---
id: "20260721-0200-do-persistent-homology-based"
title: "Do persistent-homology-based neural-network generalization diagnostics share the Byzantine trade-network study's hub-selection / sampling-artifact vulnerability?"
type: "capture"
status: "promoted"
origin: "batch"
promoted_to: ["30-notes/claim-birdal-2021-phd-estimator-samples-training-iterates-uniformly-at-random.md","30-notes/claim-gutierrez-fandino-2021-method-subsamples-nothing-uses-full-network.md","30-notes/claim-chazal-2014-persistence-diagram-subsampling-stable-under-noise-not-selection.md","30-notes/claim-stolz-2023-landmark-selection-rules-trade-density-bias-for-noise-sensitivity.md","30-notes/observation-hub-selection-artifact-absent-by-design-in-founding-ph-generalization-papers.md","40-entities/entity-tolga-birdal.md","40-entities/entity-asier-gutierrez-fandino.md","40-entities/entity-persistent-homology-dimension.md","40-entities/entity-landmark-selection-topological-data-analysis.md"]
not_promoted: ["Rieck et al. 2019 'Neural Persistence' lead (Further leads) — not read this session, no source fetched; too thin for even a watching entity stub (a name inside another paper's related-work section, not independently examined). Left as a further lead in the capstone note and this journal, not a claim or a routed question.","Corneanu et al. 2020 'Computing the Testing Error Without a Testing Set' lead — same reason, not read this session.","Chowdhury et al. 2019 'Path homologies of deep feedforward networks' lead — cited by Gutiérrez-Fandiño et al. as related work; sampling method not checked this session, not read directly.","The capture's broader [unverified] disclaimer ('could not confirm or deny for the field as a whole') — preserved verbatim in the capstone observation note's frontmatter and body rather than promoted to a claim or routed as a new question: no kept note asserts field-wide immunity, so there is no load-bearing doubt a question would be tracking.","Entity candidate Umut Şimşekli — declined promotion to a hub. Real, co-author of the Birdal paper, but the capture's own justification ('connects to heavy-tailed-SGD generalization literature already touched by the vault') does not check out — grep against 30-notes/ found no existing vault claim on heavy-tailed SGD. Also omitted from entity-tolga-birdal.md's connects_to for the same reason; a co-author credit alone isn't yet a hub-worthy 'why it matters' in his own right.","Entity candidates Frédéric Chazal and Bernadette J. Stolz — declined hub promotion. Both real, both first authors of well-sourced Tier 1 papers now grounding new claim-notes, but each appears in exactly one vault note so far — below the recurrence bar the vault has applied to comparable single-paper authors (e.g. Widrow needed six-plus notes before a hub). Left as mentions; will promote if either recurs."]
writer_model: "claude-sonnet-5"
date_created: "2026-07-21T00:00:00.000Z"
provenance: "web-research batch run, 2026-07-21"
derived_from: []
tags: ["persistent-homology","topological-data-analysis","sampling-artifact","generalization","methodology","cross-domain-bridge","neural-networks","subsampling"]
sources: [{"source_url":"https://arxiv.org/abs/2111.13171","source_author":"Tolga Birdal, Aaron Lou, Leonidas Guibas, Umut Şimşekli","source_date":"2021-11","source_tier":1,"source_note":"Full PDF fetched and read directly via extract_pdf (arxiv.org/pdf/2111.13171, tls:verified). 'Intrinsic Dimension, Persistent Homology and Generalization in Neural Networks,' NeurIPS 2021."},{"source_url":"https://arxiv.org/abs/2106.00012","source_author":"Asier Gutiérrez-Fandiño, David Pérez-Fernández, Jordi Armengol-Estapé, Marta Villegas","source_date":"2021-05","source_tier":1,"source_note":"Full PDF fetched and read directly via extract_pdf (arxiv.org/pdf/2106.00012, tls:verified). 'Persistent Homology Captures the Generalization of Neural Networks Without A Validation Set.'"},{"source_url":"https://arxiv.org/abs/1406.1901","source_author":"Frédéric Chazal, Brittany Terese Fasy, Fabrizio Lecci, Bertrand Michel, Alessandro Rinaldo, Larry Wasserman","source_date":"2014-06","source_tier":1,"source_note":"Full PDF fetched and read directly via extract_pdf (arxiv.org/pdf/1406.1901, tls:verified). 'Subsampling Methods for Persistent Homology,' ICML 2015."},{"source_url":"https://www.jmlr.org/papers/v24/21-1526.html","source_author":"Bernadette J. Stolz","source_date":"2023-02","source_tier":1,"source_note":"Full PDF fetched and read directly via extract_pdf (jmlr.org/papers/volume24/21-1526/21-1526.pdf, tls:verified). 'Outlier-Robust Subsampling Techniques for Persistent Homology,' JMLR 24 (2023)."},{"source_url":"https://arxiv.org/abs/2607.05695","source_author":"Jose de Jesus Bernal-Alvarado, David Delepine, Carlos Pinedo Guadarrama","source_date":"2026-07","source_tier":1,"source_note":"Not re-fetched this session; cited from existing vault notes [[claim-hub-selection-artifact-can-reverse-network-breakpoint-signal]] and [[claim-roman-byzantine-trade-network-decoupled-after-1082-chrysobull]], both already verified-verbatim in prior sessions."}]
---


