Pith. sign in

REVIEW 2 major objections 4 minor 72 references

A Riemannian View on Active Subspaces

T0 review · 2 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Active subspaces generalize to Riemannian manifolds by parallel-transporting gradients to a central tangent space, where intrinsic and extrinsic views agree to second order.

desk verdict A genuinely new intrinsic generalization of active subspaces with a solid core (Lemma 1, Prop. 3), but the advertised ridge-recovery rate goes beyond what the proofs actually establish. read the letter →

arxiv 2607.25163 v1 pith:SWO5NTPK submitted 2026-07-28 math.NA cs.LGcs.NAmath.DG

Zachary Grey This is my paper · ORCID
classification math.NAcs.LGcs.NAmath.DG MSC 62R3053B2162H2565F15
keywords activesubspacesRiemannianmanifoldsparalleltransportmanifold-geodesicsdimensionreductionshapeanalysispreshapespaceshyperspheres
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

The paper extends active subspaces, the eigenvalue-ordered average of squared directional derivatives that reveals which parameter directions change a scalar output most, from Euclidean domains to functions defined on Riemannian manifolds such as spheres. The key move is to define an intrinsic average tensor by parallel-transporting each Riemannian gradient along the unique geodesic to a central tangent space. The central theorem shows that this intrinsic average and the standard embedding-based gradient average, compressed to the same central tangent space, agree to second order in the geodesic radius of the sampled domain: eigenvalues agree to O(R^2) and dominant eigenspaces to O(R^2/η) under a spectral gap η. This validates active manifold-geodesics as the principled manifold generalization of active subspaces, while warning against extending activity by projecting the dominant ambient eigenvector. Because landmark and elastic-curve preshape spaces are hyperspheres, the construction applies directly to shape analysis.

What carries the argument

The central object is the parallel-transport gradient outer product (Definition 2), a (0,2)-tensor on the central tangent space obtained by radial parallel transport of the Riemannian gradient before averaging against a wrapped measure. Its eigendecomposition defines active manifold-geodesics (Definition 3): exponential images of spans of leading eigenvectors. Lemma 1 gives the exact mean-squared-directional-derivative reading of eigenvalues; Lemma 2 gives the closed-form projection-versus-transport identity on hyperspheres, which Proposition 3 integrates into the O(R^2) agreement bound.

What would settle it

On the unit 2-sphere, center at the north pole, sample a wrapped uniform measure on a geodesic ball of radius R, and use a function with known gradient, e.g. f(x)=a·x. Numerically evaluate ||E0^T C_ι E0 − G0||_2 for shrinking R and verify it scales as O(R^2) with the explicit constant 1+R^2/4 from Lemma 2 as R approaches the injectivity radius. A rate slower than quadratic or a violation of the stated constant would refute Proposition 3.

Watch

Extended reading notes

Core claim

The central claim is Proposition 3: on a Riemannian manifold, for a scalar function sampled on a geodesic ball of radius R below the injectivity radius, the intrinsic parallel-transport gradient average G0 and the extrinsic embedding-based average C_ι restricted to the central tangent space satisfy ||E0^T C_ι E0 − G0||_2 ≤ C R^2 sup ||∇f||^2, with C explicit on hyperspheres. Consequently the two spectra agree to O(R^2) and, under a spectral gap η, the dominant eigenspaces agree to O(R^2/η). Lemma 1 makes the intrinsic eigenvalues exact: each eigenvalue is the mean-squared directional derivative along the eigenvector field obtained by inverse parallel transport. Together these establish activ

Load-bearing premise

The construction requires the samples to lie in a geodesic ball of radius R below the injectivity radius of the central point, so each point connects to that point by a unique geodesic and radial parallel transport is well-defined; outside such a ball the single-frame active manifold geodesics are not defined and the eigenvalue ordering loses its stated meaning.

