Pith. sign in

REVIEW 3 major objections 6 minor 81 references

Bridging the Gap Between Hyperdimensional Computing and Kernel Methods via the Nystr\"om Method

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read NysHD composes the Nyström method with random hyperplane rounding so that any positive-semidefinite kernel becomes a low-precision HDC encoding whose inner products approximate normalized kernel values.

desk verdict Real new combination and a correct conditional theorem, but the paper overstates what 'preserving the kernel' means and skips the SVM baseline. read the letter →

arxiv 2608.06860 v1 pith:FK27WKQL submitted 2026-08-07 cs.LG cs.AI

Quanling Zhao , Anthony Hitchcock Thomas , Ari Brin , Xiaofan Yu , Tajana Rosing This is my paper · ORCID
classification cs.LGcs.AI
keywords hyperdimensionalcomputingNyströmmethodkernelapproximationpositive-semidefinitekernelsgraphstringrandomhyperplaneroundinglow-precisionembeddings
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) represents data as high-dimensional random vectors, but its accuracy depends on how raw inputs are encoded. This paper proposes NysHD, an encoding that starts from a user-chosen positive-semidefinite kernel (a similarity measure that behaves like an inner product in a high-dimensional space) and produces HDC vectors whose inner products approximate the normalized kernel values. The payoff is that the mature literature on graph, string, and other specialized kernels can be imported into HDC, which previously relied on simpler L1/L2 or angular similarity encodings. In experiments, NysHD improves HDC classification accuracy by an average of 11% on graph datasets and 17% on string datasets, and it is competitive with or faster than several deep neural network baselines on the tasks tested.

What carries the argument

The load-bearing machinery is the composition of the Nyström method and random hyperplane rounding. The Nyström method approximates the kernel matrix $G$ by $C H_Z^+ C^T$, where $C$ holds kernel values from each point to $s$ sampled landmarks, and it gives explicit features $\phi_{nys}(x_i)=\Lambda^{-1/2}Q^T C^{(i)}$ whose inner products reproduce the approximate matrix. Sign-thresholded random projection then converts these high-precision features into low-precision HDC vectors while preserving angular similarity: for unit vectors $v$ and $v'$, the expected normalized inner product of their signs is $1 - 2\cos^{-1}(v\cdot v')/\pi$. The composition turns a kernel value into an angle, and the HDC inner product tracks that angle.

What would settle it

Take a dataset where the kernel matrix has high numerical rank or contains rare clusters that uniform landmarks are unlikely to cover, then compare the spectral-norm error $\|G - C H_Z^+ C^T\|$ with the gap between the HDC inner-product matrix and the true normalized kernel matrix; if the error is large and the accuracy advantage over the existing HDC baseline disappears, the practical claim would be refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that composing the Nyström low-rank approximation with sign-thresholded random projection produces an HDC encoding that tracks a user-defined kernel. For a positive-definite kernel $K$, landmarks $Z$, and Nyström approximation $\hat{G} = C H_Z^+ C^T$, the encoding $\phi(x_i)=\sqrt{\pi/(2d)}\,\operatorname{sign}(P_{rp}\Lambda^{-1/2}Q^T C^{(i)})$ satisfies $E[\phi(x_i)\cdot\phi(x_j)] = \pi/2 - \cos^{-1}\big(\hat{G}_{ij}/\sqrt{\hat{G}_{ii}\hat{G}_{jj}}\big)$. Because $\cos^{-1}(t)$ expands as $\pi/2 - t$ to first order, this expected inner product equals the normalized kernel value $\hat{G}_{ij}/\sqrt{\hat{G}_{ii}\hat{G}_{jj}}$ up to first order. The theorem thus gives a general recipe: any positive-semidefinite similarity function can be realized as an equivalent low-precision, randomized HDC mapping, provided the Nyström approximation of the kernel matrix is accurate.

Load-bearing premise

The load-bearing premise is that uniform random sampling of $s$ landmarks from the training set gives a Nyström approximation accurate enough for classification, because the formal kernel-preservation guarantee inherits whatever approximation error the Nyström step makes.

