Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

DPQ-HD: Post-Training Compression for Ultra-Low Power Hyperdimensional Computing

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Post-training compression shrinks hyperdimensional computing models 20-100x with no retraining.

desk verdict Solid engineering story undermined by a train/test encoder mismatch the paper never addresses; the reported accuracy retention is not explained by the described pipeline. read the letter →

arxiv 2505.05413 v1 pith:YRJ7AT5X submitted 2025-05-08 cs.LG

Nilesh Prasad Pandey , Shriniwas Kulkarni , David Wang , Onat Gungor , Flavio Ponzina , Tajana Rosing This is my paper · ORCID
classification cs.LG
keywords HyperdimensionalComputingPost-TrainingCompressionBrain-InspiredLow-RankDecompositionPruningQuantizationEarlyExitInferenceEdgeAI
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

Hyperdimensional computing (HDC) classifies by projecting inputs into a high-dimensional space with a random matrix and comparing the resulting hypervectors against stored class hypervectors. This paper claims that both parts of that pipeline, the projection matrix and the class hypervectors, can be compressed after training by applying low-rank decomposition, pruning, and quantization in that order, with no retraining. The reported payoff is 20-100x less memory on image, speech, and graph classification with only a 1-2% accuracy drop, and up to 56x faster inference on a low-power microcontroller. Optimization time is also up to 100x lower than retraining-based methods because the only data needed is a small calibration set such as 128 samples. The significance is practical: HDC models become deployable on memory-starved edge devices without labeled retraining data.

What carries the argument

The load-bearing object is the product structure $P \approx P_1 P_2$ imposed on the projection matrix, because it cuts the encoder's storage and MAC count before pruning or quantization acts. Pruning then operates on the dimension $D'$ of the intermediate hypervector by deleting trailing coordinates, and quantization uses a symmetric scale chosen by an MSE search over candidate scales. The order is justified by a lemma: for any vector $x$, the error of pruning-then-quantizing is bounded by $\|\epsilon_{q\circ s}(x)\| \le \|\epsilon_q(x)\| + \|\epsilon_s(x)\|$, so pruning before quantization adds no error beyond the sum of the two individual operations. The online accelerator is a chunked cosine similarity loop that removes the two least-likely classes per step until half the classes remain, then removes one per step, with early exit when the top-two margin exceeds a threshold $\tau$ calibrated on the same small validation set.

What would settle it

For a fixed dataset and fixed uncompressed model, run DPQ-HD's calibration phase on five disjoint 128-sample subsets, produce five compressed models, and evaluate each on the same held-out test set; if the accuracy spread across the five compressed models exceeds roughly two percentage points, the claim that a 128-sample calibration set suffices to keep accuracy within 1-2% of uncompressed fails.

Watch

Extended reading notes

Core claim

The central claim is that end-to-end HDC compression is achievable without retraining by targeting every memory-heavy component. DPQ-HD replaces the random projection matrix $P \in \mathbb{R}^{F \times D}$ with a product $P_1 \in \mathbb{R}^{F \times r}$ and $P_2 \in \mathbb{R}^{r \times D'}$, prunes the hypervector dimension from $D$ to a calibrated $D'$ by dropping trailing dimensions, and quantizes both the decomposed encoder and the class hypervectors using an MSE-optimal symmetric scale. On image, speech, and graph classification workloads, the paper reports up to 20x memory reduction for image tasks and 100x for graph tasks, with total memory reduced 20-100x compared with uncompressed HDC while accuracy drops only 1-2%. A progressive inference scheme computes cosine similarity in chunks, removes unlikely classes, and exits early when the top-two margin exceeds a calibrated threshold, cutting runtime by up to 76.94%. Compared with retraining-based compression, DPQ-HD claims comparable or better accuracy per unit of memory with up to 100x less optimization time, and on an 8-bit ATmega328P microcontroller it reports 56x faster inference and 56x lower energy than the uncompressed model.

Load-bearing premise

The 128-sample calibration set used to choose decomposition rank, pruning ratio, bitwidth, and early-exit threshold is representative of the test distribution, so the settings that look good on calibration also hold accuracy on unseen data.