Editorial extensions

If this is right

  • On any Riemannian manifold, the eigenvalues of G0 retain the exact interpretation as mean-squared directional derivatives along unit-norm transported fields, so the importance ordering survives on curved domains.
  • Within a geodesic ball of radius R below the injectivity radius, intrinsic and extrinsic eigenvalue decompositions agree to O(R^2) and dominant eigenspaces to O(R^2/η), so both perspectives are locally consistent for identifying activity.
  • Extending activity by projecting the dominant ambient eigenvector can fail: wherever that eigenvector lies in the normal space, its projection vanishes and the ordering is conflated; one must instead eigendecompose the centrally projected representation or use intrinsic transport.
  • Because landmark preshape and length-normalized elastic-curve preshape spaces are hyperspheres, the construction gives a response-driven dimension reduction for functions of shape along preshape geodesics.
  • When trailing eigenvalues vanish, f is constant along the inactive distribution; if that distribution Lie-generates a foliation, f becomes a manifold ridge function over the leaves, and otherwise it is constant outright.

Reading between the lines

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

  • A practical audit rule follows: estimate the injectivity radius of a data-driven manifold and choose sampling radii R small enough that the O(R^2) term is well below the spectral gap η; otherwise the intrinsic and extrinsic spectra are indistinguishable.
  • The general-manifold constant in Proposition 3 is left qualitative; the hypersphere identity provides a template for deriving quantitative constants on other symmetric spaces.
  • The foliation-versus-constant dichotomy suggests a testable criterion: if a function's inactive directions fail to bracket-generate on a connected domain, the data are consistent with a low-dimensional manifold ridge.
  • A direct numerical check on a nonspherical manifold, such as the SPD manifold, would test whether the O(R^2) scaling holds beyond hyperspheres and whether the eigenspace error scales as O(R^2/η).
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes an intrinsic generalization of Euclidean active subspaces to scalar functions on Riemannian manifolds. The central construction, active manifold-geodesics (AMG), replaces the Euclidean gradient outer product (4) with a parallel-transported gradient outer product G0 on a central tangent space (Def. 2, Eq. (12)). Its eigenvalues are shown to be exact mean-squared directional derivatives along a transported eigenvector frame (Lemma 1). The paper then contrasts this intrinsic construction with the extrinsic embedding-based average C_ι (Eq. (16)). The main quantitative result is Prop. 3: on a geodesic ball of radius R below the injectivity radius, the centrally projected extrinsic matrix E0^T C_ι E0 and the intrinsic G0 agree to O(R^2), with eigenvalues agreeing to O(R^2) and dominant eigenspaces to O(R^2/η) under a spectral gap η. A closed-form projection-versus-transport identity is given on hyperspheres (Lemma 2). The paper also claims a 'curvature-limited quadratic rate' for ridge recovery in normal coordinates and illustrates the formalism on the 2-sphere, with applications to preshape spaces motivated.

Significance. If the results hold, the paper supplies a principled, coordinate-free extension of active subspaces to manifold-valued parameter domains, retaining the eigenvalue-ordered interpretability of the Euclidean theory. It makes a precise quantitative statement about when the intrinsic and extrinsic perspectives coincide, and it identifies a concrete failure mode of the common 'project-then-order' practice (Remark 7). The paper is largely self-contained: numbered definitions are clear, Lemma 1, Thm. 1, Lemma 2, and the hypersphere case of Prop. 3 are proven in detail, and the explicit hypersphere constant is a genuine strength. The numerical experiments, while simple, do illustrate the mechanisms. The main weaknesses are an unproven transfer of the Euclidean ridge-approximation bound to manifolds (§2.9) and a sketched, not fully tracked, argument for the general-manifold case of Prop. 3.

