Your new experience awaits. Try the new design now and help us make it even better

BRIEF RESEARCH REPORT article

Front. Appl. Math. Stat., 03 September 2025

Sec. Dynamical Systems

Volume 11 - 2025 | https://doi.org/10.3389/fams.2025.1625802

This article is part of the Research TopicInnovative Applications of Applied Mathematics in Solving Real-World ChallengesView all 6 articles

Can associative memory be modeled by quantum information?

  • Rabdan Academy, Abu Dhabi, United Arab Emirates

Associative memory is the ability to reveal similarities between unrelated items. Models of associative memory typically rely on significant assumptions about information encoding procedure, structure of underlying complex network, computational power of nodes, and communication capacity of links. Keeping assumptions plausible and at minimum, the search for association can be done on a network enhanced with quantum information processing exploiting non-locality and quantum state comparison.

Introduction

The only definite thing we know about associative memory is that the brain does it much better than we can simulate it [1]. Many advanced models that incorporate different features of associative memory have been suggested to exploit classical [2] and quantum information processing [3]. Arguably, the most successful model so far—dense associative memory by Krotov and Hopfield [4]—powered by condensed matter physics methods and non-linear dynamics seems structurally and computationally excessive compared to neurons firing almost identical electrical signals, getting the job done quickly and at low energy cost [5].

Typically, associative memory models draw on the full arsenal of physics, computer science, and engineering to incorporate as much data from neuroscience as possible [see for example [6] and references therein]. In this Brief Report, I take a different approach. Rather than attempt to mimic actual brain functionality, I look for the minimum requirements to establish an abstract association.

At a most philosophical level, the search for similarity implies comparing two or more instances of information, which could be distantly located. When formulated this way, it seems natural to engage such quantum information processing features as quantum non-locality and direct quantum state comparison [7] to search for association. While non-locality makes physical location of the information units irrelevant, the direct state comparison gives a natural distance measure to conclude on similarity.

In the following, I consider a (classical) network enhanced with quantum information processing and show that the search for association can be done making the least assumptions about the structure of the network, the processing capabilities of the nodes, and the communication capacity of the links.

Results

Suppose we are given a connected network comprising processing and storage units (nodes) and communication channels (links). Information is stored on the nodes in some form, for example, as a bit string x = {0, 1}n of size n. Information may be stored on a single node or, in case of limited node's capacity, may be distributed over a cluster of nodes. In the latter case, it is reasonable to assume a uniform rule of how to store distributed information so that nodes in two clusters of similar content also carry similar pieces of information. For example, cluster A consists of five nodes xi and i = 1...5 sharing a unit of information. Cluster B is similar to cluster A if contains xi so that xi is similar to xi, i.e., Δ(xi,xi)ε for all i in a cense of some distance measure Δ. The clusters may have different size and connectivity, but the five strings i must be pairwise similar; otherwise, the problem to find association becomes ill-defined.

The nodes can communicate information over the links they share with their neighbors. Let me first suppose the nodes can encode all information they possess into an electromagnetic signal and communicate it. Then, the signals in the network are to be all different in their amplitudes and frequencies. Massive data exchange of all-different signals would be energy inefficient. Interestingly, this assumption contradicts evidence of brain operation [1]. Rather, the brain keeps data traffic low using almost identical signals. Therefore, let me assume the signals communicated over the links are to be (identical) quantum photonic states accompanied with at most few bits of classical information.

Let me also assume non-classical information processing in the nodes. Based on the classical information encoded into bit strings xi, the nodes can generate strings of quantum states—qubits |xi. There is no physical limitation on the number of quantum copies to generate from a piece of classical information [7]. In addition, each node must be capable to generate entangled states of photon to send one photon of the pair to its neighboring node through the link, keeping the other.

The above setup is, in fact, sufficient to search for association. Node Alice has information encoded in quantum state |ψ〉A. Having shared an entangled pair with node Bob, Alice teleports her state |ψ〉A to Bob's location sending two classical bits of information per qubit with the next entangled photon as required by the teleportation protocol (see Methods). Roughly speaking, instead of communicating information itself Alice creates a quantum image of information she posses and teleports it to Bob for comparison.