Editorial extensions

If this is right

  • Any positive-semidefinite kernel can be plugged into HDC classification by computing kernel values to a small landmark set, so HDC inherits the large existing library of graph, string, and time-series kernels.
  • HDC's single-pass training and low-precision hardware advantages remain intact, because the encoding step is a one-time kernel evaluation plus sign operations.
  • On the evaluated graph benchmarks NysHD raises average HDC accuracy by 11% over GraphHD, and on string benchmarks by 17% over N-gram HDC.
  • NysHD matches or exceeds several deep graph and string models on some datasets while training much faster, indicating that HDC can close part of the accuracy gap with deep learning.
  • The first-order approximation means HDC inner products track normalized, not raw, kernel values, so scale information in the kernel is discarded and classification is driven by angular similarity.

Reading between the lines

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

  • Uniform landmark sampling is the weakest practical link: if the landmark set misses important regions of the data, the Nyström error dominates and the formal guarantee says nothing useful; adaptive landmark selection, which the paper leaves to future work, is the natural next step and could be tested by comparing spectral approximation error across sampling strategies.
  • Because the kernel choice is now the main design lever, progress in kernel design, such as alignment-tolerant kernels for time series, should transfer directly to HDC accuracy; this is a testable corollary the paper motivates but does not evaluate.
  • The sign-thresholding step is one of several possible low-precision quantizers, and replacing it with binary or ternary rounding may trade a small accuracy loss for larger hardware gains; the paper's identity gives a quantitative baseline for such experiments.
  • The result suggests a broader bridge: other kernel-approximation schemes beyond the Nyström method could be imported into HDC as long as they produce explicit finite-dimensional embeddings, though the paper only establishes the Nyström route.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes NysHD, an HDC encoding that composes Nyström embeddings of a user-supplied positive-semidefinite kernel with random hyperplane rounding (sign random projections), producing low-precision high-dimensional vectors. The main theoretical result, Theorem 1, states that the expected inner product of two encoded points equals π/2 − arccos of the normalized Nyström kernel value, and the authors use a first-order Taylor expansion of arccos to claim that normalized kernel values are approximately preserved. The paper further reports empirical gains over existing HDC encodings on graph and string datasets, alongside comparisons with deep learning baselines.

Significance. If the claims are properly qualified, this is a useful contribution to the HDC literature: it provides a clean recipe for importing non-translation-invariant kernels into HDC, and the proof of the exact conditional identity in Eq. (4) is straightforward and correct, relying only on Charikar's rounding result and the Nyström embedding identity. The empirical study covers several standard graph and string benchmarks and reports meaningful average accuracy improvements of 11% and 17% over GraphHD and N-gram HDC. These strengths are tempered by an overstatement in the 'equivalent mapping' language, a formal inconsistency in the statement of Theorem 1 regarding the expectation, and a reproducibility issue in the landmark-count configuration for two small datasets.

