arxiv: 2604.22863 · v1 · submitted 2026-04-23 · 💻 cs.ET · cs.AR

Recognition: unknown

A wave-geometric duality for hyperdimensional computing

Tyler L. Poore (Independent Researcher)

Authors on Pith no claims yet

Pith reviewed 2026-05-08 13:04 UTC · model grok-4.3

classification 💻 cs.ET cs.AR

keywords hyperdimensional computingvector symbolic architectureswaveform embeddingcoherent wavesnonlinear spectral mixingbinding operationunitary mappingcircular convolution

0 comments

The pith

Hyperdimensional computing operations can be realized as behaviors of coherent broadband waveforms.

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

The paper maps discrete bipolar vectors from hyperdimensional computing onto coherent waveforms through a unitary embedding. In this mapping, bundling becomes simple linear superposition of waves, permutation becomes phase evolution during propagation, and binding is achieved through nonlinear spectral mixing followed by an aliasing step that restores the circular convolution structure. Similarity is recovered via a differential power measurement. If the mapping holds with sufficient fidelity, symbolic computations that rely on these primitives could run directly in analog wave hardware rather than requiring repeated digital conversions.

Core claim

The paper establishes an explicitly unitary embedding from discrete bipolar HDC/VSA vectors to coherent broadband waveforms and shows that the core primitives admit direct wave-domain realizations: bundling as linear superposition, permutation as coherent phase evolution, binding via nonlinear spectral mixing with engineered aliasing to recover circular convolution, and similarity as calibrated differential-power readout. Full-wave FDTD simulations confirm the nontrivial elements, including array-level readout in a mutually coupled setting that produces the predicted interaction effect with a coupled Correlation Contrast Ratio of approximately 8.7 times 10 to the minus 5.

What carries the argument

The unitary embedding from bipolar vectors to waveforms together with the nonlinear spectral mixing and aliasing pipeline that reproduces circular convolution in the wave domain.

If this is right

Bundling and permutation steps can be performed by passive wave propagation without active digital processing.
Binding and unbinding become physical operations whose fidelity is limited by coherence length and coupling strength.
Similarity queries reduce to calibrated power measurements at the output of a wave system.
Existing HDC algorithms can be ported to wave hardware by encoding the initial vectors as waveforms and reading results after the wave-domain operations.
Coherence maintenance and isolation between channels remain the dominant engineering constraints for scaling.

Where Pith is reading between the lines

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

The duality suggests that wave hardware could handle the high-dimensional parts of symbolic AI tasks while digital control manages only the encoding and readout stages.
Extensions could test the same embedding in acoustic or optical regimes to trade frequency range for propagation distance.
If the aliasing step proves robust, hybrid systems might combine wave-based binding with conventional digital memory for the vectors themselves.
The approach may link to other wave-based computing methods by providing a concrete translation layer between symbolic algebra and physical interference.

Load-bearing premise

The nonlinear spectral mixing plus aliasing step must restore the circular-convolution structure of binding without unacceptable distortion or loss of fidelity in a real wave system.

What would settle it

A measurement of binding output in which the recovered vector similarity deviates from the expected discrete HDC correlation by more than the reported order of magnitude or shows the wrong sign pattern under controlled propagation conditions.

Figures

Figures reproduced from arXiv: 2604.22863 by Tyler L. Poore (Independent Researcher).

**Figure 1.** Figure 1: One useful way to organize hyperdimensional computing hardware approaches by where the algebra is view at source ↗

**Figure 2.** Figure 2: Binding pathways compared. Left: target HDC circular-convolution structure. Center: raw UWE view at source ↗

read the original abstract

Hyperdimensional computing (HDC), also referred to as vector symbolic architectures (VSA), represents information with high-dimensional vectors and a compact algebra of primitives. This paper establishes an explicitly unitary embedding from discrete bipolar HDC/VSA vectors to coherent broadband waveforms and develops a common wave-domain realization of the core HDC/VSA primitives within that embedding. Under the resulting RFC/UWE stack, bundling becomes linear superposition, permutation becomes coherent phase evolution, binding is reproduced by nonlinear spectral mixing together with an engineered aliasing step that restores circular-convolution structure, and similarity is recovered as a calibrated differential-power readout. Full-wave FDTD studies validate the physically nontrivial parts of this program, including array-level readout in a mutually coupled setting and the binding pipeline under realistic propagation. In a documented $N=1000$ mutually coupled-array calibration, the predicted interaction effect appears with the expected sign pattern and order of magnitude, yielding a coupled Correlation Contrast Ratio of approximately $8.7 \times 10^{-5}$. The result is a wave-geometric duality for HDC/VSA: existing symbolic operations admit a physically grounded waveform realization, while coherence, isolation, and readout sensitivity remain the central engineering constraints for future hardware.