This capture directly researches the open question logged at
[[question-tda-neural-net-sampling-artifact-risk]]: whether the two founding
neural-network persistent-homology (PH) generalization diagnostics —
[[claim-birdal-2021-persistent-homology-dimension-bounds-generalization]] and
[[claim-gutierrez-fandino-2021-persistence-diagram-distance-tracks-generalization]]
— inherit the specific failure mode the Byzantine trade-network study caught in
itself: a **hub-selection artifact**, where sampling a network by a structural
covariate (node degree) rather than at random can reverse the sign of an
inferred signal
([[claim-hub-selection-artifact-can-reverse-network-breakpoint-signal]]).
Method was to read both NN-PH papers' own method/algorithm sections directly
(full PDF, not abstract only), plus two Tier-1 papers from the general TDA
subsampling-stability literature.

## Claim: Birdal et al.'s persistent-homology-dimension estimator subsamples training iterates uniformly at random, and documents a known small-sample estimation bias rather than a sign-reversing artifact

The PHD estimator in Birdal, Lou, Guibas & Şimşekli (2021) is computed by
Algorithm 1, which repeatedly draws a subset of size *n* from the full set of
recorded training iterates and fits a power law across increasing *n*; the
algorithm's own comment labels this step explicitly: "`Wn ← sample(W, n) //
random sampling`." The paper's experimental section confirms the same
procedure in prose: "For varying n, we randomly pick n of last iterates and
compute Eα0." This is uniform random sampling over the trajectory, not
selection by any structural property (e.g., weight magnitude, layer, or
training-loss rank) analogous to the Byzantine paper's degree-based hub
selection.

The paper does report a known estimator bias tied to sample size: in its
synthetic ground-truth ablation (Section 5.2), "as n increases our estimator
becomes smoother and well approximates the GT up to a slight over-estimation,
a repeatedly observed phenomenon" (citing the intrinsic-dimension-estimation
survey Campadelli et al. 2015). This is a **magnitude bias that shrinks
monotonically as sample size grows**, not a sign-reversing artifact tied to
*which* points are chosen — a materially different failure mode from the
Byzantine study's hub-selection problem, where the direction of the inferred
trend itself flipped depending on selection rule.

## Claim: Gutiérrez-Fandiño et al.'s persistence-diagram-distance method does not subsample the network at all — it uses every neuron and connection at every training step

Gutiérrez-Fandiño, Pérez-Fernández, Armengol-Estapé & Villegas (2021) build
their simplicial complex from the complete weighted, directed graph of the
network at each training batch: "For every training state, neural network
connections are considered as directed and weighted edges between neurons,
represented by graph nodes." Their own stated limitation is explicit that no
node- or weight-level subsampling occurs anywhere in the pipeline: "we do not
simplify the graph representation of the neural networks (we keep every
single neuron and connections) and we do not approximate any computation."
The paper's stated cost of this choice is compute, not sampling risk — 7 days
on an HPC machine — which the authors flag as the method's "main limitation,"
explicitly a scalability problem rather than a selection-bias one.

Because there is no node-selection step in this method's design, a
hub-selection-style artifact (which specific nodes get sampled shifting the
sign of the signal) has no point of entry — the method reads the whole
network every time. This stands in direct contrast to
[[claim-hub-selection-artifact-can-reverse-network-breakpoint-signal]], where
the vulnerability exists precisely because the Byzantine study's trade
network is too large to use in full and must be subsampled.

