript reports headline performance figures (91% binary accuracy, 90% F1, 85% on external sets) but provides no information on dataset size, number of subjects, total recordings, ascent profiles, or ground-truth labeling method (e.g., Lake Louise score). This omission is load-bearing for the central claim of competitive or superior accuracy, as it prevents assessment of whether results reflect true signal utility or overfitting to a small/homogeneous cohort.
  2. No cross-validation procedure is described (e.g., subject-independent leave-one-subject-out versus pooled data), nor is any statistical testing or handling of class imbalance reported. These details are required to support the claims of outperforming SVM and MLP baselines in both binary and multiclass settings and the generalizability to external AMS-related datasets.
  3. The experimental design rests on the unexamined assumption that SpO2 and heart rate signals from standard wearables contain sufficient information for reliable AMS detection across individuals, ascent rates, and conditions without subject-specific calibration or additional sensors. No supporting analysis of signal quality, inter-subject variability, or failure cases is provided to substantiate this premise.
minor comments (2)
  1. The abstract states 'up to 85% accuracy on external AMS-related datasets' without naming the datasets or describing their relation to the primary study cohort, which reduces clarity of the cross-dataset claim.
  2. Notation for hypervector dimension and number of selected features is introduced but not consistently tied to the reported resource numbers (e.g., 60 Bytes memory); a table linking these parameters to measured LUT, power, and latency would improve reproducibility.

Simulated Author's Rebuttal

3 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, clarifying aspects of our experimental design and committing to revisions that strengthen the presentation of dataset details, validation procedures, and supporting analyses without altering the core contributions.

read point-by-point responses

  1. Referee: The manuscript reports headline performance figures (91% binary accuracy, 90% F1, 85% on external sets) but provides no information on dataset size, number of subjects, total recordings, ascent profiles, or ground-truth labeling method (e.g., Lake Louise score). This omission is load-bearing for the central claim of competitive or superior accuracy, as it prevents assessment of whether results reflect true signal utility or overfitting to a small/homogeneous cohort.

    Authors: We agree that these details are essential for readers to evaluate the robustness and potential limitations of the reported accuracies. The original manuscript focused primarily on the HDC methodology and hardware results, which led to the omission of expanded dataset metadata. In the revised version, we have added a dedicated subsection in the Methods section that specifies the dataset size, number of subjects, total recordings, ascent profiles, participant demographics, and ground-truth labeling via the Lake Louise score, along with the data collection protocol. This addition directly addresses concerns about cohort homogeneity and allows better assessment of generalizability. revision: yes

  2. Referee: No cross-validation procedure is described (e.g., subject-independent leave-one-subject-out versus pooled data), nor is any statistical testing or handling of class imbalance reported. These details are required to support the claims of outperforming SVM and MLP baselines in both binary and multiclass settings and the generalizability to external AMS-related datasets.

    Authors: We acknowledge the need for explicit validation details to substantiate performance claims. The revised manuscript now includes a clear description of the cross-validation approach as subject-independent leave-one-subject-out to ensure no data leakage and to test generalizability across individuals. We have also added information on class imbalance handling (via class-weighted training) and statistical comparisons (including p-values from appropriate tests such as McNemar's test) between AMS-HD and the SVM/MLP baselines for both binary and multiclass tasks. These clarifications support the reported results and external dataset evaluations. revision: yes

  3. Referee: The experimental design rests on the unexamined assumption that SpO2 and heart rate signals from standard wearables contain sufficient information for reliable AMS detection across individuals, ascent rates, and conditions without subject-specific calibration or additional sensors. No supporting analysis of signal quality, inter-subject variability, or failure cases is provided to substantiate this premise.