Bob can directly compare his quantum state |ψ〉B with that received from Alice using natural distance measures of quantum theory. There are many slightly different ways for Bob to compare the two states he has: direct state comparison, projective measurement or positive operator valued measure (POVM) [8]. Ultimately, Bob gets single number, which represents the degree of similarity (distance) between |ψ〉A and |ψ〉B, i.e., fidelity F=|ψA|ψB|2.

If fidelity is closer to unity, |ψ〉A and |ψ〉B are similar, meaning that Alice and Bob possess similar classical information. Imperfections in entangled pair generation and transmission may require repeating above procedure, which, however, does not affect the conclusion about similarity of the quantum states and corresponding classical information.

The search looks almost the same if Alice and Bob are not directly connected with a link because quantum networks permit creating non-local links with distant nodes [9, 10] by entanglement swapping [7]. Alice shares an entangled state with her neighbor Charlie, who is also neighbor for Bob, while Alice and Bob are not directly connected. Charlie separately checks similarity of his state |ψ〉C with Alice's state |ψ〉A and Bob's state |ψ〉B and if no similarity found, he performs swapping on entangled qubits received from Alice and Bob making them share an entangled pair, which may be understood as a virtual link [7]. Charlie tells Alice and Bob from which link he received the entangled pairs he swapped, so both Alice and Bob know the path they need to communicate in case their states are found to be similar.

In fact, any two nodes in the network can compare quantum copies having as many entanglement swapping on the way as the distance between them. Luckily, complex highly connected networks often exhibits small-world structure, i.e., the average distance between two nodes growth logarithmically with the network size [11], so it can be safely assumed in this model that the distance between any two nodes is relatively small, making multiple entanglement swapping practical.

Discussion

If Alice and Bob both belong to distant clusters and found similarity between their information, they may instruct other members of their clusters to check for similarity. Taking into account the arsenal of quantum information processing, comparing two clusters for similarity does not seem a problem, rather what is the optimal way, in terms of information encoding and the distance measure choice? Clusters of course may be of different size, but is it better to use just average fidelity for most similar information or have some sort of state mixing before comparing, e.g., quantum hashing?

The model as it is presented here relies on two fundamentally non-classical features: the possibility to compare two quantum states without their direct communication (teleportation), and fundamental non-locality of quantum networks that allows us to connect two distant nodes. Due to extreme simplicity of the suggested model, there are indeed many technical aspects to speculate about. An interesting option is to use POVM to establish similarity. Some sets of POVM are compatible, i.e., can be measured together with no effect on each other, while the others are not. This unique feature of quantum measurements can be of great use in models for associative memory. Consider two POVM measurements on a qubit A±(μ)=12(I±μσx) and B±(μ)=12(I±μσz), where 1−μ ∈ [0, 1] describes noisy spin measurements along x and z directions [8]. The measurements A±(μ) and B±(μ) are compatible, that is, jointly measurable for μ12, while they are not compatible, that is, influence outcomes of each other for μ>12. Bob may vary the parameter μ depending on the number of entanglement swapping that separates him from Alice. The purpose is to target the distance to search for similarity. If, for example, the distance is less than three, Bob keeps POVM compatible; otherwise, he tunes μ to the regime, when POVM are not compatible.

Recent experimental advances in single-photon manipulations and detection [1217] make the proof-of-principle experiment feasible, but understanding the collective dynamics of a complex quantum network [18, 19] with associative memory functionality requires the development of an accurate theoretical model that incorporates quantum information processing and machine learning [20, 21] with the latest neuroscientific findings.

Interestingly, synchronization in electromagnetic spikes in the brain [1] fits into the model because synchronization in entangled states exchange reduces the requirement to store entanglement before teleportation. The quantum state of a cluster may be repeatedly modified by a single node, which may be a reason to overwrite the classical memory [10]. Direct state comparison may be used not just for association search but for quantum machine learning [22]. Noise could make some paths more or less favorable for association [23]. It could even be that quantum information processing on network [24] may actually occur in the brain [25] and, exactly due to its spooky nature, may help demystify the origin of consciousness.

