arxiv: 2604.19972 · v1 · submitted 2026-04-21 · 📊 stat.ME

Recognition: unknown

Principal Nested Cones

Yanyan Zhan , Ian L. Dryden , Yuexuan Wu

Authors on Pith no claims yet

Pith reviewed 2026-05-10 01:44 UTC · model grok-4.3

classification 📊 stat.ME

keywords principal nested conesdimension reductionsize and shapecone manifoldnonlinear reductionmorphometricsstatistical shape analysis

0 comments

The pith

Principal Nested Cones reduce high-dimensional cone data to low-dimensional scores that preserve both size and shape.

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

The paper introduces Principal Nested Cones to reduce dimensions in data that sit on cones, where an overall scale factor stands apart from the scale-free structure. Size and shape measurements form one common example of this cone geometry. Most existing approaches either discard the size component or overlook the cone form entirely. PNC builds a chain of nested hypercones and projects each observation downward through them step by step. The resulting scores keep the joint size-shape pattern visible and interpretable, which matters for biological and chemical datasets where both pieces and their interactions carry meaning.

Core claim

Principal Nested Cones represent data through a sequence of nested hypercones and progressively project observations onto lower-dimensional cone spaces. The resulting PNC scores provide low-dimensional representations that jointly capture size-shape variation in an interpretable manner. A fast approximation that combines PCA-based transformation with standard PNC enables computation in ultra-high-dimensional settings. Simulation studies and applications to morphometric, developmental, and molecular data show that the method captures nonlinear size-shape structure and improves representation and reconstruction.

What carries the argument

Principal Nested Cones, a sequence of nested hypercones that separate overall scale from scale-free structure and allow stepwise projection to lower-dimensional cone spaces.

If this is right

PNC scores jointly retain size and shape information for visualization and further analysis.
Representation and reconstruction of the data improve over methods that remove size.
The scores yield interpretable insights in morphometric and developmental datasets.
A fast PCA-based version scales the approach to ultra-high-dimensional cases.
Nonlinear size-shape interactions become visible in the low-dimensional output.

Where Pith is reading between the lines

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

The nested structure could extend to other data types that show hierarchical scale separations.
It may support new models for growth processes where size and shape change together over time.
The fast approximation could be paired with existing manifold methods to handle mixed data types.

Load-bearing premise

The observations must lie on a cone manifold where overall scale separates cleanly from the remaining structure and where a nested hypercone sequence fits the size-shape interactions.

What would settle it

On simulated or real cone data, the PNC scores fail to reconstruct the original size variable or the joint size-shape pattern more accurately than methods that ignore the cone geometry.

Figures

Figures reproduced from arXiv: 2604.19972 by Ian L. Dryden, Yanyan Zhan, Yuexuan Wu.

**Figure 2.** Figure 2: Illustration of PNC modeling. Vectors xi denote observations and x ∗ i their projections onto hypercone surfaces. From Steps 1 to d − 1, observations inside the cones have negative scores and residuals, while those outside are positive. At Step d, observations clockwise from the cone axis are negative, and those counterclockwise are positive. 3.2 Model construction The core of PNC modeling is to identify a… view at source ↗

**Figure 3.** Figure 3: Simulation setup and results. Rows correspond to [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗

**Figure 4.** Figure 4: (a) PNC parameter estimates as a function of sample size [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗

**Figure 5.** Figure 5: (a) Crabs projected on the fitted cone. (b–d) 2D representations: (b) PNC polar [PITH_FULL_IMAGE:figures/full_fig_p023_5.png] view at source ↗

**Figure 6.** Figure 6: Rat skulls on the PNC cone (a) and the PNS unit sphere (b), with longitudinal [PITH_FULL_IMAGE:figures/full_fig_p025_6.png] view at source ↗