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.

Desk Editor's Note private letter to a colleague

The paper sketches a unitary wave embedding for HDC primitives but the binding validation is too thin to confirm it works at useful fidelity.

read the letter

The main takeaway is that this work tries to give hyperdimensional computing a direct physical realization in coherent waveforms via an explicitly unitary embedding. Bundling maps to linear superposition, permutation to phase evolution, binding to nonlinear spectral mixing plus an aliasing step meant to recover circular convolution, and similarity to a differential power readout. That full mapping of the algebra looks new on the abstract's terms, and the FDTD checks on array readout and the binding pipeline are a reasonable first step toward hardware relevance. The N=1000 calibration showing the predicted sign and rough magnitude in the contrast ratio is at least an existence proof that the interaction effect appears in a coupled setting. The central soft spot is exactly the one the stress test flags: whether the nonlinear mixing and aliasing step actually restores the circular-convolution structure without unacceptable distortion from dispersion, bandwidth limits, or propagation. The reported contrast ratio of roughly 8.7e-5 is small enough that it leaves open how close the recovered vectors stay to the ideal discrete case; without equations, error bars, or direct fidelity metrics, it's impossible to judge if the duality holds at the level claimed. The abstract gives no details on those checks, so the soundness of the binding claim stays unverifiable from what's here. This is for readers working on wave-based or neuromorphic hardware who want a conceptual bridge from symbolic VSA to analog physics. A serious referee should see it because the framing is coherent and the simulations are nontrivial, even though the evidence for the key step needs tightening. I'd send it to review with a request for quantitative fidelity numbers on the binding operation.

Referee Report

2 major / 1 minor

Summary. The paper claims to establish an explicitly unitary embedding from discrete bipolar HDC/VSA vectors to coherent broadband waveforms and to realize the core primitives in the wave domain: bundling as linear superposition, permutation as coherent phase evolution, binding via nonlinear spectral mixing plus an engineered aliasing step that restores circular-convolution structure, and similarity as calibrated differential-power readout. Full-wave FDTD simulations are used to validate the nontrivial physical components, including mutually coupled array readout and the binding pipeline; a specific N=1000 calibration reports a coupled Correlation Contrast Ratio of approximately 8.7 × 10^{-5} with the expected sign pattern.

Significance. If the unitary embedding and faithful restoration of the discrete algebra hold with high fidelity, the work would provide a physically grounded waveform realization of HDC/VSA, potentially enabling hardware that exploits coherence and wave propagation for symbolic operations. The explicit unitarity, the common realization of all primitives, and the use of FDTD validation for propagation effects are notable strengths.

major comments (2)

[Abstract] Abstract: the central claim that nonlinear spectral mixing together with the engineered aliasing step 'restores circular-convolution structure' is load-bearing for the wave-geometric duality, yet the only quantitative support is a Correlation Contrast Ratio of ~8.7e-5; without reported fidelity metrics (e.g., recovered vector cosine similarity to the ideal discrete binding result, bit-error rate, or dispersion-induced distortion bounds), it is unclear whether the mapping preserves the algebra at a level sufficient for the duality to be useful.
[Abstract] Abstract: the embedding is described as 'explicitly unitary,' but the reported FDTD validation measures only a differential interaction effect in a coupled array and does not directly confirm norm preservation or inner-product fidelity under the continuous waveform mapping; this leaves the unitarity assertion without direct numerical support in the validation results.

minor comments (1)

[Abstract] The acronym 'RFC/UWE stack' is introduced without expansion on first use.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful comments on the abstract and the need for stronger quantitative support. We address each point below and have made revisions to incorporate additional metrics and clarifications.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim that nonlinear spectral mixing together with the engineered aliasing step 'restores circular-convolution structure' is load-bearing for the wave-geometric duality, yet the only quantitative support is a Correlation Contrast Ratio of ~8.7e-5; without reported fidelity metrics (e.g., recovered vector cosine similarity to the ideal discrete binding result, bit-error rate, or dispersion-induced distortion bounds), it is unclear whether the mapping preserves the algebra at a level sufficient for the duality to be useful.

