Pith. sign in

REVIEW 4 major objections 5 minor 122 references

Complexity and Stability of Neural Activity Across Aging and Neurodegenerative Disease

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Healthy aging expands the brain's representational space; MCI and Alzheimer's disease collapse it.

desk verdict Large EEG study reports an aging/disease dissociation in dimensionality and Wasserstein stability, but the central dimension–stability coupling is partly a finite-sample consequence of the estimator definitions and the clinical comparisons need age and noise controls. read the letter →

arxiv 2608.05882 v1 pith:MMMBEU6J submitted 2026-08-06 q-bio.NC

Junjie Yu , Jianyu Zhang , Zian Pei , Xue Shi , Yumei Liu , Xin Jiang , Quanying Liu , Yi Guo This is my paper · ORCID
classification q-bio.NC
keywords EEGintrinsicdimensionalityWassersteindistanceneuralstabilityhealthyagingmildcognitiveimpairmentAlzheimer'sdiseaserepresentationalgeometry
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

This paper tries to establish that the brain under a fixed cognitive state does not drift freely but keeps re-entering a condition-specific region of its activity space, and that how stable that region looks over time is governed by how many effective dimensions the activity occupies. If true, the same two numbers—intrinsic dimensionality and Wasserstein distance between time-segment distributions—would describe both normal aging and neurodegeneration on one axis. Healthy aging appears to push representations into a higher-dimensional, less reproducible regime; mild cognitive impairment and Alzheimer's disease push them into a lower-dimensional, abnormally rigid regime. This matters clinically because it turns the question 'is the signal more or less variable?' into 'is the brain's representational geometry expanded or collapsed?', which is a more specific target for tracking cognitive decline.

What carries the argument

The central object is the empirical distribution of EEG activity patterns: per channel, per subject, per condition, the EEG is cut into non-overlapping 1-second windows, each a 100-dimensional vector, and the set of windows is treated as a sample from an underlying distribution. Two descriptors are computed on these distributions. Intrinsic dimensionality, estimated with a maximum-likelihood local estimator, is the effective number of degrees of freedom in the activity patterns. Wasserstein distance, computed by optimal transport, is the minimal transport cost between empirical distributions from different time segments; smaller values mean the same patterns keep reappearing. The link between them is supplied by the empirical Wasserstein convergence theorem, which says the expected distance between independently sampled empirical measures shrinks at a rate controlled by the upper Wasserstein dimension of the underlying measure. That theorem is the load-bearing identity: it makes dimensionality a predictor of stability before any neuroscience assumption is added.

What would settle it

A concrete test is to repeat the clinical comparison with an age- and education-matched healthy control group, keeping the same ICA artifact-removal pipeline and blind automated rejection; the collapse claim would be refuted if lower intrinsic dimensionality and lower Wasserstein distance in MCI/AD disappear once confounds are controlled, and strongly supported if they survive. A second check is computational: simulate EEG noise with known intrinsic dimension and confirm the estimators recover the prescribed dimensionality differences; if they cannot, the group differences could be noise artifacts.

Watch

Extended reading notes

Core claim

The central claim is that neural representations, measured as empirical distributions of 1-second multichannel EEG windows, show constrained condition-specific stability rather than unconstrained drift: within-condition Wasserstein distances stay bounded, do not grow with time, and are smaller than cross-condition distances. The paper further claims that intrinsic dimensionality—the effective number of directions in which the windowed patterns vary—constrains stability in a quantitative way, through the empirical-Wasserstein convergence bound $E W_p(\mu,\hat\mu_n) \le C n^{-1/s}$, where $s$ is tied to the upper Wasserstein dimension of the underlying distribution. Lower-dimensional distributions therefore yield smaller expected Wasserstein distances between repeated samples. Across multi-task, lifespan, and clinical datasets, the authors find a consistent positive coupling: regions and conditions with higher intrinsic dimensionality show larger within-condition displacement. The clinical result is the sharpest claim: healthy aging expands the representation (higher dimension, lower stability), while MCI and AD show a joint collapse (lower dimension, lower Wasserstein distance), which the paper interprets as pathological over-stability rather than preserved function.

Load-bearing premise

The load-bearing premise is that group differences in Wasserstein distance and intrinsic dimensionality reflect the geometry of neural representations themselves rather than differences in age, education, medication, or EEG signal quality across the compared groups.

Editorial extensions

If this is right

  • Within a fixed cognitive state, the brain's activity-pattern distribution is bounded and condition-specific; progressive representational drift is not the dominant mode of dynamics in these EEG datasets.
  • A channel's intrinsic dimensionality predicts its temporal reproducibility: richer representational spaces are less reproducible, and this coupling survives across tasks, age groups, and clinical groups, though with reduced spatial extent in AD.
  • Healthy aging can be described as representational expansion: both dimensionality and within-condition Wasserstein distance increase with age and are positively correlated across channels.
  • MCI and AD can be described as representational collapse: both measures drop relative to healthy controls, and the paper argues the lowered Wasserstein distance should be read as pathological over-stability, not healthy maintenance.
  • The two measures together give a single geometric axis on which aging and neurodegeneration move in opposite directions, which could support longitudinal tracking in the clinic.