Editorial extensions

If this is right

  • Compressing the encoder and the classifier together, rather than only one component, is what lets HDC workloads drop 20-100x in memory; compressing only the model or only the encoder would leave most of the savings on the table.
  • Deployment on MCU-class devices with very limited SRAM becomes practical: the paper reports a compressed 10k-dimensional workload running in 0.32 s at 5.05 mJ, a 56x improvement over the uncompressed baseline.
  • Because pruning happens before quantization, the combined error is bounded by the sum of each operation's error, so the pipeline can be tuned by choosing rank, pruning ratio, and bitwidth separately without expecting a negative interaction.
  • The adaptive early-exit strategy can be layered on top of compression, reducing prediction runtime by up to 76.94% while keeping accuracy intact.

Reading between the lines

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

  • The paper's Lemma 1 bound is an upper bound; on real workloads the combined error may be much smaller, which would mean even more aggressive pruning or lower bitwidths are safe than the calibration currently selects.
  • Because DPQ-HD needs only a small calibration set and no retraining, the same pipeline should transfer to online or continual HDC learning settings where labeled data is scarce, a use case the paper motivates but does not evaluate.
  • The decomposition step assumes the random projection matrix has low effective rank; encoders built from structured or learned projections, whose matrices are already compact, might gain less from decomposition and would need a different compression lever.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes DPQ-HD, a post-training compression framework for hyperdimensional computing (HDC) that combines low-rank decomposition of the random projection encoder, dimensionality pruning, and symmetric quantization, plus an adaptive early-exit inference strategy. The authors report 20-100x memory reductions with 1-2% accuracy drop on MNIST, Fashion-MNIST, CIFAR-10, ISOLET, PROTEINS, and DD, and up to 56x inference speedup on an ATmega328P microcontroller, claiming superiority over post-training baselines and parity with retraining-based state-of-the-art while avoiding retraining.

Significance. If the method worked as described, it would be a practically valuable contribution to edge HDC deployment, since post-training compression without retraining would substantially reduce deployment cost. The paper includes a broad evaluation across three HDC backbones, several datasets, and real microcontroller measurements, and the MSE-based quantization procedure is simple and clearly stated. However, the central claim depends on whether the compressed encoder is actually compatible with the class hypervectors produced by the original encoder. As written, the method description is internally inconsistent on this point, and the headline accuracy claims are contradicted by the reported comparison with QuantHD on ISOLET. The significance of the contribution is therefore not established in the current form.

major comments (3)
  1. [§3.1.1, §3.1.2, Eq. (1), Figure 2] Equation (1) replaces the trained projection matrix P with P' = P1·P2, where P1 and P2 are 'randomly initialized.' A random Gaussian P is full rank with probability 1, and independent random P1, P2 do not produce a low-rank approximation of P; they produce a new encoder whose row space is independent of P's row space. The class hypervectors W were obtained by encoding training data with P, and §3.1.2 only states that W's trailing dimensions are removed. Under the described pipeline, test vectors encoded with P' would be compared against class hypervectors from a different random subspace, and the dot products in Algorithm 2 should carry no classification signal. Figure 2a, however, reports decomposed accuracy around 80% on MNIST, which is only possible if W was recomputed with the compressed encoder (or if P' were a true approximation of P, which the text does not claim). The manuscript never states that W is recomputed. If W is recomputed, the method is not post-training and the optimization-time comparison in Figure 5 must include the cost of the re-encoding pass. This internal inconsistency invalidates the central 'no retraining' claim as written.
  2. [Abstract, §4.3.3, Figure 4b] The abstract and conclusion claim that DPQ-HD 'performs better or at par with retraining-based state-of-the-art.' Figure 4b reports QuantHD at 94.6% on ISOLET versus DPQ-HD's 91.46%, a gap of more than 3 percentage points, and MicroHD at 92.51% versus DPQ-HD's 91.46%. The text itself acknowledges that QuantHD is 'higher on ISOLET.' On accuracy alone, DPQ-HD is not at par with these retraining baselines on ISOLET. If 'at par' is meant to include the memory and speed tradeoffs, that should be stated explicitly and the claim should be rephrased to avoid implying accuracy parity.
  3. [§3.2, Lemma 1] The theoretical justification for applying pruning before quantization does not actually compare the two orders. The proof of Lemma 1 only shows, via the triangle inequality, that the error of pruning-then-quantization is bounded by the sum of the individual pruning and quantization errors. It does not show that pruning-then-quantization is no worse than quantization-then-pruning, which is the claimed ordering decision. This is a presentation issue for a supporting result rather than the central empirical claim, but the section should be rewritten to state what is actually proven.
minor comments (5)
  1. [§3.1.1] The phrase 'two level low rank decomposition' is misleading when P1 and P2 are randomly initialized, because P' is not a low-rank approximation of the original P. The terminology should be corrected or the construction should be changed to a true factorization of P.
  2. [§2.2] There is a duplicated word in 'more than than 80% of an HDC model's memory and runtime requirements.'
  3. [§4.1] The word 'effectivenss' is misspelled; it should be 'effectiveness.'
  4. [Table 1] The sentence 'DeMAT, MicroHD and DPQ-HD demonstrate significant inference performance improvements, resulting in 16.12×, 16.27× and 56× respectively, respectively' contains a duplicated 'respectively' and should be reworded.
  5. [§4.2, Figure 2] The calibration analysis reports accuracy averaged over five 128-sample subsets, but the final test accuracy is reported without error bars or a statement of how many calibration subsets were used for the final configuration. Reporting variance would strengthen the claim that the selected hyperparameters transfer.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DPQ-HD's compression pipeline is empirical and its theoretical Lemma is self-contained; no prediction is forced by construction.