major comments (2)
  1. [§2.9 and Abstract] The claim 'The same bound (8) therefore holds on the manifold up to O(R^2) corrections' is not entailed by the preceding derivation. Prop. 3 establishes O(R^2) agreement between the matrices E0^T C_ι E0 and G0, and the Weyl/Davis–Kahan consequences for eigenvalues and eigenspaces. Bound (8), however, is an L2 approximation statement with h the conditional average over inactive coordinates; its Euclidean proof requires the domain to be centered and whitened (∫ x ρ dx = 0 and ∫ xx^T ρ dx = I_n, §1.5). Def. 2 imposes no such normalization on the wrapped measure µ, and whitening would change the coordinate directions whose eigenpairs are being compared. An O(R^2) eigenvalue agreement does not by itself produce the conditional average h or control the L2 error in (8). Thus the abstract's 'derived ridge recovery at a curvature-limited quadratic rate' is an extrapolation rather than a theorem.
  2. [Prop. 3 / Appendix D, Step 4] The general-manifold part of Prop. 3 is currently a proof sketch rather than a complete proof. Step 4 states 'Taylor’s theorem with integral remainder gives F(t)[w] = E0w + tII(x̄,w) + O(t^2)' and 'applying Grönwall’s inequality gives V(t) = V(0) + O(t^2)', but the remainders are not tracked and the claimed dependence of the constant C on the second fundamental form and curvature is not demonstrated. Since Prop. 3 is the central theorem, the statement that C depends only on the second fundamental form of ι(M) and its curvature over the support needs to be substantiated with a complete argument, or the theorem should be stated with C as an unspecified geometric constant and the proof supplied in an appendix. The hypersphere case, with the explicit constant C = 1 + R^2/4, is fine.
minor comments (4)
  1. [§1.5 / §2.7] The Euclidean assumption 'without loss of generality ... centering and rescaling ... ∫ x ρ dx = 0 and ∫ xx^T ρ dx = I_n' is not mirrored in Def. 2, where µ is arbitrary. Since this mismatch underlies the ridge-transfer issue, it would help to state explicitly in §2.7 that no centering/whitening is assumed in the manifold construction.
  2. [Prop. 3 display] The same symbol C is used both as the generic constant in ||E0^T C_ι E0 - G0||_2 ≤ C ∫ ... and as the specific value C = 1 + R^2/4 on hyperspheres. The notation would be clearer if the constant carried a subscript, e.g., C_R.
  3. [Appendix B, Eq. (23)] Equation (23) contains a typographical artifact: 'τ 1−n −τ− − − − → τ→0 ∞' should read '→ ∞ as τ → 0'.
  4. [References] Reference [13] has a typo in the title: 'Emergine Ideas' should be 'Emerging Ideas'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Prop. 3 and Lemma 1 are derived from independent definitions; the §2.9 ridge-transfer is an unproved extrapolation, not a circular reduction.

full rationale

The central derivation chain is self-contained. G0 (Def. 2) and C_iota (16) are defined independently from transported and extrinsically embedded gradients, and Prop. 3 derives their O(R^2) agreement from Lemma 2 (hypersphere projection-vs-transport identity) plus standard Weyl/Davis–Kahan spectral perturbation arguments, with full proofs in Appendix D. Lemma 1's 'exact spectral interpretation' is a direct consequence of the defining quadratic form of G0 and the isometry property of parallel transport, not an imported or fitted result. No parameter is fitted and then renamed as a prediction; the numerical O(R^2) ridge recovery is a consequence claimed from Prop. 3, not an input to it. The author's own prior work ([28]–[31], [68]) appears only as applications or outlook references and is not load-bearing for the main theorems. The only fragile passage is §2.9's transfer of the Euclidean bound (8) to manifolds: 'Applied to the pullback f∘exp_p0 ... The same bound (8) therefore holds on the manifold up to O(R^2) corrections.' This skips the centering/whitening hypotheses required by (8) and is an unsupported extrapolation or correctness gap, but it is not circular—it does not reduce the manifold claim to an equivalent input by construction. The paper also explicitly scopes its own validity to geodesic balls within the injectivity radius and states in the conclusions that the construction 'remains geodesically local,' further evidence that no circularity is being hidden.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

All nonstandard inputs are explicit domain restrictions (normal neighborhoods, wrapped full-support measures, smoothness) rather than hidden fitted quantities; no new physical entities are postulated. The only user-chosen scale is R.

