Pith. sign in

REVIEW 4 major objections 6 minor 67 references

Hidden-layer connection patterns in a trained network can flag failed neutron-star EoS inferences, reaching above 90% accuracy in the best case.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Applying Topological Uncertainty to hidden-layer activations of a trained FNN detects failed neutron-star EoS inferences with over 90% success in the best-tested configuration.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A plausible but not yet robust demonstration of TU for anomaly detection in EoS inference; the qualitative signal is real, the 90% number is not. the 4 major comments →

arxiv 2508.19683 v1 pith:WRN3R6JU submitted 2025-08-27 nucl-th cs.AIcs.LG

Topological Uncertainty for Anomaly Detection in the Neural-network EoS Inference with Neutron Star Data

Kenji Fukushima , Syo Kamata This is my paper
classification nucl-th cs.AIcs.LG
keywords Topological UncertaintyAnomaly DetectionFeedforward Neural NetworkEquation of StateNeutron StarPersistent HomologyMaximum Spanning TreeWasserstein Distance
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper argues that a trained feedforward network carries a readable trace of whether its own output is trustworthy, and that this trace can be extracted without seeing the correct answer. The testbed is the inverse problem of recovering the dense-matter equation of state from neutron-star mass-radius data, where the same network inference sometimes matches the true EoS and sometimes misses it badly. The authors label training samples by the actual error—k=0 for a close match, k=1 for a failure—and build a Topological Uncertainty for each class by converting each hidden layer's weighted connections into a maximum spanning tree, then comparing the sorted edge-weight distributions with a Wasserstein distance. Their central claim is that the class giving the smaller Topological Uncertainty is the correct one, so k=1 (anomalous) data are identified by having a smaller distance to the k=1 class average than to the k=0 one. On held-out data the rule achieves above 90% success in the best tested architecture, with performance depending on the network's width, depth, and tolerance parameter.

Core claim

On the paper's own terms, the discovery is that the cross-Topological-Uncertainty rule works: for held-out neutron-star inputs, an input is assigned to the class (normal or anomalous) whose Fréchet-mean MST weight distribution is closer in Wasserstein distance, and the assignment matches the true error label often enough to exceed 90% accuracy in the best architecture and tolerance range. The TU is a post-hoc quantity: it is computed from the already-trained weights by taking, for each layer, the edge weight |Wℓ(i,j)xℓ(i)|, forming the maximum spanning tree of the layer's bipartite graph, reading off the sorted edge weights as a probability distribution, and averaging the Wasserstein distanc

What carries the argument

The maximum-spanning-tree edge-weight distribution is the TU's core object. For each layer, the trained weights and the input activation give an edge weight w(e)=|Wℓ(i,j)xℓ(i)|; edges are added in decreasing weight order to form the maximum spanning tree, and the sorted weights define a probability distribution μ(G). The Topological Uncertainty TU_k(X) is the layer-averaged p=1 Wasserstein distance between this distribution and the Fréchet mean distribution for class k, built from the labeled training subsets. The machinery turns the entire trained network into a per-class distance, so the lowest TU value acts as a predicted class label.

Load-bearing premise

The load-bearing premise is that the hidden-layer activation patterns of the trained network are reliably different for successful versus failed inferences, and that the difference seen in the paper's training runs carries over to new inputs; the paper tests this only on a small number of runs per architecture, without repeated-seed error bars or distribution-shift experiments.

What would settle it

Run the same experiment many times: train F(4,4,2,2) from, say, 20 different random initializations and evaluate R_tot at δ=0.5 on an independent Dnew for each seed. If the reported >90% accuracy reproduces in every seed, the claim is supported; if any seed gives chance-level R_k=1, or if R_k=1 collapses when Dnew is sampled from a slightly shifted mass-radius distribution, the claimed generalization fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Because TU is computed from the already-trained network, any existing FNNEoS can be turned into an anomaly detector post hoc: no retraining or architectural change is required.
  • The correct-inference rate R_k=0 and failure-detection rate R_k=1 behave differently with the tolerance δ; the largest tested network gives the best overall score, so architecture choice matters for anomaly detection.
  • Hidden-layer topology supplies signal beyond the final output: the paper's histograms show TU separates classes even when output-based confidence would not.
  • The success criterion is concrete: the anomaly is detected when the cross-TU for (j,k)=(1,1) is smaller than for (j,k)=(0,1), and this rule can be applied to any new input without its true EoS.
  • The method generalizes, in principle, to any FNN inference problem where a label function k can be defined by comparing the prediction with a known answer during calibration.