Reading between the lines

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

  • A natural next test is longitudinal: does an individual's dimensionality first rise with age and then fall as MCI converts to AD? The paper's cross-sectional design leaves that trajectory open, but its shared geometric axis predicts exactly that ordering.
  • The scalp-level gradient could be checked against source-reconstructed EEG or MEG; if posterior regions remain higher-dimensional with larger Wasserstein distances after source modeling, the result is less likely to be an artifact of volume conduction.
  • Because the clinical groups are compared without age matching, a direct extension is to re-run the HC–MCI–AD comparison against age-, education-, and medication-matched controls; the collapse interpretation would be substantially strengthened if it survives, and weakened if it does not.
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

4 major / 5 minor

Summary. The manuscript proposes a distribution-level framework for EEG analysis in which each channel's activity is represented as an empirical distribution of windowed patterns, stability is quantified by Wasserstein distance between time-segment distributions, and representational complexity is quantified by intrinsic dimensionality. Using three datasets (multi-task, lifespan, and clinical HC/MCI/AD), the authors report that within-condition Wasserstein distance is bounded and condition-specific, that higher intrinsic dimensionality is associated with larger Wasserstein distances across regions and conditions, that healthy aging increases both measures, and that MCI and AD decrease both measures. The paper interprets these joint changes as healthy representational expansion versus pathological collapse/over-stability.

Significance. If the interpretation is valid, the framework offers a clinically relevant, easily computable biomarker candidate for cognitive aging and neurodegeneration, and it provides a unified geometric axis linking dimensionality to stability. The study's strengths include the use of multiple independent EEG datasets, explicit theoretical grounding in empirical-process results, and robustness checks across neighborhood sizes K. However, the central empirical association between dimensionality and Wasserstein distance is, to a substantial degree, a mathematical consequence of the very convergence bound the authors invoke, and the clinical group comparisons lack essential demographic and medication controls. These issues are load-bearing because they directly affect the paper's main claims, so the significance is conditional on addressing them.

major comments (4)
  1. [Section 2.1.4 and Appendix A.5, Theorem A.1] The reported positive association between intrinsic dimensionality and within-condition Wasserstein distance is largely a finite-sample-statistics consequence of the cited Weed-Bach bound, not an independent empirical discovery. For two independent empirical samples of size n from the same distribution μ, the triangle inequality gives E Wp(μ̂_n, ν̂_n) ≤ 2 C n^{-1/s} with s = d*_p(μ)+ε, so the expected within-condition distance must increase with the upper Wasserstein dimension even under the null of i.i.d. sampling. Since the paper does not report the number of windows per segment, the quantitative predictions of this bound cannot be evaluated. To make the 'geometric constraint' claim non-circular, the authors should compare the observed Wasserstein distances against a surrogate null that holds the estimated dimension fixed (e.g., phase-randomized or amplitude-matched data), or explicitly test whether the empirical dimension–stability relationship exceeds the n^{-1/s} scaling predicted by the theorem.
  2. [Section 3.4 and Section 2.1] The clinical comparisons between HC, MCI, and AD do not report or statistically control for age, sex, education, medication, or comorbidities, despite Section 3.3 showing that age alone increases both intrinsic dimensionality and Wasserstein distance. Because AD patients are typically older than healthy controls, the observed 'collapse' of both measures in MCI/AD could be confounded by age, medication effects, or other clinical variables, and the direction of the age effect makes simple intuition unreliable. The authors should report the demographic and clinical characteristics of each group and repeat the group comparisons with age as a covariate or with age-matched subgroups.
  3. [Appendix A.7] The paper asserts that temporal dependence does not undermine the theoretical link, but Theorem A.1 assumes independent samples, whereas consecutive 1-second EEG windows are strongly autocorrelated. The effective number of independent samples is likely far smaller than the number of windows, which can change the constant and effective rate in the bound and may itself differ across age or disease groups. The authors should quantify the effective degrees of freedom (e.g., via block bootstrap or spectral estimates) or provide surrogate analyses that preserve autocorrelation, to confirm that the dimension–stability relationship is not an artifact of group differences in temporal dependence.
  4. [Sections 2.1.2 and Appendix A.4] The empirical estimator of intrinsic dimensionality is the Levina-Bickel nearest-neighbor MLE, while the theoretical bound in Theorem A.1 concerns the upper Wasserstein dimension defined through covering numbers. The manuscript states that the MLE is a 'proxy' but does not establish any formal connection between the two quantities for finite, noisy, temporally dependent EEG samples. Without such a link, invoking Theorem A.1 to interpret the empirical correlations is an unproven assumption; the authors should either prove or cite a quantitative relationship, or soften the theoretical justification accordingly.
minor comments (5)
  1. [Section 2.1] The text says participants were grouped according to the '2011 NIA-AA criteria' but cites Jack Jr et al. (2024), which describes revised criteria; please clarify which criteria version was used and update the citation or text accordingly.
  2. [Figure 1] The panel labels contain unusual spacing artifacts, e.g., 'Te m p o r a lFluctuations' and 'AgingAndNeurodegenerativeDisease'; please typeset the figure text properly.
  3. [Section 3.2] The posterior–anterior gradient is described qualitatively from topographies; providing quantitative summaries (e.g., effect sizes or cluster statistics) would strengthen the claim of reproducibility across conditions.
  4. [Appendix A.2, Definition 3] The definition d_ε(μ,τ) = log N_ε(μ,τ) / (−log ε) is only positive for ε < 1; please state the range of ε for which this definition is intended, and clarify how it behaves as ε → 0.
  5. [Section 2.1.1] The description of window standardization ('after standardization') is ambiguous: clarify whether each window is standardized independently or each channel is standardized globally across the recording, as this affects the interpretation of Wasserstein distances.