free parameters (1)
  • Geodesic ball radius R
    User-selected support radius in Def. 3 and Prop. 3; controls the O(R^2) locality/error but is not fitted to data.
assumptions (5)
  • domain assumption f is differentiable (C^1 or smooth as stated) with square-integrable Riemannian gradient on X.
    Invoked in Def. 2, Lemma 1, Prop. 1; gradient outer product needs a well-defined Riemannian gradient.
  • domain assumption X = supp mu is a closed geodesic ball of radius R < injectivity radius about p0, and mu is a wrapped full-support probability measure pushed forward from T_p0 M.
    Required for a unique radial geodesic and well-defined parallel transport in Def. 2/Prop. 3; the paper explicitly restricts to this setting.
  • standard math Standard Riemannian geometry facts: Levi-Civita connection, exponential map, normal coordinates, parallel transport isometry, Nash embedding, Davis-Kahan/Weyl inequalities.
    Used throughout (Appendix A, C, D); standard theorems from Lee, Do Carmo, Golub-Van Loan.
  • domain assumption For Prop. 3 on general manifolds, the second fundamental form and curvature of the embedding are uniformly bounded over the support and the O(t^2) frame expansions hold.
    Needed for the qualitative general-manifold bound; on hyperspheres the constant is explicit via Lemma 2.
  • domain assumption Numerical recovery examples assume a positive spectral gap eta for dominant eigenspace stability (Davis-Kahan).
    Stated in regime (ii) of Remark 6 and Fig. 7 note; without a gap, eigenspace recovery is not meaningful.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Riemannian View on Active Subspaces." pith.science (2026). https://pith.science/paper/SWO5NTPK

@misc{pith2026260725163,
  author       = {Pith},
  title        = {Pith review of: A Riemannian View on Active Subspaces},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SWO5NTPK}},
  note         = {Machine review of arXiv:2607.25163}
}
read the original abstract

Active subspaces provide an explainable, eigenvalue-ordered principle for studying how scalar-valued quantities of interest change the most, on average, over a reduced basis of Euclidean domains. Composition with parallel transport generalizes this principle from Euclidean space to quantities of interest defined over Riemannian manifolds, and the resulting intrinsic formulation is contrasted with the extrinsic, embedding-based gradient average of manifold learning. Either strategy is studied in an intrinsically local sense, restricted to mean-centered geodesic-balls, and within that scope the two are not identical: on the central tangent space, eigenvalues agree to second order in the geodesic radius of the sampled domain, while dominant eigenspaces agree at the same order relative to the spectral gap. Extending activity beyond that central space then calls for either recomputed decompositions over changing tangent spaces or, intrinsically, parallel transport of a single central frame. Hyperspheres are emphasized throughout as a particular manifold of interest, motivated by applications over preshape spaces for statistical shape analysis. Numerical examples over the 2-sphere illustrate the formalism, including the derived ridge recovery at a curvature-limited quadratic rate.

Figures

Figures reproduced from arXiv: 2607.25163 by the authors.