full rationale

DPQ-HD's derivation chain is not circular. The calibration phase (Sections 3.1.2, 3.4.1, and 4.2) selects decomposition rank, pruning ratio, bitwidth, and early-exit threshold on small validation subsets (e.g., 128 samples), and the reported accuracy, memory, and runtime numbers are then measured on test workloads. This is ordinary hyperparameter tuning rather than a fitted quantity being renamed as a prediction. Section 3.2's Lemma 1 proves an error bound for pruning-before-quantization directly from the definitions of the quantization and pruning maps using the triangle inequality; the bound is not assumed in the proof. Self-citations such as MicroHD [28] and DeMAT [35] appear only as comparison baselines with their own reported numbers; the central claim of DPQ-HD does not depend on accepting any internal premise of those papers as a load-bearing assumption. One non-circular correctness concern is worth flagging: Eq. (1) describes P1 and P2 as 'randomly initialized' while writing P' ≈ P1·P2, and the paper does not state whether the stored class hypervectors are re-encoded with the compressed projection before evaluation. A random product is not a low-rank approximation of the original P in any mathematical sense, so this is an internal-consistency and experimental-reproducibility concern rather than a circular derivation. Under the rule that circularity must be exhibited as a specific reduction in the paper's own equations, no circular step is scored.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central trade-off claims rest on a handful of calibration-selected hyperparameters and a domain assumption that low-rank random projections retain discriminative information. No new physical entities are introduced.

free parameters (5)
  • Decomposition rank r = 256-512 (e.g., 256 for CentroidHD/GraphHD, 512 for HDnn)
    Selected via calibration on 128 samples; controls the rank of the projection matrix approximation and directly determines memory and accuracy.
  • Pruning ratio (dimension D') = 10-70% pruned (e.g., 70% MNIST, 40% DD, 10% CIFAR-10)
    Selected via calibration; removes trailing dimensions of hypervectors and class HVs.
  • Quantization bitwidth b = 3-4 bits (per Figure 3)
    Chosen per workload; hardware-aware bitwidth for MCU with bit packing.
  • Early-exit threshold tau = Set as mean calibration margin
    Determines when adaptive inference stops; directly trades speed for accuracy.
  • Per-tensor quantization scale s = MSE-optimal candidate from {0.1s, 0.2s, ..., s}
    Searched by Algorithm 1 to minimize dequantization MSE; standard PTQ scale but tuned per tensor.
assumptions (4)
  • domain assumption Random projection encoding produces hypervectors whose dot products preserve similarity for the target tasks (HDC property).
    Used in Section 3.1.1 to justify replacing the full projection matrix with a low-rank random approximation.
  • ad hoc to paper A low-rank random matrix P1*P2 retains enough discriminatory information to classify inputs in the datasets tested.
    Not proven; the paper only shows empirically that a rank as low as 256 works on MNIST and DD. This is the core design bet of the decomposition step.
  • standard math The error decomposition in Lemma 1 (triangle inequality) is valid for the defined pruning and quantization operators.
    Section 3.2 relies on norm inequalities; the proof is correct but is a weak bound and does not establish that pruning must precede quantization.
  • domain assumption The 128-sample calibration set is representative of the test distribution for hyperparameter selection.
    All rank, pruning, bitwidth, and threshold choices depend on this set; no sensitivity analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DPQ-HD: Post-Training Compression for Ultra-Low Power Hyperdimensional Computing." pith.science (2026). https://pith.science/paper/YRJ7AT5X

@misc{pith2026250505413,
  author       = {Pith},
  title        = {Pith review of: DPQ-HD: Post-Training Compression for Ultra-Low Power Hyperdimensional Computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YRJ7AT5X}},
  note         = {Machine review of arXiv:2505.05413}
}
read the original abstract