major comments (3)
  1. [HDC Encoding via Nyström Approximation, Theorem 1 and Eq. (4)-(8)] The statement of Theorem 1 says the expectation in Eq. (4) is taken over both landmark sampling and P_rp, but the right-hand side contains Ĝ_ij, which depends on the randomly sampled landmark set. The proof only averages over P_rp for a fixed landmark set (Eqs. (5)-(8)). This is not merely a notational issue: because arccos is nonlinear, E_Z[π/2 − arccos(Ĝ_ij/sqrt(Ĝ_ii Ĝ_jj))] is not equal to π/2 − arccos(E_Z[Ĝ_ij]/sqrt(E_Z[Ĝ_ii]E_Z[Ĝ_jj])) in general. The theorem should be restated as a conditional statement for a fixed landmark set, or the right-hand side of Eq. (4) must be written with the expectation over landmark sets applied to the whole expression. As written, the theorem is not correct.
  2. [Encoding Process, Eq. (9) and abstract/contributions] Equation (9) is a Taylor expansion of arccos around r=0, and it is not a global approximation on the domain [-1,1]. For r=0.9 the predicted inner product is about 1.12 instead of 0.9, and for r=0.99 it is about 1.43 instead of 0.99. Since HDC classification scores are sums of f(r_ij)=π/2 − arccos(r_ij) over class prototypes, monotonicity of f does not make the HDC linear model equivalent to the kernel's linear model. The abstract's phrase 'equivalent mapping in HDC' and the contribution statement that inner products 'preserve normalized kernel values' therefore overstate what is established. The exact identity in Eq. (4) is a legitimate and interesting result; the paper should present it as a fixed monotone nonlinear relationship between HDC inner products and normalized Nyström kernel values, and should avoid claiming global first-order equivalence without further bounds.
  3. [Experimental Setup and Baselines] The landmark count is set to s=max(300, 2% of training data). For MUTAG and Promoter the training sets have 150 and 84 samples, respectively, so s=300 cannot be sampled without replacement as Algorithm 1 requires. The reported results for these two datasets in Tables 2 and 3 are therefore not reproducible under the stated configuration. The authors should clarify whether s was capped at the dataset size, whether sampling was with replacement for small datasets, or whether a different rule was actually used.
minor comments (6)
  1. [Algorithm 1] The line 'Prp = [w1, w1, ..., wd]^T' appears to duplicate w1; it should be [w1, w2, ..., wd]^T or an equivalent enumeration of the sampled rows.
  2. [Evaluation, first paragraph] The sentence 'we validate 1' is missing a reference; it should read 'we validate Theorem 1'.
  3. [Table 1] The entry 'Nautral Language' is a typo for 'Natural Language'.
  4. [Experimental Setup and Baselines] The phrase 'for for bio-sequence' contains a duplicated word, and the sentence about the GPU is ungrammatical ('required an Nvidia RTX 3050 GPU is used').
  5. [Figure 2] The caption 'Numerical difference of spectral norm between normalized kernel matrices compute directly from kernel function and approximated kernel matrices with our encoding' is ambiguous; please specify whether the plotted quantity is ||K − K̂||_2, |σ(K) − σ(K̂)|, or some other norm, and describe how the approximated matrix is formed from the encodings.
  6. [Experimental Setup and Baselines] The choice s=max(300, 2% of training data) is not justified, and since the number of landmarks is a free parameter, reporting sensitivity to s on at least one dataset would help the reader understand the accuracy/efficiency trade-off.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the kernel-preservation theorem is a direct composition of the Nyström identity and Charikar's external sign-projection result, with no fitted target quantity or load-bearing self-citation.

full rationale

Walking the derivation chain in 'HDC Encoding via Nyström Approximation', the paper's central Theorem 1 rests on two independently established ingredients. First, the Nyström features are defined so that φ_nys(x_i)·φ_nys(x_j) = Ĝ_ij (Eq. 2), which is the standard Nyström identity. Second, Eq. 3 is Charikar's sign-thresholded random projection expectation for unit vectors, an external result. Composing these gives Eq. 4: E[φ(x_i)·φ(x_j)] = π/2 − arccos(Ĝ_ij / sqrt(Ĝ_ii Ĝ_jj)). The subsequent first-order Taylor expansion of arccos is an approximation used to state the 'normalized kernel' preservation claim; it is not a parameter fitted to labels, nor is the target result used to define the encoding. The encoding pipeline (Algorithms 1 and 2) is fully specified from the user kernel and landmark sample, with no dependence on class labels. The paper's acknowledged limitations—uniform landmark sampling and Nyström approximation quality—bear on approximation accuracy, not on circularity. Self-citations (e.g., Thomas, Dasgupta, and Rosing 2021; Zhao et al. 2022) appear in background and efficiency contexts but are not load-bearing in the proof of Theorem 1. The abstract's 'equivalent mapping' language is arguably an overstatement because the exact relation is nonlinear and the first-order equivalence is local, but overstatement is a correctness concern, not a circularity. Therefore no circular step is present.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical or mathematical entities. It reuses the Nyström embedding and Charikar rounding. The main load-bearing choices are the landmark count s, the kernel hyperparameters, and the assumption that uniform landmark sampling gives a good Nyström approximation.