Circularity Check

1 steps flagged · score 6.0 of 10

The dimension–stability coupling is partly a mathematical consequence of the paper's own Weed–Bach bound, so the reported correlations do not independently support the central geometric claim.

  1. other [Appendix A.5 (Theorem A.1); Section 2.1.4; Sections 3.2–3.4, Figs. 3–5]
    "This result implies that the expected Wasserstein distance between empirical distributions sampled from the same underlying neural representational distribution is controlled by its intrinsic dimensionality. Lower-dimensional distributions yield smaller expected representational displacement between samples, whereas higher-dimensional distributions admit greater sampling variability."

    Within-condition stability is measured as the Wasserstein distance between two empirical distributions drawn from the same channel and condition. By the paper's own Theorem A.1, E[W_p(mu, mu_hat_n)] <= C n^{-1/s} with s = d*_p(mu)+epsilon; the triangle inequality gives E[W_p(mu_hat_n, nu_hat_n)] <= 2 C n^{-1/s}. With the number of windows n fixed, this expected distance grows as the intrinsic dimension d* grows. Therefore the positive channel-wise, condition-wise, and group-wise correlations between estimated intrinsic dimensionality and Wasserstein distance reported in Figs.

full rationale

The paper's central dimensional–stability association is partially circular: the within-condition Wasserstein distance is computed between empirical distributions from the same recording, and the cited Weed–Bach theorem (Appendix A.5) already implies that the expected such distance increases with the upper Wasserstein dimension for a fixed sample size. Consequently, the observed topographies and correlations in Figs. 3–5 are not independent evidence for the framework; they are in the direction that the theorem predicts by construction. This is not a self-citation issue: the theorem is external and real, and no parameter is fitted and then renamed a prediction. Moreover, the aging/MCI/AD dimensionality differences remain substantive empirical results, although their interpretation is weakened by unaddressed age, education, and medication confounds, which are correctness concerns rather than circularity. The paper also explicitly labels the link a 'statistical tendency,' which somewhat mitigates the circularity. Overall, one core relation reduces to the imported convergence theorem, while the group-level dimensional shifts retain independent content, giving a partial circularity score of 6.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central framework rests on the Weed-Bach convergence theorem and on the assertion that a local MLE estimates the covering dimension; both are imported from outside and connected to autocorrelated EEG data without proof. The clinical interpretations further assume diagnostic validity and comparable signal quality across groups. No new entities are postulated.

free parameters (4)
  • Neighborhood size K for intrinsic dimensionality = 50 (robustness at 10 and 100)
    Chosen by the authors; dimensionality estimates depend on it, although sensitivity analyses suggest stable topographies. It is not fitted to the target results but is a hand-selected hyperparameter.
  • EEG window length = 1 second
    Fixed across datasets; affects both the dimensionality estimate and the Wasserstein distances. No systematic optimization or justification is provided.
  • Segment length for stability analysis = not specified
    Successive non-overlapping time segments are used in Section 3.1 but their duration is not stated, so the temporal scale of 'stability' is underdetermined.
  • Order p of Wasserstein distance = not specified (likely 2)
    The p-Wasserstein distance is defined for general p but the paper does not state which p is used in the POT computations; the choice affects the magnitude and scaling of distances.
assumptions (6)
  • standard math Empirical measure convergence in Wasserstein distance (Weed and Bach, 2019)
    Invoked as Theorem A.1 to connect dimension to sampling variability. This is a published mathematical result, not derived in the paper.
  • domain assumption Levina-Bickel maximum likelihood estimator approximates the measure-based intrinsic dimension
    Section 2.1.2 and Appendix A.7 assert the estimator is an empirical proxy; no proof that it estimates the covering-number dimension of Definition 3 is given.
  • ad hoc to paper Weed-Bach convergence rates hold for temporally dependent EEG windows
    Appendix A.7 extends the i.i.d. theorem to autocorrelated neural signals without proof, claiming weak dependence is acceptable; this is an unsupported generalization.
  • domain assumption Within a fixed condition, EEG windows are samples from a stable underlying distribution
    Appendix A.1 models neural activity this way; the boundedness result in Section 3.1 is used as justification, but the modeling assumption is prior to the analysis.
  • domain assumption Clinical diagnosis and MoCA threshold 26 correctly separate HC, MCI, and AD
    Section 2.1 relies on NIA-AA criteria and clinical evaluation; no biomarker confirmation (amyloid, tau) is reported, and the 2011 criteria are cited to a 2024 reference.
  • domain assumption Reduced Wasserstein distance in AD reflects pathological over-stability rather than data quality or preserved function
    Section 3.4 interprets the reduction as loss of flexibility; alternative explanations (reduced SNR, less data variability, medication effects) are not excluded.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Complexity and Stability of Neural Activity Across Aging and Neurodegenerative Disease." pith.science (2026). https://pith.science/paper/MMMBEU6J

@misc{pith2026260805882,
  author       = {Pith},
  title        = {Pith review of: Complexity and Stability of Neural Activity Across Aging and Neurodegenerative Disease},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MMMBEU6J}},
  note         = {Machine review of arXiv:2608.05882}
}
read the original abstract