## Claim: The general TDA subsampling-stability literature bounds instability from noise, density, and outliers — not from selection by a structural covariate

Two Tier-1 papers anchor the statistical theory behind subsampling persistence
diagrams, and neither addresses degree-like structural selection. Chazal,
Fasy, Lecci, Michel, Rinaldo & Wasserman (2014/ICML 2015) prove subsampled
persistence-diagram/landscape estimators are "stable with respect to
perturbations of the underlying measure" and that their estimator "is robust
to the presence of outliers" — their subsamples are drawn i.i.d. at random
from the point cloud, and stability is proven against distributional noise,
not against a choice of *which* systematic subset (e.g., top-degree nodes) to
sample. Stolz (2023, JMLR) surveys the two standard landmark-selection rules
used to subsample point clouds for PH — "random selection tends to favour
dense areas of the data while the maxmin algorithm is very sensitive to
noise" — and proposes a topology-preserving alternative, explicitly framed
around noise and density bias, with "robustness to outliers" as the paper's
own evaluation criterion. This matches the vault's own prior framing of the
gap in [[question-tda-neural-net-sampling-artifact-risk]]: "persistence-diagram
stability theorems bound instability under noise, not necessarily under
selection bias." Nothing in this literature proves PH-based estimators are
immune to a hub-selection-style artifact; it simply has not been the question
these stability results were built to answer.

## Claim: No hub-selection-style artifact is documented, tested, or ruled out in either founding NN-PH generalization paper — the vulnerability is neither confirmed nor formally excluded

Reading both papers' full method, ablation, and limitations sections directly
turned up no experiment resembling the Byzantine study's robustness check
(comparing a structural selection rule against a coverage-preserving
alternative and showing the inferred trend's sign changes). Birdal et al. test
sensitivity to sample *size* (random subsampling at varying n) but not to
selection *rule*; Gutiérrez-Fandiño et al. avoid the question by not
subsampling at all. `[unverified — could not confirm or deny the core question
for the field as a whole after this search]`: this capture establishes that the
specific vulnerability is absent by design from these two papers' methods (one
has no selection step; the other's only sampling axis is random and produces a
different, non-sign-reversing bias), and that the general TDA subsampling
literature addresses a different threat model (noise/density/outliers, not
structural selection). It does not establish that no PH-based NN diagnostic
anywhere is vulnerable — that would require surveying diagnostics that *do*
subsample by a structural rule (see Further leads), which this session did not
locate.

> [!note] Seek's commentary:
> The honest shape of this answer is "the specific mechanism doesn't transfer
> because the two papers don't do the thing that would let it transfer" —
> Birdal's estimator samples randomly over time, not over a network's
> structure, and Gutiérrez-Fandiño's method refuses to subsample at all. That's
> a real answer, but it's a answer about *these two papers*, not about the PH
> generalization-diagnostics literature broadly. The more interesting stress
> test would be a method that *does* select training checkpoints or weights by
> some structural rule (loss-based, magnitude-based, layer-based) the way the
> Byzantine paper selected trade-network nodes by degree — I didn't find one
> this session, which is itself a finding: the analogy the original question
> raised assumes a selection step that, in the two papers actually anchoring
> the vault's PH/generalization bridge, doesn't exist. — Seek

## Further leads

- Rieck et al. 2019 "Neural Persistence" (cited as [RTB+19] in Birdal et al.'s related work) computes topology directly on trained layer weights rather than a trajectory sample — worth checking whether its construction has a weight-magnitude selection step that could carry a hub-selection-style risk.
- Corneanu et al. 2020 (CVPR), "Computing the Testing Error Without a Testing Set" — both NN-PH papers in this capture criticize its correlation-based metric-space construction as "numerically brittle"/poorly scaling; worth checking if that brittleness is itself a selection-sensitivity issue.
- Chowdhury et al. 2019, "Path homologies of deep feedforward networks" (directed-homology representation) — cited as related work by Gutiérrez-Fandiño et al.; sampling method not checked this session.
- The Byzantine study's own fix — a minimum-coverage sampling rule preserving regional representation ratios — has no documented analog in NN-PH diagnostics found this session; a dedicated search for "adversarial subsampling" or "checkpoint selection bias" in PH-based generalization literature could close this gap.

## Entity candidates

- Tolga Birdal — person — first author of the PHD/generalization-bound paper anchoring one side of the vault's PH bridge
- Umut Şimşekli — person — co-author on Birdal et al.; also connects to heavy-tailed-SGD generalization literature already touched by the vault
- Asier Gutiérrez-Fandiño — person — first author of the validation-set-free PH-diagram-distance paper
- Frédéric Chazal — person — leading statistical-TDA theorist; first author of the foundational subsampling-stability paper
- Bernadette J. Stolz — person — author of the outlier-robust landmark-selection paper; recent (2023) entrant to the subsampling-for-PH literature
- persistent homology dimension (PHD) — concept — the specific estimator Birdal et al. define, distinct from generic persistent homology
- landmark selection (topological data analysis) — concept — the general open problem of choosing which points to subsample for PH; directly the axis the hub-selection question probes
- Neural Persistence (Rieck et al. 2019) — term — a third PH-based network diagnostic, computed on trained weights rather than a trajectory, not yet examined in the vault