Figure 1
Figure 1. A sphere, ι(S 2 ) ⊂ R 3 , shown with elliptical neighborhood and associ￾ated image under the exponential map. and parametrization (10) φ . .= expp ◦ E : TpM ∼= R n → U. By naturality of the exponential [43], exp−1 p (q) is identi￾fied with a vector in ambient coordinates, so composition in (9) is simple matrix-vector multiplication. In a data￾driven setting E may be given by an approximated basis [22] or some other … view at source ↗
Figure 2
Figure 2. The two fields of the same tangential gradients over a hemisphere of ι(S 2 ). Colors cor￾respond to function values registered at the same samples across panels. Left (intrinsic): the construction (12) carries each Riemannian gradient to the central tangent space by parallel transport along its unique radial geodesic (black curves through the central point). Right (extrinsic): the construction (16) reads the same gr… view at source ↗
Figure 3
Figure 3. AMG of an ambient linear function restricted to ι(S 2 ) ⊂ R 3 . Left (a): Ambient linear function over the sphere with a subset of random points and corresponding gradients. The AMG W1 (rose great circle) is shown; the projection to Txb0 of the dominant eigenvector of (16) (the extrinsic direction) coincides with the AMG here (d ≈ 4×10−4 ) and is therefore not drawn as a separate trace. Right (b): Function evaluatio… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Transported versus projected extension of a dominant direction over a hemisphere of ι(S 2 ), colored by field magnitude. Left: the intrinsic frame field w1(p) = P −1 p [w1] of (13)—one central eigendecomposition propagated by the isometric transport—is unit-norm everyw…
Figure 5
Figure 5. Figure 5: The ambient ridge f = a ⊤xb in normal coordinates over a geodesic ball at the Karcher mean xb0: color field f ◦ expxb0 , black arrows the tangential gradient parallel-transported to xb0 (the G0 integrand). The intrinsic active axis w1 (solid rose) aligns with steepest …
Figure 6
Figure 6. Figure 6: Numerical demonstration of O(R2 ) ridge recovery over a shrinking geodesic ball—the second-order locality of Prop. 3. Top: An ambient linear function. Bottom: An ambient quadratic with preferential directions (non-ridge). Left: f ◦ expxb0 against the active geodesic co…
Figure 7
Figure 7. Figure 7: Numerical demonstration of O(R2 ) ridge recovery over a shrinking geodesic ball—the second-order locality of Prop. 3. Top: A nonlinear ridge. Bottom: A nonlinear non-ridge. Left: f ◦expxb0 against the active geodesic coordinate s1 = ⟨exp−1 xb0 (xb), w1⟩, with the sampl…
Figure 8
Figure 8. Figure 8: Convergence of approxi￾mated dominant G0 eigenspace using Al￾gorithm 1 to the orthogonal projection of the known ambient ridge direction, π0(a). Depicted is the average error￾metric over 500 re-sampled tangent gra￾dients for each of the N total samples used in Algorith…
Figure 9
Figure 9. Figure 9: Active manifold-geodesics of an ambient linear function restricted to ι(S 2 ) ⊂ R 3 . Left (a): Ambient linear function over the sphere with a subset of random points and corresponding gradients. The AMG (rose great circle) is shown with the transverse (inactive) direc…
Figure 10
Figure 10. Figure 10: Active manifold-geodesics of an ambient quadratic function with ordered directional derivatives on ι(S 2 ) ⊂ R 3 . Left (a): Ambient quadratic function over the sphere with subset of random points and corresponding gradients. The active manifold-geodesic (rose line) i…
Figure 11
Figure 11. Figure 11: Gallery of toy examples over ι(S 2 ). Rows: linear ridge; degenerate (aligned) linear ridge; quadratic ridge; ordered quadratic; nonlinear ridge; nonlinear non-ridge. Columns: normal coordinates at xb0; the function on the sphere (gradient arrows scaled per panel to t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 10 linked inside Pith

  1. [1]

    Absil, R

    P.-A. Absil, R. Mahony, and R. Sepulchre. Riemannian ge- ometry of grassmann manifolds with a view on algorithmic computation.Acta Applicandae Mathematica, 80(2):199–220, 2004

    2004

  2. [2]

    Absil, R

    P.-A. Absil, R. Mahony, and R. Sepulchre.Optimization Al- gorithms on Matrix Manifolds. Princeton University Press, 3 Market Place, Woodstock, Oxfordshire OX20 1SY, 2008

    2008

  3. [3]

    Cam- bridge University Press, Cambridge, 2020

    Andrei Agrachev, Davide Barilari, and Ugo Boscain.A Com- prehensive Introduction to Sub-Riemannian Geometry, volume 181 ofCambridge Studies in Advanced Mathematics. Cam- bridge University Press, Cambridge, 2020

    2020

  4. [4]

    What regularized auto- encoders learn from the data-generating distribution.The Journal of Machine Learning Research, 15(1):3563–3593, 2014

    Guillaume Alain and Yoshua Bengio. What regularized auto- encoders learn from the data-generating distribution.The Journal of Machine Learning Research, 15(1):3563–3593, 2014

    2014

  5. [5]

    J. J. Alonso, M. S. Eldred, P. Constantine, K. Duraisamy, C. Farhat, G. Iaccarino, and G. Jakeman. Scalable envi- ronment for quantification of uncertainty and optimization in industrial applications (SEQUOIA). In19th AIAA Non- Deterministic Approaches Conference, pages 1–19, Grapevine, TX, 2017