Objective: EEG signals fluctuate continuously even within a fixed cognitive state, but an important question is whether the brain still reuses similar activity patterns to represent information over time. Methods: To address this, we model EEG as distributions of windowed activity patterns and quantify their temporal stability using Wasserstein distance, while intrinsic dimensionality captures representational complexity. Results: Across multi-task, lifespan, and clinical EEG datasets, we find that neural representations show constrained, condition-specific stability rather than unconstrained drift. Higher intrinsic dimensionality is consistently associated with lower stability, suggesting that richer representational spaces are less reproducible over time. Both measures exhibit reproducible spatial organization, with posterior regions showing higher dimensionality and lower stability than frontal regions. Healthy aging is characterized by increased dimensionality and reduced stability, whereas mild cognitive impairment and Alzheimer's disease show a joint collapse of both. Conclusions: These findings provide a distribution-level framework for understanding neural stability across cognition, aging, and disease. Significance: This framework offers a principled approach to quantifying neural representational stability, with potential utility as a sensitive biomarker for tracking cognitive aging and neurodegeneration in clinical settings.

Figures

Figures reproduced from arXiv: 2608.05882 by the authors.

Figure 1
Figure 1. Geometric framework for analyzing neural population stability. (A) Traditional Signal View: Moment-to￾moment signal variability makes it difficult to distinguish stable states from noise. (B) Distributional View: Neural activity is conceptualized as probability distributions (clouds) in a high-dimensional space. (C) Geometric Stability: Stability is defined by constrained distributional evolution, allowing for signa… view at source ↗
Figure 2
Figure 2. Constrained distributional evolution of neural representations within fixed cognitive conditions. (A) Wasserstein distance between empirical distributions from successive time segments and the initial reference segment shows bounded, non-monotonic fluctuations over time. (B) Group-level comparison of early and late segments relative to the same initial distribution shows no significant difference in Wasserstein dist… view at source ↗
Figure 3
Figure 3. Intrinsic dimensionality and distributional stability exhibit reproducible spatial organization across cognitive conditions. (A) Whole-brain topography of intrinsic dimensionality across resting-state and task conditions, showing a posterior–anterior gradient with higher values in posterior regions and lower values in frontal regions. (B) Whole-brain topography of within-condition Wasserstein distance across the sam… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Healthy aging is associated with representational expansion and reduced distributional stability. (A) Whole-brain topography of intrinsic dimensionality across age groups. (B) Whole-brain topography of within-condition Wasserstein distance across age groups. (C) Intrin…
Figure 5
Figure 5. Figure 5: Neurodegenerative disease is associated with collapse and rigidification of representational geometry. (A) Whole-brain topography of intrinsic dimensionality in healthy controls (HC), mild cognitive impairment (MCI), and Alzheimer’s disease (AD). (B) Whole-brain topogr…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

122 extracted references · 62 canonical work pages

  1. [1]

    Scientific Data , volume=

    A test-retest resting, and cognitive state EEG dataset during multiple subject-driven states , author=. Scientific Data , volume=. 2022 , publisher=

    2022

  2. [2]

    Scientific Data , volume=

    PEARL-Neuro Database: EEG, fMRI, health and lifestyle data of middle-aged people at risk of dementia , author=. Scientific Data , volume=. 2024 , publisher=

    2024

  3. [3]

    2022 , doi =

    Julie Onton (data) and Scott Makeig (data and curation) and Arnaud Delorme (data and curation) and Dung Truong (curation) and Kay Robbins (curation) , title =. 2022 , doi =

    2022

  4. [4]

    Current opinion in neurobiology , volume=

    Neural population geometry: An approach for understanding biological and artificial neural networks , author=. Current opinion in neurobiology , volume=. 2021 , publisher=

    2021

  5. [5]

    bioRxiv , year=

    Nonlinear manifolds underlie neural population activity during behaviour , author=. bioRxiv , year=

  6. [6]

    bioRxiv , pages=

    Neural manifolds in V1 change with top-down signals from V4 targeting the foveal region , author=. bioRxiv , pages=. 2023 , publisher=

    2023

  7. [7]

    Elife , volume=

    Large-scale neural dynamics in a shared low-dimensional state space reflect cognitive and attentional dynamics , author=. Elife , volume=. 2023 , publisher=

    2023

  8. [8]

    Neural Computation , volume=

    Mapping low-dimensional dynamics to high-dimensional neural activity: A derivation of the ring model from the neural engineering framework , author=. Neural Computation , volume=. 2021 , publisher=

    2021

Show all 122 references
  1. [9]

    Current Biology , volume=

    Low-dimensional spatiotemporal dynamics underlie cortex-wide neural activity , author=. Current Biology , volume=. 2020 , publisher=

    2020

  2. [10]

    Nature neuroscience , volume=

    The intrinsic attractor manifold and population dynamics of a canonical cognitive circuit across waking and sleep , author=. Nature neuroscience , volume=. 2019 , publisher=

    2019

  3. [11]

    Nature , volume=

    Geometry of abstract learned knowledge in the hippocampus , author=. Nature , volume=. 2021 , publisher=

    2021

  4. [12]

    NeuroImage , pages=

    Assessing the effectiveness of spatial PCA on SVM-based decoding of EEG data , author=. NeuroImage , pages=. 2024 , publisher=

    2024

  5. [13]