free parameters (3)
  • number of landmarks s = max(300, 2% of training data)
    Set by rule rather than tuned per dataset; controls Nyström approximation quality and runtime.
  • kernel hyperparameters (gappy kernel k-mer size and gap, propagation kernel parameters) = not reported in the paper
    Chosen empirically per dataset; the paper does not state whether selection used a validation split, raising a mild overfitting risk.
  • HDC dimension d = not reported in experiments
    Standard HDC hyperparameter, not the focus of this work, but still a user choice affecting the rounding approximation.
assumptions (4)
  • domain assumption The kernel K is positive-definite
    Required for the Nyström pseudo-inverse and the eigen-decomposition in Algorithm 1; stated in Theorem 1.
  • standard math Charikar's sign-random-projection expectation identity (Equation 3)
    Used directly in the proof of Theorem 1.
  • domain assumption Uniform landmark sampling yields a Nyström approximation accurate enough for learning
    The paper uses uniform sampling without replacement and defers adaptive sampling to future work; the theorem does not bound the downstream classification error from approximation error.
  • standard math First-order Taylor expansion of arccos is valid for the range of normalized kernel values
    Used to connect the exact arccos expression to the normalized kernel; the actual algorithm uses the exact expression, so this is an interpretational step, not an algorithmic assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging the Gap Between Hyperdimensional Computing and Kernel Methods via the Nystr\"om Method." pith.science (2026). https://pith.science/paper/FK27WKQL

@misc{pith2026260806860,
  author       = {Pith},
  title        = {Pith review of: Bridging the Gap Between Hyperdimensional Computing and Kernel Methods via the Nystr\"om Method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FK27WKQL}},
  note         = {Machine review of arXiv:2608.06860}
}
read the original abstract

Hyperdimensional computing (HDC) is an approach from the cognitive science literature for solving information processing tasks using data represented as high-dimensional random vectors. The technique has a rigorous mathematical backing, and is easy to implement in energy-efficient and highly parallel hardware like FPGAs and "processing-in-memory" architectures. The effectiveness of HDC in machine learning largely depends on how raw data is mapped to high-dimensional space. In this work, we propose NysHD, a new method for constructing this mapping that is based on the Nystr\"om method from the literature on kernel approximation. Our approach provides a simple recipe to turn any user-defined positive-semidefinite similarity function into an equivalent mapping in HDC. There is a vast literature on the design of such functions for learning problems. Our approach provides a mechanism to import them into the HDC setting, expanding the types of problems that can be tackled using HDC. Empirical evaluation against existing HDC encoding methods shows that NysHD can achieve, on average, 11% and 17% better classification accuracy on graph and string datasets respectively.

Figures

Figures reproduced from arXiv: 2608.06860 by the authors.

Figure 1
Figure 1. Overview of HDC training and inference for [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Numerical difference of spectral norm between normalized kernel matrices compute directly from kernel function [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

81 extracted references · 58 canonical work pages

  1. [1]

    2004 , publisher=

    Kernel methods for pattern analysis , author=. 2004 , publisher=

    2004

  2. [2]

    1998 , publisher=

    Learning with kernels , author=. 1998 , publisher=

    1998

  3. [3]

    Efficient biosignal processing using hyperdimensional computing: Network templates for combined learning and classification of

    Rahimi, Abbas and Kanerva, Pentti and Benini, Luca and Rabaey, Jan M , journal =. Efficient biosignal processing using hyperdimensional computing: Network templates for combined learning and classification of. 2018 , publisher =

    work page 2018

  4. [4]

    TBCAS , year=

    A highly energy-efficient hyperdimensional computing processor for biosignal classification , author=. TBCAS , year=

    work page

  5. [5]

    2018 , publisher=

    Introduction to the theory of neural computation , author=. 2018 , publisher=

    2018

  6. [6]

    JMLR , volume=

    On the influence of the kernel on the consistency of support vector machines , author=. JMLR , volume=

    work page

  7. [7]

    IEEE Transactions on Neural networks , volume=

    Holographic reduced representations , author=. IEEE Transactions on Neural networks , volume=. 1995 , publisher=

    work page 1995

  8. [8]

    Cognitive computation , volume=

    Hyperdimensional computing: An introduction to computing in distributed representation with high-dimensional random vectors , author=. Cognitive computation , volume=. 2009 , publisher=

    2009

Show all 81 references
  1. [9]

    JAIR , volume=

    A theoretical perspective on hyperdimensional computing , author=. JAIR , volume=

  2. [10]

    DAC , pages=

    GENERIC: highly efficient learning engine on edge using hyperdimensional computing , author=. DAC , pages=

  3. [11]

    MobiCom , pages=

    FedHD: federated learning with hyperdimensional computing , author=. MobiCom , pages=

  4. [12]