    2017

  6. [6]

    J. J. Alonso, M. S. Eldred, P. G. Constantine, K. Duraisamy, C. Farhat, G. Iaccarino, and J. Jakeman. Scalable environment for quantification of uncertainty and optimization in industrial applications (SEQUOIA).19th AIAA Non-Deterministic Ap- proaches Conference, 2017

    2017

  7. [7]

    Berger, R

    E. Berger, R. Bryant, and P. Griffiths. Some isometric embed- ding and rigidity results for riemannian manifolds.Proceedings of the National Academy of Sciences, 78(8):4657–4660, 1981

    1981

  8. [8]

    Bhattacharya and V

    R. Bhattacharya and V. Patrangenaru. Large sample theory of intrinsic and extrinsic sample means on manifolds. I.The Annals of Statistics, 31:1–29, 2003

    2003

Show all 72 references
  1. [9]

    R. A. Bridges, A. D. Gruber, C. R. Felder, M. Verma, and C. Hoff. Active manifolds: A non-linear analogue to active subspaces.arXiv preprint arXiv:1904.13386, 2019

    1904 arXiv

  2. [10]

    Chatterjee

    A. Chatterjee. An introduction to the proper orthogonal de- composition.Current Science, Vol. 78(No. 7), April 2000

    2000

  3. [11]

    B. Chen, S. Dai, and Z. Yu. Intrinsic minimum average vari- ance estimation for sufficient dimension reduction with sym- metric positive definite matrices and beyond.arXiv preprint arXiv:2302.13059, 2023

    2023 arXiv

  4. [12]

    ¨Uber systeme von linearen partiellen dif- ferentialgleichungen erster ordnung.Mathematische Annalen, 117:98–105, 1939

    Wei-Liang Chow. ¨Uber systeme von linearen partiellen dif- ferentialgleichungen erster ordnung.Mathematische Annalen, 117:98–105, 1939

    1939

  5. [13]

    P. G. Constantine.Active Subspaces: Emergine Ideas in Di- mension Reduction for Parameter Studies. Society for Indus- trial and Applied Mathematics, 3600 Market Street, 6th Floor, Phildelphia, PA 19104-2688, 2015

    2015

  6. [14]

    P. G. Constantine, E. Dow, and Q. Wang. Active subspace methods in theory and practice: Applications to kriging sur- faces.SIAM Journal on Scientific Computing, 36(4):A1500– A1524, 2014

    2014

  7. [15]

    G Constantine, A

    P. G Constantine, A. Eftekhari, J. Hokanson, and R. A. Ward. A near-stationary subspace for ridge approximation.Computer Methods in Applied Mechanics and Engineering, 326:402–421, 2017

    2017

  8. [16]

    P. G. Constantine, M. Emory, J. Larsson, and G. Iaccarino. Exploiting active subspaces to quantify uncertainty in the nu- merical simulation of the hyshot ii scramjet.Journal of Com- putational Physics, 302:1–20, 2015

    2015

  9. [17]

    P. G. Constantine and D. Gleich. Computing active subspaces with monte carlo.arXiv preprint arXiv:1408.0545, 2014

    2014 arXiv

  10. [18]

    P. G. Constantine, Q. Wang, A. Doostan, and G. Iaccarino. A surrogate accelerated bayesian inverse analysis of the HyShot II flight data.52nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, Denver, Col- orado, 2011. AIAA 2011-2037

    2011

  11. [19]

    R. D. Cook.Regression Graphics: Ideas for Studying Regres- sions through Graphics, volume Vol. 482. John Wiley & Sons. Inc., 605 Third Avenue, New York, 2009

    2009

  12. [20]

    M. P. Do Carmo.Differential Geometry of Curves and Sur- faces: Revised and Updated Second Edition. Courier Dover Publications, 2016

    2016

  13. [21]

    Edelman, T

    A. Edelman, T. A. Arias, and S. T. Smith. The geometry of algorithms with orthogonality constraints.SIAM journal on Matrix Analysis and Applications, 20(2):303–353, 1998