    Biomedical Signal Processing and Control , volume=

    Characterizing EEG signals of meditative states using persistent homology and Hodge spectral entropy , author=. Biomedical Signal Processing and Control , volume=. 2024 , publisher=

    2024

  6. [14]

    Engineering Applications of Artificial Intelligence , volume=

    T-distributed stochastic neighbor embedding echo state network with state matrix dimensionality reduction for time series prediction , author=. Engineering Applications of Artificial Intelligence , volume=. 2023 , publisher=

    2023

  7. [15]

    , author=

    Visualizing data using t-SNE. , author=. Journal of machine learning research , volume=

  8. [16]

    arXiv preprint arXiv:1802.03426 , year=

    Umap: Uniform manifold approximation and projection for dimension reduction , author=. arXiv preprint arXiv:1802.03426 , year=

    arXiv

  9. [17]

    arXiv preprint arXiv:1312.6114 , year=

    Auto-encoding variational bayes , author=. arXiv preprint arXiv:1312.6114 , year=

    arXiv

  10. [18]

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

    On time-series topological data analysis: New data and opportunities , author=. Proceedings of the IEEE conference on computer vision and pattern recognition workshops , pages=

  11. [19]

    arXiv preprint arXiv:2204.08624 , year=

    Topology and geometry of data manifold in deep learning , author=. arXiv preprint arXiv:2204.08624 , year=

    arXiv

  12. [20]

    arXiv preprint arXiv:2012.13255 , year=

    Intrinsic dimensionality explains the effectiveness of language model fine-tuning , author=. arXiv preprint arXiv:2012.13255 , year=

    2012 arXiv

  13. [21]

    Advances in Neural Information Processing Systems , volume=

    Intrinsic dimension of data representations in deep neural networks , author=. Advances in Neural Information Processing Systems , volume=

  14. [22]

    arXiv preprint arXiv:1906.00443 , year=

    Dimensionality compression and expansion in deep neural networks , author=. arXiv preprint arXiv:1906.00443 , year=

    1906 arXiv

  15. [23]

    arXiv preprint arXiv:2211.16599 , year=

    Compression supports low-dimensional representations of behavior across neural circuits , author=. arXiv preprint arXiv:2211.16599 , year=

    arXiv

  16. [24]

    Current opinion in neurobiology , volume=

    Why neurons mix: high dimensionality for higher cognition , author=. Current opinion in neurobiology , volume=. 2016 , publisher=

    2016

  17. [25]

    Cell Reports , volume=

    Learning to represent continuous variables in heterogeneous neural networks , author=. Cell Reports , volume=. 2022 , publisher=

    2022

  18. [26]

    Elife , volume=

    Contribution of behavioural variability to representational drift , author=. Elife , volume=. 2022 , publisher=

    2022

  19. [27]

    science , volume=

    Nonlinear dimensionality reduction by locally linear embedding , author=. science , volume=. 2000 , publisher=

    2000

  20. [28]

    science , volume=

    A global geometric framework for nonlinear dimensionality reduction , author=. science , volume=. 2000 , publisher=

    2000

  21. [29]

    Advances in neural information processing systems , volume=

    Maximum likelihood estimation of intrinsic dimension , author=. Advances in neural information processing systems , volume=

  22. [30]

    Science Advances , volume=

    Distinct manifold encoding of navigational information in the subiculum and hippocampus , author=. Science Advances , volume=. 2024 , publisher=

    2024

  23. [31]

    Journal of vision , volume=

    Topological analysis of population activity in visual cortex , author=. Journal of vision , volume=. 2008 , publisher=

    2008

  24. [32]

    , title =

    Piczak, Karol J. , title =. Proceedings of the 23rd ACM International Conference on Multimedia , pages =. 2015 , isbn =. doi:10.1145/2733373.2806390 , abstract =

    2015

  25. [33]

    2022 , doi =

    Nur Syairah Ab Rani and Nurfaizatul Aisyah Ab Aziz and Mohammed Farouq Reza and Muzaimi Mustapha , title =. 2022 , doi =

    2022

  26. [34]

    2021 , doi =

    Mojtaba Lahijanian and Mohammad Javad Sedghizadeh and Hamid Aghajan and Zahra Vahabi , title =. 2021 , doi =

    2021

  27. [35]

    2023 , doi =

    Dzianok Patrycja and Antonova Ingrida and Wojciechowski Jakub and Dreszer Joanna and Kublik Ewa , title =. 2023 , doi =

    2023

  28. [36]

    2024 , doi =

    Dzianok Patrycja and Kublik Ewa , title =. 2024 , doi =

    2024

  29. [37]

    Luat and Neena I

    Kazuki Sakakura and Naoto Kuroda and Masaki Sonoda and Takumi Mitsuhashi and Ethan Firestone and Aimee F. Luat and Neena I. Marupudi and Sandeep Sood and Eishi Asano , title =. 2023 , doi =

    2023

  30. [38]

    2022 , doi =

    Yulin Wang and Wei Duan and Debo Dong and Lihong Ding and Xu Lei , title =. 2022 , doi =

    2022

  31. [39]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  32. [40]

    Probability theory and related fields , volume=

    On the shape of the convex hull of random points , author=. Probability theory and related fields , volume=. 1988 , publisher=

    1988

  33. [41]