Where Pith is reading between the lines

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

  • Editorial inference: the same construction should transfer to other regression FNNs—gravitational-wave parameter estimation or heavy-ion surrogate models—provided a calibration set with known errors can be labeled, and the paper's own conclusions gesture at this.
  • Editorial inference: since the paper trains each architecture only once, the above-90% figure is a point estimate; repeated-seed and distribution-shift tests would tell whether the separation of k=0/k=1 Fréchet means is stable.
  • Editorial inference: the tolerance parameter δ is doing double duty—it defines the labels that train the TU classes and sets the yardstick for success—so a fully unsupervised deployment would need a principled way to choose δ without peeking at answers.
  • Editorial inference: replacing the argmin decision rule with a calibrated threshold on TU_1 − TU_0 could yield anomaly scores with controllable false-alarm rates, which the paper does not explore.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes using Topological Uncertainty (TU), a persistent-homology-based summary of a trained feedforward network's edge-weight distributions, as a post-hoc anomaly detector for neural-network inference of the neutron-star equation of state. The authors label training inputs as successful (k=0) or unsuccessful (k=1) according to whether the L_infinity logarithmic-pressure deviation between predicted and true EoS is below a tolerance delta (Eqs. (19)-(20)), construct per-class Frechet means of sorted MST edge weights (Eq. (10)), and define cross-TU distances (Eqs. (11),(23)). An input is flagged anomalous if it is closer to the k=1 Frechet mean. On a held-out mock dataset balanced at 500 per class, they report detection rates exceeding 90% for the best architecture F(4,4,2,2) and favorable delta, and they study dependence on delta and architecture.

Significance. If the result holds, the paper offers a cheap, architecture-agnostic way to flag unreliable neural predictions, which is of genuine value for EoS inference and for physics surrogate models more broadly. The exposition of the TU construction is clear, the numerical setup is coherent, and the use of Wasserstein distances between activation-graph persistence distributions is a sensible extension of the TU idea. However, the quantitative evidence for the headline claim is currently thin: there are no repeated training runs or error bars, the best delta and architecture are selected on the same test set, and no comparison is made with output-based uncertainty baselines. The paper's claimed robustness is not supported by its own results, and the novelty claim needs to be calibrated against Ref. [50]. These issues are load-bearing for the central 'exceeds 90%' statement, though the underlying method appears plausible.