**Figure 7.** Figure 7: (a) Seven molecules represented by 2,498 3D landmarks. Colors indicate landmark [PITH_FULL_IMAGE:figures/full_fig_p026_7.png] view at source ↗

read the original abstract

In many applications, the data lie on a type of cone, where there is a distinction between an overall scale variable and the remaining scale-free structure. For example, the joint size and shape of objects are points on a cone, where size represents scale, and shape is the scale-free structure. Dimension reduction is central in such applications, as shape data are often high-dimensional. Interactions between shape and size are widespread and of significant interest in real-world applications. However, most existing methods either lack a single notion of size or focus solely on shape, effectively removing size information. We propose Principal Nested Cones (PNC), a nonlinear dimension reduction framework that preserves both shape and size. PNC represents data through a sequence of nested hypercones and progressively projects observations onto lower-dimensional cone spaces. The resulting PNC scores provide low-dimensional representations that jointly capture size-shape variation in an interpretable manner. To enable scalable computation in ultra-high-dimensional settings, we develop a fast approximation combining PCA-based transformation with standard PNC. Simulation studies and real data applications demonstrate that PNC captures nonlinear size-shape structure, improves representation and reconstruction, and yields interpretable insights across morphometric, developmental, and molecular datasets.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

PNC introduces nested hypercones for size-shape dimension reduction with supporting derivations and empirical checks.

read the letter

Principal Nested Cones is a new framework for dimension reduction on data that lives on cones, where you can separate an overall scale from the rest of the structure. This seems particularly aimed at size and shape analysis in morphometrics and similar fields. The paper does well by spelling out the hypercone construction, working out the projection operators, and giving an optimization routine to fit the nested sequence to data. Simulations are used to check that it can recover the underlying size-shape relationships when they are known. On real data from morphometric, developmental, and molecular sources, it shows better reconstruction than linear methods while keeping the size information in the low-dimensional scores. The approximation that uses a PCA pre-transformation is described as a way to make it scalable, not as a replacement for the full method. One area that could be softer is the dependence on the data actually having that cone geometry with a separable scale factor. The authors note this is a domain-specific assumption rather than something that holds for all data, which is reasonable. Still, outside those settings the method might not add much beyond existing nonlinear dimension reduction techniques. The empirical sections include direct comparisons, which helps, but more detail on robustness to the number of cones or starting points would strengthen it. This is the sort of paper that would be useful to researchers who work with shape data or other cone-structured observations and want interpretable low-dimensional representations that do not discard size. It engages with the relevant literature and has no apparent internal problems in the construction or the way the results are reported. I think it should go to peer review because the idea is new, the math is developed, and there is evidence from both simulations and applications to back up the claims.

Referee Report

2 major / 2 minor

Summary. The paper proposes Principal Nested Cones (PNC), a nonlinear dimension reduction framework for data lying on cone manifolds, such as size-shape data where an overall scale variable is distinguished from scale-free structure. PNC constructs a sequence of nested hypercones, derives projection operators onto successively lower-dimensional cone spaces, and obtains low-dimensional PNC scores that jointly encode size and shape variation in an interpretable way. A scalable approximation combining PCA pre-transformation with standard PNC is introduced for ultra-high dimensions. The method is supported by an explicit optimization procedure for fitting the nested sequence, simulation studies recovering known size-shape structure, and real-data applications in morphometric, developmental, and molecular datasets showing improved reconstruction over linear baselines while preserving scale/shape separation.

Significance. If the geometric construction and empirical results hold, PNC offers a principled alternative to linear methods like PCA for applications where size and shape interact nonlinearly, providing both better representation and direct interpretability of the scale component. The explicit definition of cone geometry, derivation of projections, and optimization procedure, together with direct comparisons in simulations and multiple real datasets, constitute clear strengths. The acknowledgment that the cone assumption is domain-specific rather than universal is appropriate.

major comments (2)

The central claim relies on the data lying on a cone manifold with separable scale and scale-free components. While the manuscript acknowledges this as domain-specific, the simulation design and real-data examples do not include controlled violations of the cone assumption to quantify degradation in score interpretability or reconstruction error when the assumption fails.
The fast approximation via PCA pre-transformation is described as a computational device. However, the error analysis for this step (how the subsequent cone fitting and projections deviate from the exact PNC solution in the original space) is not quantified, which is load-bearing for the claim of scalability in ultra-high dimensions.