    DATE , pages=

    Hydrea: Towards more robust and efficient machine learning systems with hyperdimensional computing , author=. DATE , pages=. 2021 , organization=

    2021

  5. [13]

    ISLPED , pages=

    Shear er: highly-efficient hyperdimensional computing by software-hardware enabled multifold approximation , author=. ISLPED , pages=

  6. [14]

    GLSVLSI , pages=

    Hdnn-pim: Efficient in memory design of hyperdimensional computing with feature extraction , author=. GLSVLSI , pages=

  7. [15]

    ASP-DAC , pages=

    Xcelhd: An efficient gpu-powered hyperdimensional computing with parallelized training , author=. ASP-DAC , pages=. 2022 , organization=

    2022

  8. [16]

    On the Nystr

    Drineas, Petros and Mahoney, Michael W and Cristianini, Nello , journal=. On the Nystr

  9. [17]

    Machine learning , volume=

    Propagation kernels: efficient graph kernels from propagated information , author=. Machine learning , volume=. 2016 , publisher=

    2016

  10. [18]

    Biocomputing 2002 , pages=

    The spectrum kernel: A string kernel for SVM protein classification , author=. Biocomputing 2002 , pages=. 2001 , publisher=

    2002

  11. [19]

    Kernel methods in machine learning , author=

  12. [20]

    2008 IEEE International Joint Conference on Neural Networks , pages=

    Support vector machines and dynamic time warping for time series , author=. 2008 IEEE International Joint Conference on Neural Networks , pages=. 2008 , organization=

    2008

  13. [21]

    BHI , pages=

    Hdna: Energy-efficient dna sequencing using hyperdimensional computing , author=. BHI , pages=. 2018 , organization=

    2018

  14. [22]

    EMBC , pages=

    Detection of epileptic seizures from surface eeg using hyperdimensional computing , author=. EMBC , pages=. 2020 , organization=

    2020

  15. [23]

    , pages=

    Computing on Functions Using Randomized Vector Representations , author=. , pages=

  16. [24]

    NeurIPS , volume=

    Random features for large-scale kernel machines , author=. NeurIPS , volume=

  17. [25]

    DATE , pages=

    Onlinehd: Robust, efficient, and single-pass online learning using hyperdimensional system , author=. DATE , pages=. 2021 , organization=

    2021

  18. [26]

    DATE , pages=

    GraphHD: Efficient graph classification using hyperdimensional computing , author=. DATE , pages=. 2022 , organization=

    2022

  19. [27]

    Computer networks and ISDN systems , volume=

    The anatomy of a large-scale hypertextual web search engine , author=. Computer networks and ISDN systems , volume=. 1998 , publisher=

    1998

  20. [28]

    2012 , howpublished =

    Almeida,Tiago and Hidalgo,Jos , title =. 2012 , howpublished =

    2012

  21. [29]

    BMC bioinformatics , volume=

    Ion-pumping microbial rhodopsin protein classification by machine learning approach , author=. BMC bioinformatics , volume=. 2023 , publisher=

    2023

  22. [30]

    Knowledge and Information Systems , volume=

    Comparison of descriptor spaces for chemical compound retrieval and classification , author=. Knowledge and Information Systems , volume=. 2008 , publisher=

    2008

  23. [31]

    Bioinformatics , volume=

    Protein function prediction via graph kernels , author=. Bioinformatics , volume=. 2005 , publisher=

    2005

  24. [32]

    Proceedings of the IEEE , volume=

    Gradient-based learning applied to document recognition , author=. Proceedings of the IEEE , volume=. 1998 , publisher=

    1998

  25. [33]

    arXiv preprint arXiv:1708.07747 , year=

    Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms , author=. arXiv preprint arXiv:1708.07747 , year=

    arXiv

  26. [34]

    NeurIPS , volume=

    Kernel methods through the roof: handling billions of points efficiently , author=. NeurIPS , volume=

  27. [35]

    TCBB , volume=

    Hardware acceleration of the STRIKE string kernel algorithm for estimating protein to protein interactions , author=. TCBB , volume=. 2021 , publisher=

    2021

  28. [36]

    Sampling methods for the Nystr

    Kumar, Sanjiv and Mohri, Mehryar and Talwalkar, Ameet , journal=. Sampling methods for the Nystr. 2012 , publisher=