major comments (4)
  1. [Sec. III C, Figs. 6 and 7] The headline claim that the anomaly-detection success rate exceeds 90% rests on a single training run per architecture and on selecting delta and F(.) after looking at the same test set. No repeated seeds, confidence intervals, or validation-based model-selection protocol are provided. The strong sensitivity of the result is visible in the paper itself: F(2,2,2,2) and F(4,4,2,2) have nearly equal training/validation losses (Table II), but their R_{k=1}(delta) curves in Fig. 6 are dramatically different, and the R_{k=1} histogram for F(2,2,2,2) in Fig. 7 shows almost no separation. The reported 'best case' 90% could therefore be a seed or delta-selection artifact. Please provide repeated training runs and either a pre-registered validation selection or per-seed error bars, or temper the quantitative claim.
  2. [Sec. III B, Eqs. (22)-(25)] R_tot is computed on a test set with 500 examples per class, which is not the natural base rate. For delta > 0.2, Fig. 4 shows RLQ(delta) >= 0.8, meaning the k=1 class is only 10-20% of a naturally sampled dataset. Weighting the same per-class detection rates by this base rate can lower the overall success rate considerably, so the balanced R_tot overstates what a user would experience in practice. Please report R_tot under the empirical class prior, or report ROC/AUC and per-class rates separately, and discuss how delta is chosen in an unsupervised setting.
  3. [Sec. I, Sec. IV, Ref. [50]] The statement that this is the 'first successful application of the TU to Anomaly Detection' is overstated. Ref. [50] introduced Topological Uncertainty precisely for monitoring trained neural networks through persistence of activation graphs, and that framework already includes detecting inputs for which the network is unreliable. The present paper even cites Ref. [50] as the original TU proposal. The novelty should be framed as a first physics application or a first demonstration in EoS inference, not as the first anomaly-detection use of TU.
  4. [Sec. IV vs Sec. III C] The conclusion says the method is 'robust against moderate variations of hyperparameters', but this is contradicted by Fig. 6. For example, F(2,2,2,2), with train/val losses close to those of the best F(4,4,2,2), has R_{k=1} below F(2,2) at many delta values. The paper gives no mechanism for this non-monotonic architecture sensitivity. Either provide an explanation or soften the robustness claim to 'performance varies with architecture and delta, and the best configuration can exceed 90%'.
minor comments (6)
  1. [Sec. II B] Typo: 'The number of the edges is |V| = d_l + d_{l+1}' should be 'number of vertices'. The edge count is |E| = d_l * d_{l+1}, as correctly stated immediately after.
  2. [Sec. II B, Eq. (9)] The Wasserstein-1 formula Dist(mu,nu) = (1/N) sum |w_i - w_i'| is only valid after sorting the edge weights within each MST. This is implicitly done later in Sec. III C ('in our convention of the ordering'), but Eq. (9) should state the sorting explicitly to avoid confusion.
  3. [Sec. III B] The notation 'Ntr = 3' is inconsistent with the previously defined truncation parameter Ntrun in Eq. (19). Please use Ntrun consistently.
  4. [Sec. III C] Minor typos: 'F(2,2,)' should be 'F(2,2)', and 'RLR should increase' should be 'RLQ'.
  5. [Fig. 5] The normalization of the plotted distribution (division by N_l = |V_l| - 1) is mentioned only in the text; the figure caption should state it explicitly for readability.
  6. [Sec. IV] Typo: 'classifiction' should be 'classification'.

Circularity Check

0 steps flagged

No circularity found: the TU anomaly-detection claim is evaluated on held-out data against externally defined labels.

full rationale

The paper's central claim—that TU values computed from hidden-layer MST edge weights separate successful (k=0) from unsuccessful (k=1) EoS inferences—is not definitionally forced. Labels k are assigned by Eq. (20) from the L-infinity log-pressure deviation between the predicted and true EoSs; the class Frechet means mu_k in Eq. (10) are built only from the training sub-datasets Dtrain_k, and the performance R_k and R_tot in Eq. (25) is measured on an independent test set Dnew with ground-truth labels. No fitted parameter is renamed as a prediction: the only fitted objects are the FNN weights and the class means, and the test inputs are held out. The cited prior work of the authors [25-28] supplies the FNN-EoS data-generation scheme but not the TU criterion; the TU construction itself is credited to external work [50,54]. The claimed >90% success rate is therefore an empirical generalization result, not a consequence of the definitions. The main caveats—single training runs, architecture/delta selection on the same test set, and absence of error bars—concern statistical reliability rather than circular reasoning, and the paper itself acknowledges that 'further work is needed to establish quantitative reliability guarantees.'

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

No new physical entities are postulated. The method introduces the cross-TU quantity, but that is a methodological construct, not an entity needing independent evidence. The central claim rests on three hand-chosen parameters (δ, Ntrun, architecture) and on the empirical separability of the class-conditional edge-weight distributions.