minor comments (2)

Notation for the nested cone sequence and the associated projection operators could be introduced with a single diagram or table early in the methods section to improve readability.
The abstract states that PNC 'improves representation and reconstruction' but does not specify the quantitative metrics used; these should be named explicitly.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment and constructive comments. We respond to each major comment below, indicating the changes we will make to the manuscript.

read point-by-point responses

Referee: The central claim relies on the data lying on a cone manifold with separable scale and scale-free components. While the manuscript acknowledges this as domain-specific, the simulation design and real-data examples do not include controlled violations of the cone assumption to quantify degradation in score interpretability or reconstruction error when the assumption fails.

Authors: We agree that demonstrating robustness (or lack thereof) under violations of the cone assumption would provide useful context for readers. Our current simulations and applications focus on settings where the assumption holds, as this is the regime for which PNC is designed. In the revision we will add a controlled simulation study that perturbs data away from the cone manifold (e.g., by introducing additive noise to the scale factor or directional distortions) and quantify the resulting increases in reconstruction error together with the loss of interpretability in the size component. revision: yes
Referee: The fast approximation via PCA pre-transformation is described as a computational device. However, the error analysis for this step (how the subsequent cone fitting and projections deviate from the exact PNC solution in the original space) is not quantified, which is load-bearing for the claim of scalability in ultra-high dimensions.

Authors: This is a fair criticism. Deriving a tight theoretical error bound is nontrivial because of the nonlinearity of the cone projections that follow the PCA step. In the revised manuscript we will add an empirical error analysis: on moderate-dimensional data sets where exact PNC remains computationally feasible, we will directly compare the approximate scores and reconstructions against the exact PNC solution and report the observed deviations. These results will be used to support the practical accuracy of the approximation for ultra-high-dimensional regimes. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper explicitly defines the cone geometry, derives projection operators onto nested hypercones, and supplies an optimization procedure for fitting the sequence. The PCA pre-transformation is presented as a computational approximation rather than a theoretical requirement. Simulations recover known structure independently, and real-data examples compare reconstruction against linear baselines while preserving scale/shape separation. No load-bearing step reduces by construction to fitted inputs or self-citations; the data-cone assumption is treated as domain-specific and externally falsifiable.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Only the abstract is available, so the ledger is necessarily incomplete. The central claim rests on the existence of cone-structured data and the utility of nested hypercone projections; no explicit free parameters, axioms, or invented entities are stated.

invented entities (1)

Principal Nested Cones no independent evidence
purpose: nonlinear dimension reduction that preserves joint size-shape variation
New modeling construct introduced by the paper

pith-pipeline@v0.9.0 · 5505 in / 1155 out tokens · 42170 ms · 2026-05-10T01:44:58.324111+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

33 extracted references · 2 canonical work pages

[1]

2001 , publisher =

A Course in Metric Geometry , author =. 2001 , publisher =

2001
[2]

Marron, J. S. and Dryden, I. L. , title =. 2021 , publisher =

2021
[3]

, organization =

Dryden, Ian L. , organization =. 2025 , note =

2025
[4]

Australian Journal of Zoology , volume=

A multivariate study of variation in two species of rock crab of the genus Leptograpsus , author=. Australian Journal of Zoology , volume=. 1974 , publisher=

1974
[5]

Office of the Medical Investigator , year=

New Mexico decedent image database , author=. Office of the Medical Investigator , year=
[6]

On expected figures and a strong law of large numbers for random elements in quasi-metric spaces , author=. Transactions of the Seventh Prague Conference on Information Theory, Statistical Decision Functions, Random Processes and of the 1974 European Meeting of Statisticians: held at Prague, from August 18 to 23, 1974 , pages=. 1977 , organization=

1974
[7]