    Authors: We agree that additional supporting analysis would improve the manuscript. The revised version incorporates a new subsection analyzing signal quality (e.g., noise levels and artifact rates), quantifying inter-subject variability in SpO2 and HR responses, and examining representative failure cases where classification performance declined. While our results demonstrate that these signals alone can yield competitive accuracy in the tested scenarios, we have expanded the discussion to explicitly note limitations around the absence of subject-specific calibration and the potential benefits of additional sensors in future work. This provides a more balanced substantiation of the premise. revision: partial

Circularity Check

0 steps flagged

No circularity: performance metrics are direct experimental measurements

full rationale

The paper presents an empirical HDC framework for AMS detection using SpO2 and heart rate signals, with results consisting of measured accuracy (up to 91%), F1-score, resource usage reductions on FPGA, and energy/battery metrics on mobile platforms. These quantities are obtained from physical implementations and cross-dataset evaluations rather than being algebraically derived from the model's own fitted parameters or equations. No self-definitional steps, fitted-input predictions, or load-bearing self-citations appear in the described methods (mutual information selection, hypervector encoding, positional projection); the central claims remain independent of any reduction to inputs by construction. Dataset and validation details affect generalizability but do not create circularity in the reported derivation chain.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The framework rests on standard HDC assumptions and common signal-processing choices rather than new physical postulates; a modest number of implementation parameters are expected but not enumerated in the abstract.

free parameters (2)
  • hypervector dimension
    Typical HDC design choice that trades accuracy against memory and compute; value not stated in abstract.
  • number of selected features
    Mutual-information threshold or count is tuned to the SpO2/HR signals; exact count or cutoff not reported.
axioms (1)
  • domain assumption Hyperdimensional vectors with simple bundling and binding operations can separate AMS from non-AMS states in SpO2 and heart-rate data.
    Invoked by the choice of HDC as the core classifier.

pith-pipeline@v0.9.0 · 5908 in / 1544 out tokens · 41173 ms · 2026-05-21T14:25:32.797856+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

57 extracted references · 57 canonical work pages

  1. [1]

    AMS-HD: Acute Mountain Sickness Detection with Hyperdimensional Computing,

    A. K. M. Masum, R. Pradhananga, J. I. Schmidt, M. S. Moghadam, M. H. Najafi, B. Unluturk, U. Guler, and S. Aygun, “AMS-HD: Acute Mountain Sickness Detection with Hyperdimensional Computing,” in 2025 IEEE International Symposium on Circuits and Systems (ISCAS), 2025, pp. 1–5

    work page 2025

  2. [2]

    Ventilatory and pulmonary vascular response to hypoxia and sus- ceptibility to high altitude pulmonary oedema,

    E. Hohenhaus, A. Paul, R. McCullough, H. Kucherer, and P. Bartsch, “Ventilatory and pulmonary vascular response to hypoxia and sus- ceptibility to high altitude pulmonary oedema,”European Respiratory Journal, vol. 8, no. 11, pp. 1825–1833, 1995

    work page 1995

  3. [3]

    Respiratory monitoring: Current state of the art and future roads,

    I. Costanzo, D. Sen, L. Rhein, and U. Guler, “Respiratory monitoring: Current state of the art and future roads,”IEEE Reviews in Biomedical Engineering, vol. 15, pp. 103–121, 2022. 13

    work page 2022

  4. [4]

    Using machine learning to determine the correlation between physiological and environmental parameters and the induction of acute mountain sickness,

    C.-Y . Wei, P.-N. Chen, S.-S. Linet al., “Using machine learning to determine the correlation between physiological and environmental parameters and the induction of acute mountain sickness,” vol. 22, no. 5, p. 628, 2022

    work page 2022

  5. [5]

    W. Li, M. Zhang, Y . Hu, P. Shen, Z. Bai, C. Huangfu, Z. Ni, D. Sun, N. Wang, P. Zhanget al., “Acute mountain sickness prediction: a concerto of multidimensional phenotypic data and machine learning strategies in the framework of predictive, preventive, and personalized medicine,”EPMA Journal, pp. 1–20, 2025

    work page 2025

  6. [6]

    Recent advances in predicting acute mountain sickness: from multidimensional cohort studies to cutting-edge model applications,