free parameters (3)
  • tolerance parameter δ = scanned over 0.1 to 0.8; best results around 0.3-0.5
    Defines the label k=Kδ in Eq. (20); the reported >90% success rate is conditional on δ.
  • truncation index Ntrun = 3
    Truncates the log-pressure deviation in Eq. (19); anomalies at n=4,5 are explicitly ignored.
  • network architecture F(h2,...,hL) = nine architectures tested; best F(4,4,2,2)
    Performance depends strongly on architecture; the 'best case' claim is a scan over architectures.
axioms (4)
  • domain assumption The trained FNN activations carry information about inference failure that is extractable from the activation-weighted weight distribution |W x|.
    The whole TU construction in Sec. II B assumes this; it is plausible but not proven.
  • standard math The persistence diagram of H0 from the MST filtration is a stable summary (stability from Ref [55]).
    Cited from Cohen-Steiner et al.; the paper relies on stability to justify TU.
  • standard math The one-to-one correspondence between the EoS and the M-R relation (Lindblom).
    Invoked in Sec. III A via Ref [46] to justify the inverse problem.
  • domain assumption The labeled split Dtrain_k is representative of held-out Dnew_k (no distribution shift between training and test data).
    Both are generated from the same mock distribution; real data may differ.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Topological Uncertainty for Anomaly Detection in the Neural-network EoS Inference with Neutron Star Data." pith.science (2026). https://pith.science/paper/WRN3R6JU

@misc{pith2026250819683,
  author       = {Pith},
  title        = {Pith review of: Topological Uncertainty for Anomaly Detection in the Neural-network EoS Inference with Neutron Star Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WRN3R6JU}},
  note         = {Machine review of arXiv:2508.19683}
}
Share X Bluesky LinkedIn Reddit HN
abstract

We study the performance of the Topological Uncertainty (TU) constructed with a trained feedforward neural network (FNN) for Anomaly Detection. Generally, meaningful information can be stored in the hidden layers of the trained FNN, and the TU implementation is one tractable recipe to extract buried information by means of the Topological Data Analysis. We explicate the concept of the TU and the numerical procedures. Then, for a concrete demonstration of the performance test, we employ the Neutron Star data used for inference of the equation of state (EoS). For the training dataset consisting of the input (Neutron Star data) and the output (EoS parameters), we can compare the inferred EoSs and the exact answers to classify the data with the label $k$. The subdataset with $k=0$ leads to the normal inference for which the inferred EoS approximates the answer well, while the subdataset with $k=1$ ends up with the unsuccessful inference. Once the TU is prepared based on the $k$-labled subdatasets, we introduce the cross-TU to quantify the uncertainty of characterizing the $k$-labeled data with the label $j$. The anomaly or unsuccessful inference is correctly detected if the cross-TU for $j=k=1$ is smaller than that for $j=0$ and $k=1$. In our numerical experiment, for various input data, we calculate the cross-TU and estimate the performance of Anomaly Detection. We find that performance depends on FNN hyperparameters, and the success rate of Anomaly Detection exceeds $90\%$ in the best case. We finally discuss further potential of the TU application to retrieve the information hidden in the trained FNN.

Figures

Figures reproduced from arXiv: 2508.19683 by Kenji Fukushima, Syo Kamata.

Figure 1
Figure 1. Figure 1: FIG. 1. Examples of the successful (left) and the unsuccessful (right) results of the EoS inference. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Schematic figure to illustrate the way how to construct the TU from the FNN framework. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Logarithmic separation between the predicted EoSs and the exact answers quantified by [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: shows the distribution, ¯wk,i, normalized by Nℓ in the Fr´echet mean, ¯µk,ℓ in Eq. (10) for k = 0 (solid lines) and k = 1 (dashed lines). The normalization is Nℓ = |Vℓ | − 1 for fℓ∈{1,...,L}. The panels display the results for F(2,2) (left-top), F(4,4) (right-top), F(2,2,2,2) (left-bottom), and F(4,4,2,2) (right-bottom) with the label k for the tolerance δ = 0.5. It should be noted that the TU measures the… view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Histogram of TU [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

67 extracted references · 27 canonical work pages · 18 internal anchors

  1. [1]

    So far, in the literature, there are not many discussions for analyzing the bulk properties of the trained FNNs

    The trained FNNs contain more information in the hidden layers than that accessed in the output layer only. So far, in the literature, there are not many discussions for analyzing the bulk properties of the trained FNNs

    work page

  2. [2]

    The bulk properties could characterize the learning quality to distinguish the successful and unsuccessful applications of the FNNs