Hyperdimensional Computing (HDC) is emerging as a promising approach for edge AI, offering a balance between accuracy and efficiency. However, current HDC-based applications often rely on high-precision models and/or encoding matrices to achieve competitive performance, which imposes significant computational and memory demands, especially for ultra-low power devices. While recent efforts use techniques like precision reduction and pruning to increase the efficiency, most require retraining to maintain performance, making them expensive and impractical. To address this issue, we propose a novel Post Training Compression algorithm, Decomposition-Pruning-Quantization (DPQ-HD), which aims at compressing the end-to-end HDC system, achieving near floating point performance without the need of retraining. DPQ-HD reduces computational and memory overhead by uniquely combining the above three compression techniques and efficiently adapts to hardware constraints. Additionally, we introduce an energy-efficient inference approach that progressively evaluates similarity scores such as cosine similarity and performs early exit to reduce the computation, accelerating prediction inference while maintaining accuracy. We demonstrate that DPQ-HD achieves up to 20-100x reduction in memory for image and graph classification tasks with only a 1-2% drop in accuracy compared to uncompressed workloads. Lastly, we show that DPQ-HD outperforms the existing post-training compression methods and performs better or at par with retraining-based state-of-the-art techniques, requiring significantly less overall optimization time (up to 100x) and faster inference (up to 56x) on a microcontroller

Figures

Figures reproduced from arXiv: 2505.05413 by the authors.

Figure 1
Figure 1. Illustration of DPQ-HD highlighting decomposition, [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Effect of decomposition rank on calibration accuracy for (a) MNIST and pruning ratio for (b-i) DD and (b-ii) Fashion [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of uncompressed HDC workloads trained using (a) CentroidHD, (b) GraphHD, and (c) HDnn, and their [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Comparison of accuracy and memory overhead for (a) post-training baselines: Naive Quantization, Eff-SparseHD [ [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Comparison of DPQ-HD with MicroHD [28] and QuantHD [15] in memory usage, compressed model accuracy, and offline optimization time related to naive quantization. loss when applied extensively without retraining. This approach enables DPQ-HD to preserve high accuracy eve…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Gram-Space: Structure-Preserving Codebook Compression for Memory-Efficient Neuro-Symbolic AI

    cs.LG 2026-08 reject novelty 3.0 of 10

    Projecting a codebook into its own M-dimensional orthonormal basis does not compress it when the basis is stored, so the claimed 15.75x memory reduction is not supported.

Reference graph

Works this paper leans on

39 extracted references · 26 canonical work pages · cited by 1 Pith paper

  1. [1]

    Hussam Amrouch, Mohsen Imani, Xun Jiao, Yiannis Aloimonos, Cornelia Fer- muller, Dehao Yuan, Dongning Ma, Hamza E Barkam, Paul R Genssler, and Peter Sutor. 2022. Brain-inspired hyperdimensional computing for ultra-efficient edge ai. In 2022 International Conference on Hardware/Software Codesign and System Synthesis (CODES+ ISSS). IEEE, 25–34

    work page 2022

  2. [2]

    Arthur Asuncion, David Newman, et al. 2007. UCI machine learning repository

    2007

  3. [3]

    K Behnam, X Hanyang, M Justin, and R Tajana. 2021. tiny-hd: Ultra-efficient hyperdimensional computing engine for iot applications. In IEEE/ACM Design Automation and Test in Europe Conference (DATE), IEEE, IEEE , Vol. 10

    work page 2021

  4. [4]

    Simone Benatti, Fabio Montagna, Victor Kartsch, Abbas Rahimi, Davide Rossi, and Luca Benini. 2019. Online learning and classification of EMG-based gestures on a parallel ultra-low power platform using hyperdimensional computing. IEEE transactions on biomedical circuits and systems 13, 3 (2019), 516–528

    work page 2019

  5. [5]

    Kim Isaac I Buelagala, Ginzy S Javier, Sean Alfred A Lipardo, James Carlo E Sorsona, Sherry Joy Alvionne S Baquiran, Lawrence Roman A Quizon, Allen Jason A Tan, Ryan Albert G Antonio, Fredrick Angelo R Galapon, and Anastacia B Alvarez. 2023. Energy-Efficient Sparse Hyperdimensional Computing for Speech Recognition. In 2023 20th International SoC Design Co...

    work page 2023

  6. [6]