    B. Wang, S. Chen, J. Song, D. Huang, and G. Xiao, “Recent advances in predicting acute mountain sickness: from multidimensional cohort studies to cutting-edge model applications,”Frontiers in Physiology, vol. 15, p. 1397280, 2024

    work page 2024

  7. [7]

    Machine learning-based prediction model for myocardial ischemia under high altitude exposure: a cohort study,

    Y . Chen, X. Zhang, Q. Ye, X. Zhang, N. Cao, S.-Y . Li, J. Yu, S.-T. Zhao, J. Zhang, X.-M. Xuet al., “Machine learning-based prediction model for myocardial ischemia under high altitude exposure: a cohort study,”Scientific Reports, vol. 14, no. 1, p. 686, 2024

    work page 2024

  8. [8]

    Hyperdimensional computing: An introduction to comput- ing in distributed representation with high-dimensional random vectors,

    P. Kanerva, “Hyperdimensional computing: An introduction to comput- ing in distributed representation with high-dimensional random vectors,” Cognitive Computation, vol. 1, no. 2, pp. 139–159, 2009

    work page 2009

  9. [9]

    Modeling dependencies in multiple parallel data streams with hyperdimensional computing,

    O. R ¨as¨anen and S. Kakouros, “Modeling dependencies in multiple parallel data streams with hyperdimensional computing,”IEEE Signal Processing Letters, vol. 21, no. 7, pp. 899–903, 2014

    work page 2014

  10. [10]

    A linear-time, optimization-free, and edge device-compatible hypervector encoding,

    S. Aygun, M. H. Najafi, and M. Imani, “A linear-time, optimization-free, and edge device-compatible hypervector encoding,” inDATE’23, 2023

    work page 2023

  11. [11]

    Sobol sequence optimization for hardware- efficient vector symbolic architectures,

    S. Aygun and M. H. Najafi, “Sobol sequence optimization for hardware- efficient vector symbolic architectures,”IEEE Trans. on Computer-Aided Design of Integrated Circuits and Systems, 2024

    work page 2024

  12. [12]

    All You Need is Unary: End-to-End Unary Bit-stream Processing in Hyperdimensional Computing,

    M. Moghadam, S. Aygun, F. S. Banitaba, and M. H. Najafi, “All You Need is Unary: End-to-End Unary Bit-stream Processing in Hyperdimensional Computing,” ser. ISLPED ’24, 2024, p. 1–6. [Online]. Available: https://doi.org/10.1145/3665314.3670834

    work page doi:10.1145/3665314.3670834 2024

  13. [13]

    uHD: Unary processing for lightweight and dynamic hyperdimensional computing,

    S. Aygun, M. S. Moghadam, and M. H. Najafi, “uHD: Unary processing for lightweight and dynamic hyperdimensional computing,” in2024 Design, Automation & Test in Europe Conference & Exhibition (DATE). IEEE, 2024, pp. 1–6

    work page 2024

  14. [14]

    ID-VSA: Independent and Dynamic Vector Symbolic Architecture for Energy- Efficient Edge Al,

    M. Moghadam, A. K. M. Masum, S. Aygun, and M. H. Najafi, “ID-VSA: Independent and Dynamic Vector Symbolic Architecture for Energy- Efficient Edge Al,” in2025 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED), 2025, pp. 1–7

    work page 2025

  15. [15]

    On-the-Fly Hadamard Hypervec- tor Processing for Efficient Hyperdimensional Computing,

    A. K. M. Masum, M. S. Moghadam, S. H. Moon, A. M. M. Ahmed, M. H. Najafi, D. Reis, and S. Aygun, “On-the-Fly Hadamard Hypervec- tor Processing for Efficient Hyperdimensional Computing,” inDesign Automation Conference (DAC), 2025

    work page 2025

  16. [16]

    Toward energy-efficient stochastic circuits using parallel sobol sequences,

    S. Liu and J. Han, “Toward energy-efficient stochastic circuits using parallel sobol sequences,”IEEE TVLSI, vol. 26, no. 7, pp. 1326–1339, 2018

    work page 2018

  17. [17]