    2012

  29. [37]

    DATE , pages=

    FSL-HD: Accelerating Few-Shot Learning on ReRAM using Hyperdimensional Computing , author=. DATE , pages=. 2023 , organization=

    2023

  30. [38]

    Proceedings of the 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays , pages=

    F5-hd: Fast flexible fpga-based framework for refreshing hyperdimensional computing , author=. Proceedings of the 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays , pages=

    2019

  31. [39]

    IEEE Transactions on Computers , volume=

    Openhd: A gpu-powered framework for hyperdimensional computing , author=. IEEE Transactions on Computers , volume=. 2022 , publisher=

    2022

  32. [40]

    TCAD , year=

    : Accelerating Event-based Workloads with HyperDimensional Computing and Spiking Neural Networks , author=. TCAD , year=

  33. [41]

    Using the Nystr

    Williams, Christopher and Seeger, Matthias , journal=. Using the Nystr

  34. [42]

    Quantum Interaction: 10th International Conference, QI 2016, San Francisco, CA, USA, July 20-22, 2016, Revised Selected Papers 10 , pages=

    Language geometry using random indexing , author=. Quantum Interaction: 10th International Conference, QI 2016, San Francisco, CA, USA, July 20-22, 2016, Revised Selected Papers 10 , pages=. 2017 , organization=

    2016

  35. [43]

    HST , pages=

    Hyperdimensional computing encoding schemes for improved image classification , author=. HST , pages=. 2022 , organization=

    2022

  36. [44]

    MICRO , pages=

    Dual: Acceleration of clustering algorithms using digital-based processing in-memory , author=. MICRO , pages=. 2020 , organization=

    2020

  37. [45]

    Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Random projection in dimensionality reduction: applications to image and text data , author=. Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  38. [46]

    Cybernetics and Systems Analysis , volume=

    Formation of similarity-reflecting binary vectors with random binary projections , author=. Cybernetics and Systems Analysis , volume=. 2015 , publisher=

    2015

  39. [47]

    DAC , pages=

    Bric: Locality-based encoding for energy-efficient brain-inspired hyperdimensional computing , author=. DAC , pages=

  40. [48]

    , author=

    The perceptron: a probabilistic model for information storage and organization in the brain. , author=. Psychological review , volume=. 1958 , publisher=

    1958

  41. [49]

    NeurIPS , editor=

    Understanding Hyperdimensional Computing for Parallel Single-Pass Learning , author=. NeurIPS , editor=. 2022 , url=

    2022

  42. [50]

    ACM Computing Surveys , volume=

    A survey on hyperdimensional computing aka vector symbolic architectures, part ii: Applications, cognitive models, and challenges , author=. ACM Computing Surveys , volume=. 2023 , publisher=

    2023

  43. [51]

    Computational intelligence and neuroscience , volume=

    Encoding sequential information in semantic space models: Comparing holographic reduced representation and random permutation , author=. Computational intelligence and neuroscience , volume=. 2015 , publisher=

    2015

  44. [52]

    NeurIPS , volume=

    Dynamic time-alignment kernel in support vector machine , author=. NeurIPS , volume=

  45. [53]

    2010 2nd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing , pages=

    The angular kernel in machine learning for hyperspectral data classification , author=. 2010 2nd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing , pages=. 2010 , organization=

    2010

  46. [54]

    Word , volume=

    Distributional structure , author=. Word , volume=. 1954 , publisher=

    1954

  47. [55]

    arXiv preprint arXiv:1905.04579 , year=

    Are powerful graph neural nets necessary? a dissection on graph classification , author=. arXiv preprint arXiv:1905.04579 , year=

    1905 arXiv

  48. [56]

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

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  49. [57]

    Proceedings of the thiry-fourth annual ACM symposium on Theory of computing , pages=

    Similarity estimation techniques from rounding algorithms , author=. Proceedings of the thiry-fourth annual ACM symposium on Theory of computing , pages=

  50. [58]

    NeurIPS , volume=

    Recursive sampling for the nystrom method , author=. NeurIPS , volume=

  51. [59]

    ASP-DAC , year=

    Fach: Fpga-based acceleration of hyperdimensional computing by reducing computational complexity , author=. ASP-DAC , year=