Methods

The four maximally entangled two-qubit states form Bell basis and are given by

|ψ±=|00±|11 2,             |ϕ±=|01±|10 2.

To teleport a qubit state |κ〉, Alice shares with Bob state |ψ±AB, so that each of then hold one qubit of the entangled pair. Then, Alice measures |κ〉 and her entangled qubit in the Bell basis and communicates result of her measurement to Bob via a classical channel.

Entanglement swapping is performed by Charlie, who shares entangled states both with Alice |ψ±AC and Bob |ψ±CB. Charlie measures his qubits from the entangled pairs in the Bell basis and communicates the result of his measurement to the other parties. As a result, Alice and Bob share an entangled state, while Charlie is excluded from their communication.

I thank Marc Timme, Jacob Sherson, and Mile Gu for discussions.

Data availability statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Author contributions

MS: Writing – review & editing, Writing – original draft.

Funding

The author(s) declare that no financial support was received for the research and/or publication of this article.

Conflict of interest

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that no Gen AI was used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

1. Lindsay G. Models of the Mind: How Physics, Engineering and Mathematics Have Shaped Our Understanding of the Brain. New York, NY: Bloomsbury Sigma (2021). doi: 10.5040/9781472966445

PubMed Abstract | Crossref Full Text | Google Scholar

2. Steinberg J, Sompolinsky H. Associative memory of structured knowledge. Sci Rep. (2022) 12:21808. doi: 10.1038/s41598-022-25708-y

PubMed Abstract | Crossref Full Text | Google Scholar

3. Ventura D, Martinez T. Quantum associative memory. arXiv preprint arXiv:quant-ph/9807053. doi: 10.48550/arXiv.quant-ph/9807053

Crossref Full Text | Google Scholar

4. Krotov D, Hopfield J. Dense associative memory for pattern recognition. Neural Inf Process Syst. (2016) 29:1172–80. Available online at: https://proceedings.neurips.cc/paper_files/paper/2016/file/eaae339c4d89fc102edd9dbdb6a28915-Paper.pdf

Google Scholar

5. Cutsuridisa V, Wennekers T. Hippocampus, microcircuits and associative memory. Neural Netw. (2009) 22:1120–8. doi: 10.1016/j.neunet.2009.07.009

PubMed Abstract | Crossref Full Text | Google Scholar

6. Behera AK, Rao M, Sastry S, Vaikuntanathan S. Enhanced associative memory, classification, and learning with active dynamics. Physical Review X. (2023) 13:41043. doi: 10.1103/PhysRevX.13.041043

Crossref Full Text | Google Scholar

7. Nielsen MA, Chuang IL. Quantum Computation and Quantum Information. Cambridge University Press: Cambridge (2000).

Google Scholar

8. Gühne O, Haapasalo E, Kraft T, Pellonpää J-P, Uola R. Incompatible measurements in quantum information science. Rev Mod Phys. (2023) 95:11003. doi: 10.1103/RevModPhys.95.011003

Crossref Full Text | Google Scholar

9. Siomau M. Structural complexity of quantum networks. AIP Conf Proc. (2016) 1742:30017. doi: 10.1063/1.4953138

Crossref Full Text | Google Scholar

10. Siomau M. Any quantum network is structurally controllable by a single driving signal. Quantum Inf Process. (2019) 18:1. doi: 10.1007/s11128-018-2112-6

Crossref Full Text | Google Scholar

11. Bassett DS, Bullmore ET. Small-world brain networks revisited. Neuroscientist. (2017) 23:499–516. doi: 10.1177/1073858416667720

PubMed Abstract | Crossref Full Text | Google Scholar

12. Shen HZ, Shang C, Zhou YH, Yi XX. Unconventional single-photon blockade in non-Markovian systems. Phys Rev A. (2018) 98:23856. doi: 10.1103/PhysRevA.98.023856