    Robust Data Processing for Vector Symbolic Computing,

    M. Moghadam, A. K. M. Masum, S. Aygun, and M. H. Najafi, “Robust Data Processing for Vector Symbolic Computing,” inProceedings of the Great Lakes Symposium on VLSI 2025, ser. GLSVLSI ’25. New York, NY , USA: Association for Computing Machinery, 2025, p. 823–828. [Online]. Available: https://doi.org/10.1145/3716368.3735287

    work page doi:10.1145/3716368.3735287 2025

  18. [18]

    A survey on hyperdimensional computing aka vector symbolic architectures, part ii: Applications, cognitive models, and challenges,

    D. Kleyko, D. Rachkovskij, E. Osipov, and A. Rahimi, “A survey on hyperdimensional computing aka vector symbolic architectures, part ii: Applications, cognitive models, and challenges,” vol. 55, no. 9, 2023. [Online]. Available: https://doi.org/10.1145/3558000

    work page doi:10.1145/3558000 2023

  19. [19]

    Chapter 8 - hyperdimensional computing nanosystem: in-memory computing using monolithic 3d integration of rram and cnfet,

    A. Rahimi, T. F. Wu, H. Li, J. M. Rabaey, H.-S. P. Wong, M. M. Shulaker, and S. Mitra, “Chapter 8 - hyperdimensional computing nanosystem: in-memory computing using monolithic 3d integration of rram and cnfet,” inMemristive Devices for Brain-Inspired Computing, 2020

    work page 2020

  20. [20]

    A framework for collaborative learning in secure high- dimensional space,

    M. Imani, Y . Kim, S. Riazi, J. Messerly, P. Liu, F. Koushanfar, and T. Rosing, “A framework for collaborative learning in secure high- dimensional space,” inIEEE CLOUD, 2019, pp. 435–446

    work page 2019

  21. [21]

    Hardware optimizations of dense binary hyperdimensional computing: Rematerialization of hyper- vectors, binarized bundling, and combinational associative memory,

    M. Schmuck, L. Benini, and A. Rahimi, “Hardware optimizations of dense binary hyperdimensional computing: Rematerialization of hyper- vectors, binarized bundling, and combinational associative memory,”J. Emerg. Technol. Comput. Syst., vol. 15, no. 4, Oct. 2019

    work page 2019

  22. [22]

    Achieving software- equivalent accuracy for hyperdimensional computing with ferroelectric- based in-memory computing,

    A. Kazemi, F. M ¨uller, M. M. Sharifi, H. Errahmouni, G. Gerlach, T. K¨ampfe, M. Imani, X. S. Hu, and M. Niemier, “Achieving software- equivalent accuracy for hyperdimensional computing with ferroelectric- based in-memory computing,”Scientific Reports, vol. 12, no. 1, p. 19201, Nov 2022

    work page 2022

  23. [23]

    A robust and energy-efficient classifier using brain-inspired hyperdimensional computing,

    A. Rahimi, P. Kanerva, and J. M. Rabaey, “A robust and energy-efficient classifier using brain-inspired hyperdimensional computing,” inISLPED, 2016, pp. 64–69

    work page 2016

  24. [24]

    No-Multiplication Deterministic Hyperdimensional Encoding for Resource-Constrained Devices,

    M. S. Moghadam, S. Aygun, and M. H. Najafi, “No-Multiplication Deterministic Hyperdimensional Encoding for Resource-Constrained Devices,”IEEE Embedded Systems Letters, vol. 15, no. 4, pp. 210–213, 2023

    work page 2023

  25. [25]

    Assess- ment of acute mountain sickness: How to integrate the advantages of the lake louise score and the chinese ams score,

    Y . Wu, P. Li, Z. Zhong, J. Xie, S. Zhou, Y . Gao, and J. Chen, “Assess- ment of acute mountain sickness: How to integrate the advantages of the lake louise score and the chinese ams score,” 2020

    work page 2020

  26. [26]

    Establishing a prediction model of severe acute mountain sickness using machine learning of support vector machine recursive feature elimination,