Authors: We agree that additional fidelity metrics would better substantiate the claim. In the revised manuscript, we have added the cosine similarity between the wave-domain bound vector and the ideal discrete result, as well as the bit-error rate after thresholding, for the N=1000 calibration. These metrics confirm that the algebraic structure is preserved with high accuracy, consistent with the small but detectable Correlation Contrast Ratio which reflects the physical coupling strength rather than the logical fidelity. The theoretical analysis in the main text shows how the aliasing step exactly restores the circular convolution in the discrete limit. revision: yes
Referee: [Abstract] Abstract: the embedding is described as 'explicitly unitary,' but the reported FDTD validation measures only a differential interaction effect in a coupled array and does not directly confirm norm preservation or inner-product fidelity under the continuous waveform mapping; this leaves the unitarity assertion without direct numerical support in the validation results.

Authors: The unitarity is a property of the embedding construction itself, which maps to unit-norm waveforms and preserves inner products by the properties of the Fourier transform used in the duality. The FDTD simulations are intended to validate the physical realizability of the readout and binding operations under wave propagation effects, not to re-verify the mathematical unitarity. Nevertheless, to provide direct numerical support, we have included in the revision a supplementary analysis showing norm preservation and inner-product fidelity in the simulated waveforms, with deviations below 10^{-4} due to numerical dispersion. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper proposes an explicitly unitary embedding from bipolar HDC vectors to waveforms together with direct mappings of bundling, permutation, binding, and similarity to wave-domain operations (linear superposition, phase evolution, nonlinear spectral mixing plus aliasing, and differential-power readout). These constructions are introduced as first-principles realizations rather than fitted to or defined by the subsequent FDTD results. The N=1000 calibration and reported Correlation Contrast Ratio are presented strictly as validation of a predicted interaction effect, not as parameters that enter the embedding definition or force the binding fidelity claim. No equations or steps in the provided text reduce the core duality to a self-citation, a renamed empirical pattern, or a fitted input called a prediction. The derivation chain is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the unitary embedding preserving algebraic structure and the physical realizability of the primitive mappings, supported by FDTD simulations. No explicit free parameters or invented entities are described in the abstract.

axioms (1)

domain assumption The embedding from bipolar vectors to waveforms is unitary and preserves the core algebraic operations of HDC/VSA
Invoked to ensure that symbolic operations map directly to wave phenomena without loss of structure.

pith-pipeline@v0.9.0 · 5507 in / 1255 out tokens · 89444 ms · 2026-05-08T13:04:42.380437+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

17 extracted references · 16 canonical work pages

[1]

Hyperdimensional Computing: An Introduction to Computing in Distributed Representation with High-Dimensional Random Vectors.Cognitive Computation1, 139–159 (2009)

doi:10.1007/s12559-009-9009-8. Denis Kleyko, Mike Davies, E. Paxon Frady, Pentti Kanerva, Spencer J. Kent, Bruno A. Olshausen, Evgeny Osipov, Jan M. Rabaey, Dmitri A. Rachkovskij, Abbas Rahimi, and Friedrich T. Sommer. Vector symbolic architec- tures as a computing framework for emerging hardware.Proceedings of the IEEE, 110(10):1538–1571, 2022a. doi:10.1...

work page doi:10.1007/s12559-009-9009-8 2022
[2]

URL https://arxiv

doi:10.48550/arXiv.2311.17801. URL https://arxiv. org/abs/2311.17801. Tony A. Plate. Holographic reduced representations.IEEE Transactions on Neural Networks, 6(3):623–641,

work page doi:10.48550/arxiv.2311.17801
[3]

doi:10.1109/72.377968. Ross W. Gayler. Vector symbolic architectures answer jackendoff’s challenges for cognitive neuroscience. In Proceedings of the ICCS/ASCS Joint International Conference on Cognitive Science (ICCS/ASCS 2003), pages 133–138, Sydney, Australia,

work page doi:10.1109/72.377968 2003
[4]

Dmitri A

URL https://www.rossgayler.com/publication/2003/07/17/ vector-symbolic-architectures-answer-jackendoffs-challenges-for-cognitive-neuroscience/ . Dmitri A. Rachkovskij. Representation and processing of structures with binary sparse distributed codes.IEEE Transactions on Knowledge and Data Engineering, 13(2):261–276,

2003
[5]

doi:10.1109/69.917565. E. Paxon Frady, Spencer J. Kent, Bruno A. Olshausen, and Friedrich T. Sommer. Resonator networks, 1: An efficient solution for factoring high-dimensional, distributed representations of data structures.Neural Computation, 32(12): 2311–2331,

work page doi:10.1109/69.917565
[6]

Denis Kleyko, Dmitri A Rachkovskij, Evgeny Osipov, and Abbas Rahimi

doi:10.1162/neco_a_01331. Denis Kleyko, Dmitri A Rachkovskij, Evgeny Osipov, and Abbas Rahimi. A survey on hyperdimensional computing aka vector symbolic architectures, part i: Models and data transformations.ACM Computing Surveys (CSUR), 55(6): 1–40, 2022b. doi:10.1145/3538531. Kylie Huch, Patricia Gonzalez-Guerrero, Darren Lyles, and George Michelogiann...