Crossref Full Text | Google Scholar

13. Shen HZ, Wang Q, Wang J, Yi XX. Nonreciprocal unconventional photon blockade in a driven dissipative cavity with parametric amplification. Phys Rev A. (2020) 101:13826. doi: 10.1103/PhysRevA.101.013826

Crossref Full Text | Google Scholar

14. Yang JF, Shen HZ. Exceptional-point-engineered dispersive readout of a driven three-level atom weakly interacting with coupled cavities in non-Markovian environments. Phys Rev A. (2024) 109:53712. doi: 10.1103/PhysRevA.109.053712

Crossref Full Text | Google Scholar

15. Shen HZ, Yang JF, Yi XX. Unconventional photon blockade with non-Markovian effects in driven dissipative coupled cavities. Phys Rev A. (2024) 109:43714. doi: 10.1103/PhysRevA.109.043714

Crossref Full Text | Google Scholar

16. Li ZY, Shen HZ. Non-Markovian dynamics with a giant atom coupled to a semi-infinite photonic waveguide. Phys Rev A. (2024) 109:23712. doi: 10.1103/PhysRevA.109.023712

Crossref Full Text | Google Scholar

17. Zhou Y-H, Liu T, Su Q-P, Zhang X-Y, Wu Q-C, Chen D-X, et al. Universal photon blockade. Phys Rev Lett. (2025) 134:183601. doi: 10.1103/PhysRevLett.134.183601

PubMed Abstract | Crossref Full Text | Google Scholar

18. Gyongyosi L, Imre S. Advances in the quantum internet. Commun ACM. (2022) 65:52–63. doi: 10.1145/3524455

Crossref Full Text | Google Scholar

19. Gyongyosi L, Imre S. Networked quantum services. Quantum Inf Comput. (2025) 25:97–140. doi: 10.2478/qic-2025-0006

Crossref Full Text | Google Scholar

20. Gyongyosi L, Imre S. Optimizing high-efficiency quantum memory with quantum machine learning for near-term quantum devices. Sci Rep. (2020) 10:135. doi: 10.1038/s41598-019-56689-0

PubMed Abstract | Crossref Full Text | Google Scholar

21. Gyongyosi L, Imre S. Scalable distributed gate-model quantum computers. Sci Rep. (2021) 11:5172. doi: 10.1038/s41598-020-76728-5

PubMed Abstract | Crossref Full Text | Google Scholar

22. Siomau M. A quantum model for autonomous learning automata. Quantum Inf Process. (2014) 13:1211–21. doi: 10.1007/s11128-013-0723-5

Crossref Full Text | Google Scholar

23. Gardiner C, Zoller P. Quantum Noise, 3rd ed. Springer: New York (2004).

PubMed Abstract | Google Scholar

24. Nokkala J, Piilo J, Bianconi G. Complex quantum networks: a topical review. J Phys A. (2024) 57:233001. doi: 10.1088/1751-8121/ad41a6

Crossref Full Text | Google Scholar

25. Lambert N, Chen Y-N, Cheng Y-C, Li C-M, Chen G-Y, Nori F. Quantum biology. Nat Phys. (2013) 9:10–8. doi: 10.1038/nphys2474

Crossref Full Text | Google Scholar

Keywords: association, memory, quantum information, entanglement, network

Citation: Siomau M (2025) Can associative memory be modeled by quantum information? Front. Appl. Math. Stat. 11:1625802. doi: 10.3389/fams.2025.1625802

Received: 09 May 2025; Accepted: 13 August 2025;
Published: 03 September 2025.

Edited by:

Priyambada Tripathi, Vidyashilp University, India

Reviewed by:

H. Z. Shen, Northeast Normal University, China
Laszlo Gyongyosi, Budapest University of Technology and Economics, Hungary

Copyright © 2025 Siomau. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Michael Siomau, bXNpb21hdUByYS5hYy5hZQ==

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.