    M. Yang, Y . Wu, X.-b. Yang, T. Liu, Y . Zhang, Y . Zhuo, Y . Luo, and N. Zhang, “Establishing a prediction model of severe acute mountain sickness using machine learning of support vector machine recursive feature elimination,” vol. 13, no. 1, p. 4633, 2023. [Online]. Available: https://doi.org/10.1038/s41598-023-31797-0

    work page doi:10.1038/s41598-023-31797-0 2023

  27. [27]

    The use of pulse oximetry in the assessment of acclimatization to high altitude,

    T. D ¨unnwald, R. Kienast, D. Niederseer, and M. Burtscher, “The use of pulse oximetry in the assessment of acclimatization to high altitude,” Sensors, vol. 21, no. 4, p. 1263, 2021

    work page 2021

  28. [28]

    A slow feature based lstm network for susceptibility assessment of acute mountain sickness with heterogeneous data,

    L. Wang, R. Xiao, J. Chen, L. Zhu, D. Shi, and J. Wang, “A slow feature based lstm network for susceptibility assessment of acute mountain sickness with heterogeneous data,”Biomedical Signal Processing and Control, vol. 80, p. 104355, 2023

    work page 2023

  29. [29]

    Event-triggered pseudo supervised meta learning for susceptibility assessment of acute mountain sickness,

    L. Wang, D. Shi, L. Zhu, and J. Wang, “Event-triggered pseudo supervised meta learning for susceptibility assessment of acute mountain sickness,” in2024 39th YAC

    work page

  30. [30]

    Lstm: A search space odyssey,

    K. Greff, R. K. Srivastava, J. Koutn ´ık, B. R. Steunebrink, and J. Schmid- huber, “Lstm: A search space odyssey,”IEEE transactions on neural networks and learning systems, vol. 28, no. 10, pp. 2222–2232, 2016

    work page 2016

  31. [31]

    Word2hypervec: From word embeddings to hypervectors for hyperdimensional comput- ing,

    A. G. Ayar, S. Aygun, M. H. Najafi, and M. Margala, “Word2hypervec: From word embeddings to hypervectors for hyperdimensional comput- ing,” inGreat Lakes Symposium on VLSI, 2024, p. 355–356

    work page 2024

  32. [32]

    Inflammatory gene expression during acute high-altitude exposure,

    K. Pham, S. Frost, K. Parikh, N. Puvvula, B. Oeung, and E. C. Heinrich, “Inflammatory gene expression during acute high-altitude exposure,”The Journal of physiology, vol. 600, no. 18, pp. 4169–4186, 2022

    work page 2022

  33. [33]

    Mutual information between discrete and continuous data sets,

    B. C. Ross, “Mutual information between discrete and continuous data sets,”PloS one, vol. 9, no. 2, p. e87357, 2014

    work page 2014

  34. [34]

    Performing stochastic computation deterministically,

    M. H. Najafi, D. Jenson, D. J. Lilja, and M. D. Riedel, “Performing stochastic computation deterministically,”IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 27, no. 12, 2019

    work page 2019

  35. [35]

    Higher-dimensional hadamard matrices,

    P. Shlichta, “Higher-dimensional hadamard matrices,”IEEE Transac- tions on Information Theory, vol. 25, no. 5, pp. 566–572, 1979

    work page 1979

  36. [36]

    Hadamard matrices and hadamard designs,

    R. Craigen and H. Kharaghani, “Hadamard matrices and hadamard designs,” inHandbook of combinatorial designs. Chapman and Hall/CRC, 2006, pp. 299–305

    work page 2006

  37. [37]

    K. J. Horadam,Hadamard matrices and their applications. Princeton university press, 2012

    work page 2012

  38. [38]

    Generalized hadamard matrices,

    A. T. Butson, “Generalized hadamard matrices,”Proceedings of the American Mathematical Society, vol. 13, no. 6, pp. 894–898, 1962

    work page 1962

  39. [39]