    Cold Case: The Lost MNIST Digits

    Chhavi Yadav and L\' e on Bottou. Cold Case: The Lost MNIST Digits. Advances in Neural Information Processing Systems 32 , year =

  34. [42]

    ATT Labs [Online]

    MNIST handwritten digit database , author=. ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist , volume=

  35. [43]

    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

  36. [44]

    Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages=

    An analysis of single-layer networks in unsupervised feature learning , author=. Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages=. 2011 , organization=

    2011

  37. [45]

    Alex Krizhevsky , title=

  38. [46]

    Journal of computational neuroscience , volume=

    Neural manifold analysis of brain circuit dynamics in health and disease , author=. Journal of computational neuroscience , volume=. 2023 , publisher=

    2023

  39. [47]

    Neuron , volume=

    Neural manifolds for the control of movement , author=. Neuron , volume=. 2017 , publisher=

    2017

  40. [48]

    PLoS computational biology , volume=

    Estimating the dimensionality of the manifold underlying multi-electrode neural recordings , author=. PLoS computational biology , volume=. 2021 , publisher=

    2021

  41. [49]

    2024 , publisher=

    Radical flexibility of neural representations in service of flexible behaviour , author=. 2024 , publisher=

    2024

  42. [50]

    Neuron , volume=

    A neural representation of prior information during perceptual inference , author=. Neuron , volume=. 2008 , publisher=

    2008

  43. [51]

    Human brain mapping , volume=

    Nonlinear manifold learning in functional magnetic resonance imaging uncovers a low-dimensional space of brain dynamics , author=. Human brain mapping , volume=. 2021 , publisher=

    2021

  44. [52]

    arXiv preprint arXiv:1810.04805 , year=

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

    arXiv

  45. [53]

    AI Open , year=

    GPT understands, too , author=. AI Open , year=

  46. [54]

    Neural networks , volume=

    Independent component analysis: algorithms and applications , author=. Neural networks , volume=. 2000 , publisher=

    2000

  47. [55]

    Tzimourta and Theodora Afrantou and Panagiotis Ioannidis and Nikolaos Grigoriadis and Dimitrios G

    Andreas Miltiadous and Katerina D. Tzimourta and Theodora Afrantou and Panagiotis Ioannidis and Nikolaos Grigoriadis and Dimitrios G. Tsalikakis and Pantelis Angelidis and Markos G. Tsipouras and Evripidis Glavas and Nikolaos Giannakeas and Alexandros T. Tzallas , title =. 202...

    2024

  48. [56]

    Pattern recognition , volume=

    DANCo: An intrinsic dimensionality estimator exploiting angle and norm concentration , author=. Pattern recognition , volume=. 2014 , publisher=

    2014

  49. [57]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Low bias local intrinsic dimension estimation from expected simplex skewness , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2014 , publisher=

    2014

  50. [58]

    IEEE Transactions on Signal Processing , volume=

    On local intrinsic dimension estimation and its applications , author=. IEEE Transactions on Signal Processing , volume=. 2009 , publisher=

    2009

  51. [59]

    Data Mining and Knowledge Discovery , volume=

    Extreme-value-theoretic estimation of local intrinsic dimensionality , author=. Data Mining and Knowledge Discovery , volume=. 2018 , publisher=

    2018

  52. [60]

    Proceedings of the 2019 SIAM international conference on data mining , pages=

    Intrinsic dimensionality estimation within tight localities , author=. Proceedings of the 2019 SIAM international conference on data mining , pages=. 2019 , organization=

    2019

  53. [61]

    Scientific reports , volume=

    Estimating the intrinsic dimension of datasets by a minimal neighborhood information , author=. Scientific reports , volume=. 2017 , publisher=

    2017

  54. [62]

    Neurology , volume=

    Early A accumulation and progressive synaptic loss, gliosis, and tangle formation in AD brain , author=. Neurology , volume=. 2004 , publisher=

    2004

  55. [63]

    Nature Reviews Neurology , volume=

    Synaptic degeneration in Alzheimer disease , author=. Nature Reviews Neurology , volume=. 2023 , publisher=

    2023

  56. [64]

    Neurology , volume=

    Synaptic loss in Alzheimer's disease and other dementias , author=. Neurology , volume=. 1989 , publisher=

    1989

  57. [65]

    Acta neuropathologica , volume=

    Synaptic degeneration in Alzheimer’s disease , author=. Acta neuropathologica , volume=. 2009 , publisher=

    2009

  58. [66]

    Alzheimer's & Dementia , volume=

    Default mode network dedifferentiation predicts cognitive performance in Alzheimer disease: Neuroimaging/Optimal neuroimaging measures for tracking disease progression , author=. Alzheimer's & Dementia , volume=. 2020 , publisher=

    2020

  59. [67]

    Entropy , volume=

    Complexity analysis of EEG, MEG, and fMRI in mild cognitive impairment and Alzheimer’s disease: a review , author=. Entropy , volume=. 2020 , publisher=

    2020

  60. [68]

    Journal of Neurology , volume=

    Neural compensation in manifest neurodegeneration: systems neuroscience evidence from social cognition in frontotemporal dementia , author=. Journal of Neurology , volume=. 2023 , publisher=

    2023

  61. [69]

    Free Neuropathology , volume=

    Selective cellular and regional vulnerability in frontotemporal lobar degeneration: a scoping review , author=. Free Neuropathology , volume=

  62. [70]

    Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring , volume=

    Occipital atrophy signature in prodromal Lewy bodies disease , author=. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring , volume=. 2023 , publisher=

    2023

  63. [71]

    Neurology , volume=

    Occipital hypoperfusion on SPECT in dementia with Lewy bodies but not AD , author=. Neurology , volume=. 2001 , publisher=

    2001

  64. [72]

    Neurology , volume=

    Correlation of visual hallucinations with occipital rCBF changes by donepezil in DLB , author=. Neurology , volume=. 2006 , publisher=

    2006

  65. [73]

    Entropy , volume=

    Entropy and Complexity Tools Across Scales in Neuroscience: A Review , author=. Entropy , volume=. 2025 , publisher=

    2025

  66. [74]

    Neural Network World , volume=

    Analysis of EEG signals using Lyapunov exponents , author=. Neural Network World , volume=. 2006 , publisher=

    2006

  67. [75]

    Human brain mapping , volume=

    Changes in EEG multiscale entropy and power-law frequency scaling during the human sleep cycle , author=. Human brain mapping , volume=. 2019 , publisher=

    2019

  68. [76]

    Knowledge-based systems , volume=

    Application of entropies for automated diagnosis of epilepsy using EEG signals: A review , author=. Knowledge-based systems , volume=. 2015 , publisher=

    2015

  69. [77]

    Chaos, Solitons & Fractals , volume=

    A new algorithm for Largest Lyapunov Exponent determination for noisy chaotic signal studies with application to Electroencephalographic signals analysis for epilepsy and epileptic seizures detection , author=. Chaos, Solitons & Fractals , volume=. 2022 , publisher=

    2022

  70. [78]

    Nature Reviews Neuroscience , volume=

    A unifying perspective on neural manifolds and circuits for cognition , author=. Nature Reviews Neuroscience , volume=. 2023 , publisher=

    2023

  71. [79]

    Current Opinion in Behavioral Sciences , volume=

    The dimensionality of neural representations for control , author=. Current Opinion in Behavioral Sciences , volume=. 2021 , publisher=

    2021

  72. [80]

    NeuroImage , volume=

    Estimating the functional dimensionality of neural representations , author=. NeuroImage , volume=. 2018 , publisher=

    2018

  73. [81]

    Nature Reviews Neuroscience , volume=

    Neural tuning and representational geometry , author=. Nature Reviews Neuroscience , volume=. 2021 , publisher=

    2021

  74. [82]

    arXiv preprint arXiv:1804.08838 , year=

    Measuring the intrinsic dimension of objective landscapes , author=. arXiv preprint arXiv:1804.08838 , year=

    arXiv

  75. [83]

    Current opinion in neurobiology , volume=

    Representational drift: Emerging theories for continual learning and experimental future directions , author=. Current opinion in neurobiology , volume=. 2022 , publisher=

    2022

  76. [84]

    Journal of Cognitive Neuroscience , volume=

    Dynamics are the only constant in working memory , author=. Journal of Cognitive Neuroscience , volume=. 2022 , publisher=

    2022

  77. [85]

    Current Biology , volume=

    Neuronal circuits underlying persistent representations despite time varying activity , author=. Current Biology , volume=. 2012 , publisher=

    2012

  78. [86]

    Neuron , volume=

    Neural variability and sampling-based probabilistic representations in the visual cortex , author=. Neuron , volume=. 2016 , publisher=

    2016

  79. [87]

    Trends in cognitive sciences , volume=

    Neural variability: friend or foe? , author=. Trends in cognitive sciences , volume=. 2015 , publisher=

    2015

  80. [88]

    Current opinion in neurobiology , volume=

    Variability in neural activity and behavior , author=. Current opinion in neurobiology , volume=. 2014 , publisher=

    2014

  81. [89]

    Trends in cognitive sciences , volume=

    How variability shapes learning and generalization , author=. Trends in cognitive sciences , volume=. 2022 , publisher=

    2022

  82. [90]

    Brain imaging and behavior , volume=

    Understanding variability in the BOLD signal and why it matters for aging , author=. Brain imaging and behavior , volume=. 2014 , publisher=

    2014

  83. [91]

    Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring , volume=

    Cognitive variability—A marker for incident MCI and AD: An analysis for the Alzheimer's Disease Neuroimaging Initiative , author=. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring , volume=. 2016 , publisher=

    2016

  84. [92]

    Journal of Clinical and Experimental Neuropsychology , volume=

    Intraindividual variability as a marker of neurological dysfunction: a comparison of Alzheimer's disease and Parkinson's disease , author=. Journal of Clinical and Experimental Neuropsychology , volume=. 2006 , publisher=

    2006

  85. [93]

    Philosophical Transactions of the Royal Society B: Biological Sciences , volume=

    Variance and invariance of neuronal long-term representations , author=. Philosophical Transactions of the Royal Society B: Biological Sciences , volume=. 2017 , publisher=

    2017

  86. [94]

    International Journal of Psychophysiology , volume=

    Is cortical distribution of spectral power a stable individual characteristic? , author=. International Journal of Psychophysiology , volume=. 2009 , publisher=

    2009

  87. [95]

    Journal of biological physics , volume=

    The dynamics of EEG entropy , author=. Journal of biological physics , volume=. 2010 , publisher=

    2010

  88. [96]