    Cheng-Yang Chang, Yu-Chuan Chuang, Chi-Tse Huang, and An-Yeu Wu. 2023. Recent progress and development of hyperdimensional computing (hdc) for edge intelligence. IEEE Journal on Emerging and Selected Topics in Circuits and Systems 13, 1 (2023), 119–136

    work page 2023

  7. [7]

    Wei-Chen Chen, H-S Philip Wong, and Sara Achour. 2024. Bitwise Adaptive Early Termination in Hyperdimensional Computing Inference. In Proceedings of the 61st ACM/IEEE Design Automation Conference . 1–6

    work page 2024

  8. [8]

    Tsai-Kan Chien, Lih-Yih Chiou, Shyh-Shyuan Sheu, Jing-Cian Lin, Chang-Chia Lee, Tzu-Kun Ku, Ming-Jinn Tsai, and Chih-I Wu. 2016. Low-power MCU with embedded ReRAM buffers as sensor hub for IoT applications. IEEE Journal on Emerging and Selected Topics in Circuits and Systems 6, 2 (2016), 247–257

    work page 2016

Show all 39 references
  1. [9]

    Li Deng. 2012. The mnist database of handwritten digit images for machine learning research [best of the web]. IEEE signal processing magazine 29, 6 (2012), 141–142

    2012

  2. [10]

    Paul D Dobson and Andrew J Doig. 2003. Distinguishing enzyme structures from non-enzymes without alignments. Journal of molecular biology 330, 4 (2003), 771–783

    2003

  3. [11]

    Arpan Dutta, Saransh Gupta, Behnam Khaleghi, Rishikanth Chandrasekaran, Weihong Xu, and Tajana Rosing. 2022. Hdnn-pim: Efficient in memory design of hyperdimensional computing with feature extraction. In Proceedings of the Great Lakes Symposium on VLSI 2022 . 281–286

    2022

  4. [12]

    Onat Gungor, Tajana Rosing, and Baris Aksanli. 2024. A2HD: Adaptive Adversar- ial Training for Hyperdimensional Computing-Based Intrusion Detection Against Adversarial Attacks. In 2024 IEEE International Conference on Cyber Security and Resilience (CSR). 107–113. https://doi.o...

    2024

  5. [13]

    Nicolás Hernández, Francisco Almeida, and Vicente Blanco. 2024. Optimizing convolutional neural networks for IoT devices: performance and energy efficiency of quantization techniques. The Journal of Supercomputing 80, 9 (2024), 12686– 12705

    2024

  6. [14]

    Alejandro Hernández-Cano, Namiko Matsumoto, Eric Ping, and Mohsen Imani

  7. [15]

    Mohsen Imani, Samuel Bosch, Sohum Datta, Sharadhi Ramakrishna, Sahand Salamat, Jan M Rabaey, and Tajana Rosing. 2019. Quanthd: A quantization framework for hyperdimensional computing. IEEE Transactions on Computer- Aided Design of Integrated Circuits and Systems 39, 10 (2019),...

    2019

  8. [16]

    Mohsen Imani, Deqian Kong, Abbas Rahimi, and Tajana Rosing. 2017. Voicehd: Hyperdimensional computing for efficient speech recognition. In 2017 IEEE international conference on rebooting computing (ICRC) . IEEE, 1–8

    2017

  9. [17]

    Pentti Kanerva. 2009. Hyperdimensional computing: An introduction to com- puting in distributed representation with high-dimensional random vectors. Cognitive computation 1 (2009), 139–159

    2009

  10. [18]

    Behnam Khaleghi, Jaeyoung Kang, Hanyang Xu, Justin Morris, and Tajana Rosing

  11. [19]

    N Kishore Kumar and Jan Schneider. 2017. Literature survey on low rank approx- imation of matrices. Linear and Multilinear Algebra 65, 11 (2017), 2212–2244

    2017

  12. [20]

    Denis Kleyko, Dmitri A Rachkovskij, Evgeny Osipov, and Abbas Rahimi. 2022. A survey on hyperdimensional computing aka vector symbolic architectures, part i: Models and data transformations. Comput. Surveys 55, 6 (2022), 1–40

    2022

  13. [21]

    Alex Krizhevsky, Vinod Nair, and Geoffrey Hinton. 2010. Cifar-10 (canadian institute for advanced research). URL http://www. cs. toronto. edu/kriz/cifar. html 5, 4 (2010), 1

    2010

  14. [22]

    Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Wei-Ming Chen, Wei-Chen Wang, Guangxuan Xiao, Xingyu Dang, Chuang Gan, and Song Han. 2024. Awq: Activation-aware weight quantization for on-device llm compression and accel- eration. Proceedings of Machine Learning and Systems 6 ...