    Hadamard matrices and their applica- tions,

    A. Hedayat and W. D. Wallis, “Hadamard matrices and their applica- tions,”The annals of statistics, pp. 1184–1238, 1978

    work page 1978

  40. [40]

    Hadamard transform image coding,

    W. K. Pratt, J. Kane, and H. C. Andrews, “Hadamard transform image coding,”Proceedings of the IEEE, vol. 57, no. 1, pp. 58–68, 1969

    work page 1969

  41. [41]

    Low-cost sorting network circuits using unary processing,

    M. H. Najafi, D. J. Lilja, M. D. Riedel, and K. Bazargan, “Low-cost sorting network circuits using unary processing,”IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 26, no. 8, pp. 1471– 1480, 2018

    work page 2018

  42. [42]

    OTFGEncoder - HDC: Hardware-efficient Encoding Techniques for Hyperdimensional Com- puting,

    M. S. Roodsari, J. Krautter, and M. Tahoori, “OTFGEncoder - HDC: Hardware-efficient Encoding Techniques for Hyperdimensional Com- puting,” in2024 Design, Automation & Test in Europe Conference & Exhibition (DATE), 2024, pp. 1–2

    work page 2024

  43. [43]

    Aliasing in signature analysis testing with multiple input shift registers,

    M. Damiani, P. Olivo, M. Favalli, S. Ercolani, and B. Ricco, “Aliasing in signature analysis testing with multiple input shift registers,”IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 9, no. 12, pp. 1344–1353, 1990

    work page 1990

  44. [44]

    The relationship between anxiety and acute mountain sickness,

    C. J. Boos, M. Bass, J. P. O’Hara, E. Vincent, A. Mellor, L. Sevier, H. Abdul-Razakq, M. Cooke, M. Barlow, and D. R. Woods, “The relationship between anxiety and acute mountain sickness,”PLoS One, vol. 13, no. 6, p. e0197147, 2018

    work page 2018

  45. [45]

    Prevalence and knowledge about acute mountain sickness in the western alps,

    M. M. Berger, A. H ¨using, N. Niessen, L. M. Schiefer, M. Schneider, P. B¨artsch, and K.-H. J ¨ockel, “Prevalence and knowledge about acute mountain sickness in the western alps,”PLoS One, vol. 18, no. 9, p. e0291060, 2023

    work page 2023

  46. [46]

    tiny machine learning on android devices: Continuous health monitoring with wearables,

    C. Dupuis, A. K. M. Masum, M. H. Najafi, U. Guler, and S. Aygun, “tiny machine learning on android devices: Continuous health monitoring with wearables,” inIEEE International Midwest Symposium on Circuits and Systems (MWSCAS 2025), 2025

    work page 2025

  47. [47]

    E3HDC: En- ergy Efficient Encoding for Hyper-Dimensional Computing on Edge De- vices,

    M. S. Roodsari, J. Krautter, V . Meyers, and M. Tahoori, “E3HDC: En- ergy Efficient Encoding for Hyper-Dimensional Computing on Edge De- vices,” in2024 34th International Conference on Field-Programmable Logic and Applications (FPL), 2024, pp. 274–280. 14

    work page 2024

  48. [48]

    Hd2fpga: Automated framework for accelerating hyperdimen- sional computing on fpgas,

    T. Zhang, S. Salamat, B. Khaleghi, J. Morris, B. Aksanli, and T. S. Rosing, “Hd2fpga: Automated framework for accelerating hyperdimen- sional computing on fpgas,” in2023 24th International Symposium on Quality Electronic Design (ISQED), 2023, pp. 1–9

    work page 2023

  49. [49]

    F5-hd: Fast flexible fpga-based framework for refreshing hyperdimensional computing,

    S. Salamat, M. Imani, B. Khaleghi, and T. Rosing, “F5-hd: Fast flexible fpga-based framework for refreshing hyperdimensional computing,” in Proceedings of the 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays, ser. FPGA ’19. New York, NY , USA: Association for Computing Machinery, 2019, p. 53–62. [Online]. Available: https://doi.org...