    European Journal of Neuroscience , volume=

    Brain entropy, fractal dimensions and predictability: A review of complexity measures for EEG in healthy and neuropsychiatric populations , author=. European Journal of Neuroscience , volume=. 2022 , publisher=

    2022

  89. [97]

    Journal of Clinical Neurophysiology , volume=

    EEG, temporal correlations, and avalanches , author=. Journal of Clinical Neurophysiology , volume=. 2010 , publisher=

    2010

  90. [98]

    Clinical Neurophysiology , volume=

    Human EEG shows long-range temporal correlations of oscillation amplitude in Theta, Alpha and Beta bands across a wide age range , author=. Clinical Neurophysiology , volume=. 2010 , publisher=

    2010

  91. [99]

    Annual review of neuroscience , volume=

    Computation through neural population dynamics , author=. Annual review of neuroscience , volume=. 2020 , publisher=

    2020

  92. [100]

    Nature Communications , volume=

    Unsupervised approach to decomposing neural tuning variability , author=. Nature Communications , volume=. 2023 , publisher=

    2023

  93. [101]

    Nature Communications , volume=

    Temporal coding carries more stable cortical visual representations than firing rate over time , author=. Nature Communications , volume=. 2025 , publisher=

    2025

  94. [102]

    Alzheimer's & Dementia , volume=

    Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association Workgroup , author=. Alzheimer's & Dementia , volume=. 2024 , publisher=

    2024

  95. [103]

    GeroScience , volume=

    EEG entropy insights in the context of physiological aging and Alzheimer’s and Parkinson’s diseases: a comprehensive review , author=. GeroScience , volume=. 2024 , publisher=

    2024

  96. [104]

    Entropy , volume=

    Scikit-dimension: a python package for intrinsic dimension estimation , author=. Entropy , volume=. 2021 , publisher=

    2021

  97. [105]

    Journal of Machine Learning Research , volume=

    Pot: Python optimal transport , author=. Journal of Machine Learning Research , volume=

  98. [106]

    Bernoulli , volume=

    Sharp asymptotic and finite-sample rates of convergence of empirical measures in Wasserstein distance , author=. Bernoulli , volume=. 2019 , publisher=

    2019

  99. [107]

    Trends in cognitive sciences , volume=

    Representational geometry: integrating cognition, computation, and the brain , author=. Trends in cognitive sciences , volume=. 2013 , publisher=

    2013

  100. [108]

    Current opinion in neurobiology , volume=

    On simplicity and complexity in the brave new world of large-scale neuroscience , author=. Current opinion in neurobiology , volume=. 2015 , publisher=

    2015

  101. [109]

    Current opinion in neurobiology , volume=

    Towards the neural population doctrine , author=. Current opinion in neurobiology , volume=. 2019 , publisher=

    2019

  102. [110]

    Nature neuroscience , volume=

    Dimensionality reduction for large-scale neural recordings , author=. Nature neuroscience , volume=. 2014 , publisher=

    2014

  103. [111]

    PloS one , volume=

    Age-related changes in electroencephalographic signal complexity , author=. PloS one , volume=. 2015 , publisher=

    2015

  104. [112]

    Electroencephalography and clinical neurophysiology , volume=

    Discrimination of Alzheimer's disease and normal aging by EEG data , author=. Electroencephalography and clinical neurophysiology , volume=. 1997 , publisher=

    1997

  105. [113]

    PloS one , volume=

    Electroencephalographic fractal dimension in healthy ageing and Alzheimer’s disease , author=. PloS one , volume=. 2016 , publisher=

    2016

  106. [114]

    Fractals , volume=

    Age-based variations of fractal structure of EEG signal in patients with epilepsy , author=. Fractals , volume=. 2018 , publisher=

    2018

  107. [115]

    Dementia and geriatric cognitive disorders , volume=

    Global dimensional complexity of multichannel EEG in mild Alzheimer's disease and age-matched cohorts , author=. Dementia and geriatric cognitive disorders , volume=. 1997 , publisher=

    1997

  108. [116]

    IEEE Transactions on Biomedical Engineering , volume=

    M/EEG-based bio-markers to predict the MCI and Alzheimer's disease: a review from the ML perspective , author=. IEEE Transactions on Biomedical Engineering , volume=. 2019 , publisher=

    2019

  109. [117]

    Journal of Neuroscience , volume=

    The importance of being variable , author=. Journal of Neuroscience , volume=. 2011 , publisher=

    2011

  110. [118]

    The Neuroscientist , volume=

    The dynamical balance of the brain at rest , author=. The Neuroscientist , volume=. 2011 , publisher=

    2011

  111. [119]

    Neuron , volume=

    The metastable brain , author=. Neuron , volume=. 2014 , publisher=

    2014

  112. [120]

    Clinical neurophysiology , volume=

    EEG dynamics in patients with Alzheimer's disease , author=. Clinical neurophysiology , volume=. 2004 , publisher=

    2004

  113. [121]

    Clinical neurophysiology , volume=

    Nonlinear dynamical analysis of EEG and MEG: review of an emerging field , author=. Clinical neurophysiology , volume=. 2005 , publisher=

    2005

  114. [122]

    International Journal of Psychophysiology , volume=

    Brain neural synchronization and functional coupling in Alzheimer's disease as revealed by resting state EEG rhythms , author=. International Journal of Psychophysiology , volume=. 2016 , publisher=

    2016

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.