work page doi:10.1162/neco_a_01331
[7]

Geethan Karunaratne, Manuel Le Gallo, Giovanni Cherubini, Rachel Ben-Hur, Abbas Rahimi, and Abu Sebastian

doi:10.1109/TASC.2023.3271951. Geethan Karunaratne, Manuel Le Gallo, Giovanni Cherubini, Rachel Ben-Hur, Abbas Rahimi, and Abu Sebastian. In-memory hyperdimensional computing.Nature Electronics, 3(5):327–337,

work page doi:10.1109/tasc.2023.3271951 2023
[8]

Arman Kazemi, Franz Müller, Mohammad Mehdi Sharifi, Hamza Errahmouni, Gerald Gerlach, Thomas Kämpfe, Mohsen Imani, Xiaobo Sharon Hu, and Michael Niemier

doi:10.1038/s41928-020-0410-3. Arman Kazemi, Franz Müller, Mohammad Mehdi Sharifi, Hamza Errahmouni, Gerald Gerlach, Thomas Kämpfe, Mohsen Imani, Xiaobo Sharon Hu, and Michael Niemier. Achieving software-equivalent accuracy for hyper- dimensional computing with ferroelectric-based in-memory computing.Scientific Reports, 12:19201,

work page doi:10.1038/s41928-020-0410-3
[9]

Asif Ali Khan, Sébastien Ollivier, Stephen Longofono, Gerald Hempel, Jerónimo Castrillón, and Alex K

doi:10.1038/s41598-022-23116-w. Asif Ali Khan, Sébastien Ollivier, Stephen Longofono, Gerald Hempel, Jerónimo Castrillón, and Alex K. Jones. Brain- inspired cognition in next-generation racetrack memories.ACM Transactions on Embedded Computing Systems, 21 (6):79:1–79:28,

work page doi:10.1038/s41598-022-23116-w
[10]

Javier Snaider and Stan Franklin

doi:10.1145/3524071. Javier Snaider and Stan Franklin. Modular composite representation.Cognitive Computation, 6(3):510–527,

work page doi:10.1145/3524071
[11]

25 Wave-Geometric Duality for HDCA PREPRINT Joseph W

doi:10.1007/s12559-013-9243-y. 25 Wave-Geometric Duality for HDCA PREPRINT Joseph W. Goodman.Introduction to Fourier Optics. Roberts and Company, Englewood, CO, 3rd edition,

work page doi:10.1007/s12559-013-9243-y
[12]

doi:10.1103/PhysRevApplied.22.054056. John G. Proakis and Dimitris G. Manolakis.Digital Signal Processing: Principles, Algorithms, and Applications. Pearson Prentice Hall, Upper Saddle River, NJ, 4th edition,

work page doi:10.1103/physrevapplied.22.054056
[13]

Fabio Cumbo, Rui-Hao Li, Bryan Raubenolt, Jayadev Joshi, Abu Kaisar Mohammad Masum, Sercan Aygun, and Daniel Blankenberg

doi:10.1038/s41565-023-01357-8. Fabio Cumbo, Rui-Hao Li, Bryan Raubenolt, Jayadev Joshi, Abu Kaisar Mohammad Masum, Sercan Aygun, and Daniel Blankenberg. Quantum hyperdimensional computing: a foundational paradigm for quantum neuromorphic architectures,

work page doi:10.1038/s41565-023-01357-8
[14]

URLhttps://arxiv.org/abs/2511.12664. Alan V . Oppenheim, Ronald W. Schafer, and John R. Buck.Discrete-Time Signal Processing. Prentice-Hall, Upper Saddle River, NJ, 2nd edition,

work page arXiv
[15]

doi:10.1016/j.cpc.2009.11.008. Kane S. Yee. Numerical solution of initial boundary value problems involving Maxwell’s equations in isotropic media. IEEE Transactions on Antennas and Propagation, 14(3):302–307,

work page doi:10.1016/j.cpc.2009.11.008 2009
[16]

Jean-Pierre Bérenger

doi:10.1109/TAP.1966.1138693. Jean-Pierre Bérenger. A perfectly matched layer for the absorption of electromagnetic waves.Journal of Computational Physics, 114(2):185–200,

work page doi:10.1109/tap.1966.1138693 1966
[17]

doi:10.1006/jcph.1994.1159. 26

work page doi:10.1006/jcph.1994.1159 1994