    Even if the output from the FNNs is identical, the bulk response including the hidden layers may be different. The bulk properties could characterize the learning quality to distinguish the successful and unsuccessful applications of the FNNs. In the present work, we often refer to the second question as Anomaly Detection. To give intuitive explanations, ...

    work page

  3. [3]

    G. Baym, T. Hatsuda, T. Kojo, P. D. Powell, Y. Song, and T. Takatsuka, Rept. Prog. Phys. 81, 056902 (2018), arXiv:1707.04966 [astro-ph.HE]

    Pith/arXiv arXiv 2018

  4. [4]

    Andronic, P

    A. Andronic, P. Braun-Munzinger, K. Redlich, and J. Stachel, Nature 561, 321 (2018), arXiv:1710.09425 [nucl-th]

    Pith/arXiv arXiv 2018

  5. [5]

    C. S. Fischer, Prog. Part. Nucl. Phys. 105, 1 (2019), arXiv:1810.12938 [hep-ph]

    Pith/arXiv arXiv 2019

  6. [6]

    Aarts et al., Prog

    G. Aarts et al., Prog. Part. Nucl. Phys. 133, 104070 (2023), arXiv:2301.04382 [hep-lat]

    Pith/arXiv arXiv 2023

  7. [7]

    Fukushima, J

    K. Fukushima, J. Subatomic Part. Cosmol. 3, 100066 (2025), arXiv:2501.01907 [hep-ph]

    Pith/arXiv arXiv 2025

  8. [8]

    Bollweg, D

    D. Bollweg, D. A. Clarke, J. Goswami, O. Kaczmarek, F. Karsch, S. Mukherjee, P. Petreczky, C. Schmidt, and S. Sharma (HotQCD), Phys. Rev. D 108, 014510 (2023), arXiv:2212.09043 [hep-lat]

    Pith/arXiv arXiv 2023

  9. [9]

    Continuum extrapolated high order baryon fluctuations

    S. Borsanyi, Z. Fodor, J. N. Guenther, S. D. Katz, P. Parotto, A. Pasztor, D. Pesznyak, K. K. Szabo, and C. H. Wong, Phys. Rev. D 110, L011501 (2024), arXiv:2312.07528 [hep-lat]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  10. [10]

    D. T. Son and M. A. Stephanov, Phys. Rev. Lett. 86, 592 (2001), arXiv:hep-ph/0005225

    Pith/arXiv arXiv 2001

  11. [11]

    B. B. Brandt, F. Cuteri, and G. Endrodi, JHEP 07, 055 (2023), arXiv:2212.14016 [hep-lat]

    Pith/arXiv arXiv 2023

  12. [12]

    Abbott, W

    R. Abbott, W. Detmold, F. Romero-L´ opez, Z. Davoudi, M. Illa, A. Parre˜ no, R. J. Perry, P. E. Shanahan, and M. L. Wagman (NPLQCD), Phys. Rev. D 108, 114506 (2023), arXiv:2307.15014 [hep-lat]

    Pith/arXiv arXiv 2023

  13. [13]

    V. V. Braguta, E. M. Ilgenfritz, A. Y. Kotov, A. V. Molochkov, and A. A. Nikolaev, Phys. Rev. D 94, 114510 (2016), arXiv:1605.04090 [hep-lat]

    Pith/arXiv arXiv 2016

  14. [14]

    Iida and E

    K. Iida and E. Itou, PTEP 2022, 111B01 (2022), arXiv:2207.01253 [hep-ph]

    Pith/arXiv arXiv 2022

  15. [15]

    Buballa and S

    M. Buballa and S. Carignano, Prog. Part. Nucl. Phys. 81, 39 (2015), arXiv:1406.1367 [hep-ph]

    Pith/arXiv arXiv 2015

  16. [16]

    T.-G. Lee, E. Nakano, Y. Tsue, T. Tatsumi, and B. Friman, Phys. Rev. D 92, 034024 (2015), arXiv:1504.03185 [hep-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  17. [17]

    Phonons, pions and quasi-long-range order in spatially modulated chiral condensates

    Y. Hidaka, K. Kamikado, T. Kanazawa, and T. Noumi, Phys. Rev. D 92, 034003 (2015), arXiv:1505.00848 [hep-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  18. [18]