Journal of molecular biology , volume=

A simplified representation of protein conformations for rapid simulation of protein folding , author=. Journal of molecular biology , volume=. 1976 , publisher=

1976
[8]

The Annals of Statistics , volume=

Barycentric subspace analysis on manifolds , author=. The Annals of Statistics , volume=. 2018 , publisher=

2018
[9]

2nd edition

Statistical Shape Analysis: with applications in R. 2nd edition. , author=. 2016 , publisher=

2016
[10]

Shape manifolds,

Kendall, David G , journal=. Shape manifolds,. 1984 , publisher=

1984
[11]

Pennec, Xavier and Fillard, Pierre and Ayache, Nicholas , journal=. A. 2006 , publisher=

2006
[12]

Arsigny, Vincent and Fillard, Pierre and Pennec, Xavier and Ayache, Nicholas , journal=. Log-. 2006 , publisher=

2006
[13]

Bookstein, F. L. , title =. 1991 , publisher =

1991
[14]

Vemuri , journal =

Chun-Hao Yang and Baba C. Vemuri , journal =. Nested. 2020 , note =

2020
[15]

Kendall, D. G. , title =. Statistical Science , year =
[16]

Proceedings of the 12th annual conference on Computer graphics and interactive techniques , pages=

Animating rotation with quaternion curves , author=. Proceedings of the 12th annual conference on Computer graphics and interactive techniques , pages=
[17]

1994 , publisher=

A Mathematical Introduction to Robotic Manipulation , author=. 1994 , publisher=

1994
[18]

Marron, J. S. and Jung, S. and Dryden, I. L. , year =. Speculation on the Generality of the Backward Stepwise View of. Proceedings of MIR 2010: 11th ACM SIGMM International Conference on Multimedia Information Retrieval , publisher =

2010
[19]

2025 , eprint=

Principal nested spheres for high-dimensional data , author=. 2025 , eprint=

2025
[20]

Dryden, I. L. and Kim, K. and Laughton, C. A. and Le, H. , year =. Principal nested shape space analysis of molecular dynamics data , journal =
[21]

Pavlos Zoubouloglou and Eduardo García-Portugués and J. S. Marron , title =. Journal of Computational and Graphical Statistics , volume =. 2023 , publisher =. doi:10.1080/10618600.2022.2119985 , URL =

work page doi:10.1080/10618600.2022.2119985 2023
[22]

2025 , url =

R: A Language and Environment for Statistical Computing , author =. 2025 , url =

2025
[23]

2025 , eprint=

Principal Subsimplex Analysis , author=. 2025 , eprint=

2025
[24]

Nature structural & molecular biology , volume=

Announcing the worldwide protein data bank , author=. Nature structural & molecular biology , volume=. 2003 , publisher=

2003
[25]

Development genes and evolution , volume=

Size, shape, and form: concepts of allometry in geometric morphometrics , author=. Development genes and evolution , volume=. 2016 , publisher=

2016
[26]

, author=

Analysis of a complex of statistical variables into principal components. , author=. Journal of educational psychology , volume=. 1933 , publisher=

1933
[27]

Wiley interdisciplinary reviews: computational statistics , volume=

Principal component analysis , author=. Wiley interdisciplinary reviews: computational statistics , volume=. 2010 , publisher=

2010
[28]

2016 , publisher=

Functional and shape data analysis , author=. 2016 , publisher=

2016
[29]

Test , volume=

Shape-based functional data analysis , author=. Test , volume=. 2024 , publisher=

2024
[30]

2013 , publisher=

Modern applied statistics with S , author=. 2013 , publisher=

2013
[31]

and Marron, J

Jung, Sungkyu and Dryden, Ian L. and Marron, J. S. , TITLE =. Biometrika , FJOURNAL =. 2012 , NUMBER =. doi:10.1093/biomet/ass022 , URL =

work page doi:10.1093/biomet/ass022 2012
[32]

I. T. Jolliffe , title =
[33]

2025 , note =

shapes: Statistical Shape Analysis , author =. 2025 , note =

2025