    work page doi:10.1145/3289602.3293913 2019

  50. [50]

    Quanthd: A quantization framework for hyperdimen- sional computing,

    M. Imani, S. Bosch, S. Datta, S. Ramakrishna, S. Salamat, J. M. Rabaey, and T. Rosing, “Quanthd: A quantization framework for hyperdimen- sional computing,”IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 39, no. 10, pp. 2268–2278, 2020

    work page 2020

  51. [51]

    Smartwatch measurement of blood oxygen saturation for predicting acute mountain sickness: Diagnostic accuracy and reliability,

    Z. Zeng, L. Li, L. Hu, K. Wang, and L. Li, “Smartwatch measurement of blood oxygen saturation for predicting acute mountain sickness: Diagnostic accuracy and reliability,” vol. 10, p. 20552076241284910, 2024, pMCID: PMC11440541; eCollection 2024 Jan–Dec

    work page 2024

  52. [52]

    Smartwatch-based maximum oxygen consumption measurement for predicting acute mountain sickness: Diagnostic accuracy evaluation study,

    X. Ye, M. Sun, S. Yu, J. Yang, Z. Liu, H. Lv, B. Wu, J. He, X. Wang, and L. Huang, “Smartwatch-based maximum oxygen consumption measurement for predicting acute mountain sickness: Diagnostic accuracy evaluation study,”JMIR Mhealth Uhealth, vol. 11, p. e43340, Jul 2023. [Online]. Available: https://mhealth.jmir.org/2023/1/e43340

    work page 2023

  53. [53]

    Active ascent accelerates the time course but not the overall incidence and severity of acute mountain sickness at 3,600 m,

    B. A. Beidleman, P. S. Figueiredo, S. D. Landspurg, J. K. Femling, J. D. Williams, J. E. Staab, M. J. Buller, J. P. Karl, A. J. Reilly, T. J. Mayschak, E. Y . Atkinson, T. J. Mesite, and R. W. Hoyt, “Active ascent accelerates the time course but not the overall incidence and severity of acute mountain sickness at 3,600 m,” vol. 135, no. 2, pp. 436–444, 20...

    work page doi:10.1152/japplphysiol.00216.2023 2023

  54. [54]

    Evaluation of leading smartwatches for the detection of hypoxemia: Comparison to reference oximeter,

    S. Walzel, R. Mikus, V . Rafl-Huttova, M. Rozanek, T. E. Bachman, and J. Rafl, “Evaluation of leading smartwatches for the detection of hypoxemia: Comparison to reference oximeter,” vol. 23, no. 22, p. 9164, 2023, pMCID: PMC10674783

    work page 2023

  55. [55]

    Healthcare monitoring of mountaineers by low power wireless sensor networks,

    R. K. Garg, J. Bhola, and S. K. Soni, “Healthcare monitoring of mountaineers by low power wireless sensor networks,”Informatics in Medicine Unlocked, vol. 27, p. 100775, 2021

    work page 2021

  56. [56]

    Commercial smartwatch with pulse oximeter detects short-time hy- poxemia as well as standard medical-grade device: Validation study,

    J. Rafl, T. E. Bachman, V . Rafl-Huttova, S. Walzel, and M. Rozanek, “Commercial smartwatch with pulse oximeter detects short-time hy- poxemia as well as standard medical-grade device: Validation study,” DIGITAL HEALTH, vol. 8, p. 20552076221132127, 2022

    work page 2022

  57. [57]

    Smartphone-enabled heart rate variability and acute mountain sickness,

    A. M. B. Mellor, J. Bakker-Dyos, J. O’Hara, D. R. Woods, D. A. Holdsworth, and C. J. Boos, “Smartphone-enabled heart rate variability and acute mountain sickness,”Clinical Journal of Sport Medicine, vol. 28, no. 1, pp. 76–81, Jan. 2018. Abu Kaisar Mohammad Masum(S’25) received his B.Sc. degree in Computer Science and En- gineering from Daffodil Internatio...

    work page 2018