    R. D. Pisarski and F. Rennecke, Phys. Rev. Lett. 127, 152302 (2021), arXiv:2103.06890 [hep- ph]

    Pith/arXiv arXiv 2021

  19. [19]

    McLerran and R

    L. McLerran and R. D. Pisarski, Nucl. Phys. A 796, 83 (2007), arXiv:0706.2191 [hep-ph]. 20

    Pith/arXiv arXiv 2007

  20. [20]

    M. G. Alford, A. Schmitt, K. Rajagopal, and T. Sch¨ afer, Rev. Mod. Phys. 80, 1455 (2008), arXiv:0709.4635 [hep-ph]

    Pith/arXiv arXiv 2008

  21. [21]

    T. E. Riley et al., Astrophys. J. Lett. 887, L21 (2019), arXiv:1912.05702 [astro-ph.HE]; As- trophys. J. Lett. 918, L27 (2021), arXiv:2105.06980 [astro-ph.HE]

    Pith/arXiv arXiv 2019

  22. [22]

    M. C. Miller et al., Astrophys. J. Lett. 887, L24 (2019), arXiv:1912.05705 [astro-ph.HE]; Astrophys. J. Lett. 918, L28 (2021), arXiv:2105.06979 [astro-ph.HE]

    Pith/arXiv arXiv 2019

  23. [23]

    B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. Lett. 119, 161101 (2017), arXiv:1710.05832 [gr-qc]; Phys. Rev. Lett. 121, 161101 (2018), arXiv:1805.11581 [gr-qc]

    Pith/arXiv arXiv 2017

  24. [24]

    A. W. Steiner, J. M. Lattimer, and E. F. Brown, Astrophys. J. 722, 33 (2010), arXiv:1005.0811 [astro-ph.HE]

    Pith/arXiv arXiv 2010

  25. [25]

    F. Ozel, D. Psaltis, T. Guver, G. Baym, C. Heinke, and S. Guillot, Astrophys. J. 820, 28 (2016), arXiv:1505.05155 [astro-ph.HE]

    Pith/arXiv arXiv 2016

  26. [26]

    Neutron Star Mass-Radius Constraints of the Quiescent Low-mass X-ray Binaries X7 and X5 in the Globular Cluster 47 Tuc

    S. Bogdanov, C. O. Heinke, F. ¨Ozel, and T. G¨ uver, Astrophys. J. 831, 184 (2016), arXiv:1603.01630 [astro-ph.HE]

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  27. [27]

    Methodology study of machine learning for the neutron star equation of state

    Y. Fujimoto, K. Fukushima, and K. Murase, Phys. Rev. D 98, 023019 (2018), arXiv:1711.06748 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  28. [28]

    Fujimoto, K

    Y. Fujimoto, K. Fukushima, and K. Murase, Phys. Rev. D 101, 054016 (2020), arXiv:1903.03400 [nucl-th]

    Pith/arXiv arXiv 2020

  29. [29]

    Extensive Studies of the Neutron Star Equation of State from the Deep Learning Inference with the Observational Data Augmentation

    Y. Fujimoto, K. Fukushima, and K. Murase, JHEP 03, 273 (2021), arXiv:2101.08156 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  30. [30]

    Fujimoto, K

    Y. Fujimoto, K. Fukushima, S. Kamata, and K. Murase, Phys. Rev. D 110, 034035 (2024), arXiv:2401.12688 [nucl-th]

    Pith/arXiv arXiv 2024

  31. [31]

    Ferreira and C

    M. Ferreira and C. Providˆ encia, JCAP07, 011 (2021), arXiv:1910.05554 [nucl-th]

    Pith/arXiv arXiv 2021

  32. [32]

    Morawski and M

    F. Morawski and M. Bejger, Astron. Astrophys. 642, A78 (2020), arXiv:2006.07194 [astro- ph.HE]

    Pith/arXiv arXiv 2020

  33. [33]

    Structure of Quark Star: A Comparative Analysis of Bayesian Inference and Neural Network Based Modeling