  52. [60]

    JACM , volume=

    Improved approximation algorithms for maximum cut and satisfiability problems using semidefinite programming , author=. JACM , volume=. 1995 , publisher=

    1995

  53. [61]

    A primer on kernel methods , author=

  54. [62]

    and Reynolds, R

    Harley, C. and Reynolds, R. and Noordewier, M. , title =. 1990 , howpublished =

    1990

  55. [63]

    Journal of molecular biology , volume=

    Distinguishing enzyme structures from non-enzymes without alignments , author=. Journal of molecular biology , volume=. 2003 , publisher=

    2003

  56. [64]

    Journal of chemical information and computer sciences , volume=

    Spline-fitting with a genetic algorithm: A method for developing classification structure- activity relationships , author=. Journal of chemical information and computer sciences , volume=. 2003 , publisher=

    2003

  57. [65]

    correlation with molecular orbital energies and hydrophobicity , author=

    Structure-activity relationship of mutagenic aromatic and heteroaromatic nitro compounds. correlation with molecular orbital energies and hydrophobicity , author=. Journal of medicinal chemistry , volume=. 1991 , publisher=

    1991

  58. [66]

    Journal of medicinal chemistry , volume=

    Derivation and validation of toxicophores for mutagenicity prediction , author=. Journal of medicinal chemistry , volume=. 2005 , publisher=

    2005

  59. [67]

    SSPR & SPR 2008, Orlando, USA, December 4-6, 2008

    IAM graph database repository for graph based pattern recognition and machine learning , author=. SSPR & SPR 2008, Orlando, USA, December 4-6, 2008. Proceedings , pages=. 2008 , organization=

    2008

  60. [68]

    AAAI , volume=

    An end-to-end deep learning architecture for graph classification , author=. AAAI , volume=

  61. [69]

    arXiv preprint arXiv:1810.00826 , year=

    How powerful are graph neural networks? , author=. arXiv preprint arXiv:1810.00826 , year=

    arXiv

  62. [70]

    Expert Systems with Applications , volume=

    Graph isomorphism U-Net , author=. Expert Systems with Applications , volume=. 2024 , publisher=

    2024

  63. [71]

    , author=

    Fast string kernels using inexact matching for protein sequences. , author=. JMLR , volume=

  64. [72]

    Proceedings of the National Academy of Sciences , volume=

    Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences , author=. Proceedings of the National Academy of Sciences , volume=. 2021 , publisher=

    2021

  65. [73]

    arXiv preprint arXiv:1810.04805 , year=

    Bert: Pre-training of deep bidirectional transformers for language understanding , author=. arXiv preprint arXiv:1810.04805 , year=

    arXiv

  66. [74]

    Science China technological sciences , volume=

    Pre-trained models for natural language processing: A survey , author=. Science China technological sciences , volume=. 2020 , publisher=

    2020

  67. [75]

    JMLR , volume=

    Exploring the limits of transfer learning with a unified text-to-text transformer , author=. JMLR , volume=

  68. [76]

    arXiv preprint arXiv:1711.05101 , year=

    Decoupled weight decay regularization , author=. arXiv preprint arXiv:1711.05101 , year=

    arXiv

  69. [77]

    arXiv preprint arXiv:2007.08663 , year=

    Tudataset: A collection of benchmark datasets for learning with graphs , author=. arXiv preprint arXiv:2007.08663 , year=

    2007 arXiv

  70. [78]

    arXiv preprint arXiv:2007.13462 , year=

    A short letter on the dot product between rotated fourier transforms , author=. arXiv preprint arXiv:2007.13462 , year=

    2007 arXiv

  71. [79]

    IEEE Journal of Selected Topics in Signal Processing , year=

    The Hyperdimensional Transform: a Holographic Representation of Functions , author=. IEEE Journal of Selected Topics in Signal Processing , year=

  72. [80]

    PAKDD 2010, Hyderabad, India, June 21-24, 2010

    Normalized kernels as similarity indices , author=. PAKDD 2010, Hyderabad, India, June 21-24, 2010. Proceedings. Part II 14 , pages=. 2010 , organization=

    2010

  73. [81]

    Primate splice-junction gene sequences (DNA) with associated imperfect domain theory , author=

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.