    2024

  15. [23]

    Markus Nagel, Marios Fournarakis, Rana Ali Amjad, Yelysei Bondarenko, Mart Van Baalen, and Tijmen Blankevoort. 2021. A white paper on neural network quantization. arXiv preprint arXiv:2106.08295 (2021)

    2021 arXiv

  16. [24]

    Igor Nunes, Mike Heddes, Tony Givargis, Alexandru Nicolau, and Alex Veiden- baum. 2022. GraphHD: Efficient graph classification using hyperdimensional computing. In 2022 Design, Automation & Test in Europe Conference & Exhibition (DATE). IEEE, 1485–1490

    2022

  17. [25]

    Nilesh Prasad Pandey, Marios Fournarakis, Chirag Patel, and Markus Nagel. 2023. Softmax bias correction for quantized generative models. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 1453–1458

    2023

  18. [26]

    Nilesh Prasad Pandey, Markus Nagel, Mart van Baalen, Yin Huang, Chirag Patel, and Tijmen Blankevoort. 2023. A practical mixed precision algorithm for post- training quantization. arXiv preprint arXiv:2302.05397 (2023)

    2023 arXiv

  19. [27]

    Ian R Peitzsch, Evan W Gretok, and Alan D George. 2024. Putting the “Space” in Hyperspace: Investigating Hyperdimensional Computing for Space Applications. (2024)

    2024

  20. [28]

    Flavio Ponzina and Tajana Rosing. 2024. MicroHD: An Accuracy-Driven Op- timization of Hyperdimensional Computing Algorithms for TinyML systems. arXiv preprint arXiv:2404.00039 (2024)

    2024 arXiv

  21. [29]

    Haseena Rahmath P, Vishal Srivastava, Kuldeep Chaurasia, Roberto G Pacheco, and Rodrigo S Couto. 2024. Early-exit deep neural network-a comprehensive survey. Comput. Surveys 57, 3 (2024), 1–37

    2024

  22. [30]

    Alexander Redding, Xiaofan Yu, Shengfan Hu, Pat Pannuto, and Tajana Rosing

  23. [31]

    Raghubir Singh and Sukhpal Singh Gill. 2023. Edge AI: a survey. Internet of Things and Cyber-Physical Systems 3 (2023), 71–92

    2023

  24. [32]

    Srinivasa R Sridhara. 2011. Ultra-low power microcontrollers for portable, wear- able, and implantable medical electronics. In 16th Asia and South Pacific Design Automation Conference (ASP-DAC 2011). IEEE, 556–560

    2011

  25. [33]

    Anthony Thomas, Sanjoy Dasgupta, and Tajana Rosing. 2021. A theoretical perspective on hyperdimensional computing. Journal of Artificial Intelligence Research 72 (2021), 215–249

    2021

  26. [34]

    Han Xiao, Kashif Rasul, and Roland Vollgraf. 2017. Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms. arXiv preprint arXiv:1708.07747 (2017)

    2017 arXiv

  27. [35]

    Weihong Xu, Jaeyoung Kang, and Tajana Rosing. 2023. Fsl-hd: Accelerating few-shot learning on reram using hyperdimensional computing. In 2023 Design, Automation & Test in Europe Conference & Exhibition (DATE) . IEEE, 1–6

    2023

  28. [36]

    Tianyang Yu, Bi Wu, Ke Chen, Gong Zhang, and Weiqiang Liu. 2023. Fully Learnable Hyperdimensional Computing Framework with Ultra-tiny Accelerator for Edge-side Applications. IEEE Trans. Comput. (2023)

    2023

  29. [2021]

    In 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE)

    Onlinehd: Robust, efficient, and single-pass online learning using hyper- dimensional system. In 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE). IEEE, 56–61

    2021

  30. [2022]

    In Proceedings of the 59th ACM/IEEE Design Automation Conference

    Generic: highly efficient learning engine on edge using hyperdimensional computing. In Proceedings of the 59th ACM/IEEE Design Automation Conference . 1117–1122

  31. [2023]

    In Proceedings of the 2nd Workshop on Networked Sensing Systems for a Sustainable Society

    EmbHD: A Library for Hyperdimensional Computing Research on MCU- Class Devices. In Proceedings of the 2nd Workshop on Networked Sensing Systems for a Sustainable Society . 187–192

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.