    S. Traversi and P. Char, Astrophys. J. 905, 9 (2020), arXiv:2007.10239 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  34. [34]

    P. G. Krastev, Galaxies 10, 16 (2022), arXiv:2112.04089 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  35. [35]

    S. Soma, L. Wang, S. Shi, H. St¨ ocker, and K. Zhou, JCAP 08, 071 (2022), arXiv:2201.01756 [hep-ph]

    Pith/arXiv arXiv 2022

  36. [36]

    Thete, K

    A. Thete, K. Banerjee, and T. Malik, Phys. Rev. D 108, 063028 (2023), arXiv:2208.13163 [nucl-th]

    Pith/arXiv arXiv 2023

  37. [37]

    Farrell, P

    D. Farrell, P. Baldi, J. Ott, A. Ghosh, A. W. Steiner, A. Kavitkar, L. Lindblom, D. Whiteson, 21 and F. Weber, JCAP 02, 016 (2023), arXiv:2209.02817 [astro-ph.HE]

    Pith/arXiv arXiv 2023

  38. [38]

    S. Soma, L. Wang, S. Shi, H. St¨ ocker, and K. Zhou, Phys. Rev. D 107, 083028 (2023), arXiv:2209.08883 [astro-ph.HE]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  39. [39]

    Extracting nuclear matter properties from the neutron star matter equation of state using deep neural networks

    M. Ferreira, V. Carvalho, and C. Providˆ encia, Phys. Rev. D 106, 103023 (2022), arXiv:2209.09085 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  40. [40]

    Machine-Learning Love: classifying the equation of state of neutron stars with Transformers

    G. Gon¸ calves, M. Ferreira, J. a. Aveiro, A. Onofre, F. F. Freitas, C. Providˆ encia, and J. A. Font, JCAP 12, 001 (2023), arXiv:2210.08382 [astro-ph.IM]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  41. [41]

    Chatterjee, H

    S. Chatterjee, H. Sudhakaran, and R. Mallick, Eur. Phys. J. C 84, 1291 (2024), arXiv:2302.13648 [astro-ph.HE]

    Pith/arXiv arXiv 2024

  42. [42]

    P. G. Krastev, Symmetry 15, 1123 (2023), arXiv:2303.17146 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  43. [43]

    Deducing Neutron Star Equation of State from Telescope Spectra with Machine-learning-derived Likelihoods

    D. Farrell, P. Baldi, J. Ott, A. Ghosh, A. W. Steiner, A. Kavitkar, L. Lindblom, D. Whiteson, and F. Weber, JCAP 12, 022 (2023), arXiv:2305.07442 [astro-ph.HE]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  44. [44]

    W. Zhou, J. Hu, Y. Zhang, and H. Shen, Astrophys. J. 950, 186 (2023), arXiv:2305.03323 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  45. [45]

    Decoding Neutron Star Observations: Revealing Composition through Bayesian Neural Networks

    V. Carvalho, M. Ferreira, T. Malik, and C. Providˆ encia, Phys. Rev. D 108, 043031 (2023), arXiv:2306.06929 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  46. [46]

    From NS observations to nuclear matter properties: a machine learning approach

    V. Carvalho, M. Ferreira, and C. Providˆ encia, Phys. Rev. D 109, 123038 (2024), arXiv:2401.05770 [nucl-th]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  47. [47]

    N. K. Patra, T. Malik, H. Pais, K. Zhou, B. K. Agrawal, and C. Providˆ encia, Phys. Lett. B 865, 139470 (2025), arXiv:2502.20226 [nucl-th]

    Pith/arXiv arXiv 2025

  48. [48]

    Lindblom, Astrophys

    L. Lindblom, Astrophys. J. 398, 569 (1992)

    work page 1992

  49. [49]

    H. H. J. Salima Omar, Asri Ngadi, International Journal of Computer Applications 79, 33 (2013)

    work page 2013

  50. [50]

    A. B. Nassif, M. A. Talib, Q. Nasir, and F. M. Dakalbab, IEEE Access 9, 78658 (2021)

    work page 2021

  51. [51]

    G. Pang, C. Shen, L. Cao, and A. V. D. Hengel, ACM Comput. Surv. 54 (2021), 10.1145/3439950

    doi:10.1145/3439950 2021

  52. [52]

    Topological uncer- tainty: Monitoring trained neural networks through persistence of activation graphs,

    T. Lacombe, Y. Ike, M. Carriere, F. Chazal, M. Glisse, and Y. Umeda, “Topological uncer- tainty: Monitoring trained neural networks through persistence of activation graphs,” (2021), arXiv:2105.04404 [stat.ML]

    Pith/arXiv arXiv 2021

  53. [53]

    Carlsson, Nature Reviews Physics 2, 697 (2020)

    G. Carlsson, Nature Reviews Physics 2, 697 (2020)

    work page 2020

  54. [54]

    An introduction to topological data analysis: fundamental and 22 practical aspects for data scientists,

    F. Chazal and B. Michel, “An introduction to topological data analysis: fundamental and 22 practical aspects for data scientists,” (2021), arXiv:1710.04019 [math.ST]

    Pith/arXiv arXiv 2021

  55. [55]

    Topological data analysis and machine learning

    D. Leykam and D. G. Angelakis, Adv. Phys. X 8, 2202331 (2023), arXiv:2206.15075 [cond- mat.mes-hall]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  56. [56]

    Characterizing the Shape of Activation Space in Deep Neural Networks

    T. Gebhart, P. Schrater, and A. Hylton, “Characterizing the shape of activation space in deep neural networks,” (2019), arXiv:1901.09496 [cs.LG]

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  57. [57]

    Cohen-Steiner, H

    D. Cohen-Steiner, H. Edelsbrunner, and J. Harer, Discrete & Computational Geometry 37, 103 (2007)

    work page 2007

  58. [58]

    Hendrycks and K

    D. Hendrycks and K. Gimpel, in International Conference on Learning Representations (ICLR) (2017) arXiv:1610.02136

    Pith/arXiv arXiv 2017

  59. [59]

    Liang, Y

    S. Liang, Y. Li, and R. Srikant, in International Conference on Learning Representations (ICLR) (2018) arXiv:1706.02690

    Pith/arXiv arXiv 2018

  60. [60]

    W. Liu, X. Wang, J. D. Owens, and Y. Li, in Advances in Neural Information Processing Systems (NeurIPS) (2020) arXiv:2010.03759

    Pith/arXiv arXiv 2020

  61. [61]

    Rieck, M

    B. Rieck, M. Togninalli, C. Bock, M. Moor, M. Horn, T. Gumbsch, and K. Borgwardt, in International Conference on Learning Representations (ICLR)(2019) arXiv:1812.09764

    Pith/arXiv arXiv 2019

  62. [62]

    Br¨ uel Gabrielsson, B

    R. Br¨ uel Gabrielsson, B. J. Nelson, A. Dwaraknath, and P. Skraba, in AISTATS 2020, Proceedings of Machine Learning Research, Vol. 108 (PMLR, 2020) pp. 1553–1563

    work page 2020

  63. [63]

    M. Moor, M. Horn, B. Rieck, and K. Borgwardt, in International Conference on Machine Learning (ICML), Proceedings of Machine Learning Research, Vol. 119 (PMLR, 2020) pp. 7045–7054

    work page 2020

  64. [64]

    J. S. Read, B. D. Lackey, B. J. Owen, and J. L. Friedman, Phys. Rev. D 79, 124032 (2009), arXiv:0812.2163 [astro-ph]

    Pith/arXiv arXiv 2009

  65. [65]

    R. C. Tolman, Phys. Rev. 55, 364 (1939)

    1939

  66. [66]

    J. R. Oppenheimer and G. M. Volkoff, Phys. Rev. 55, 374 (1939)

    1939

  67. [67]

    Brandes, C

    L. Brandes, C. Modi, A. Ghosh, D. Farrell, L. Lindblom, L. Heinrich, A. W. Steiner, F. Weber, and D. Whiteson, JCAP 09, 009 (2024), arXiv:2403.00287 [astro-ph.HE]. 23

    Pith/arXiv arXiv 2024

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.