    1998

  14. [22]

    P. T. Fletcher, C. Lu, S. M. Pizer, and S. Joshi. Principal geodesic analysis for the study of nonlinear statistics of shape. IEEE transactions on medical imaging, 23(8):995–1005, 2004

    2004

  15. [23]

    C. R. Genovese, M. Perone-Pacifico, I. Verdinelli, and L. Wasserman. Nonparametric ridge estimation.The Annals of Statistics, 42(4):1511–1545, 2014

    2014

  16. [24]

    Glaws, P

    A. Glaws, P. Constantine, and R. D. Cook. Inverse regression for ridge recovery i: Theory.arXiv preprint arXiv, 1702, 2017

    2017

  17. [25]

    Glaws and P

    A. Glaws and P. G. Constantine. Gauss–christoffel quadrature for inverse regression: applications to computer experiments. Statistics and Computing, Jun 2018

    2018

  18. [26]

    Glaws, P

    A. Glaws, P. G. Constantine, J. N. Shadid, and T. M. Wildey. Dimension reduction in magnetohydrodynamics power gener- ation models: Dimensional analysis and active subspaces.Sta- tistical Analysis and Data Mining: The ASA Data Science Journal, 10(5):312–325, 2017

    2017

  19. [27]

    Golub and C.F

    G.H. Golub and C.F. Van Loan.Matrix Computations. Johns Hopkins Univesity Press, 3rd ed. edition, 1996

    1996

  20. [28]

    Z. J. Grey and P. G Constantine. Active subspaces of air- foil shape parameterizations.AIAA Journal, 56(5):2003–2017, 2018. A RIEMANNIAN VIEW ON ACTIVE SUBSPACES 25 Figure 11.Gallery of toy examples overι(S 2).Rows: linear ridge; degenerate (aligned) linear ridge; quadratic ri...

    2003

  21. [29]

    Z. J. Grey, P. G. Constantine, and A. White. Enabling aero- engine thermal model calibration using active subspaces.AIAA Propulsion and Energy Forum, 2019

    2019

  22. [30]

    Z. J. Grey, O. A. Doronina, and A. Glaws. Separable shape ten- sors for aerodynamic design.Journal of Computational Design and Engineering, 10(1):468–487, 2023

    2023

  23. [31]

    Z. J. Grey, N. Fisher, and A. Glaws. Explainable binary clas- sification of separable shape ensembles.Journal of Mathemat- ical Imaging and Vision, 2025. in revision (revised manuscript pending reviewer responses)

    2025

  24. [32]

    Hang and X Wang

    F. Hang and X Wang. Rigidity theorems for compact mani- folds with boundary and positive ricci curvature.Journal of Geometric Analysis, 19(3):628–642, 2009

    2009

  25. [33]

    Hastie and W

    T. Hastie and W. Stuetzle. Principal curves.Journal of the American Statistical Association, 84(406):502–516, 1989

    1989

  26. [34]

    Partially isometric matrices.J

    John Z Hearon. Partially isometric matrices.J. Res. Nat. Bur. Standards Sect. B, 71:225–228, 1967

    1967

  27. [35]

    M Hokanson and P

    J. M Hokanson and P. G Constantine. Data-driven polynomial ridge approximation using variable projection.SIAM Journal on Scientific Computing, 40(3):A1566–A1589, 2018

    2018

  28. [36]

    J. M. Hokanson and P. G Constantine. Data-driven polynomial ridge approximation using variable projection.SIAM Journal on Scientific Computing, 40(3):A1566–A1589, 2018

    2018

  29. [37]

    J. M. Hokanson and P. G Constantine. The lipschitz matrix: A tool for parameter space dimension reduction.arXiv preprint arXiv:1906.00105, 2019

    1906 arXiv

  30. [38]

    Huckemann, T

    S. Huckemann, T. Hotz, and A. Munk. Intrinsic shape analysis: Geodesic principal component analysis for Riemannian man- ifolds modulo isometric Lie group actions.Statistica Sinica, 20(1):1–58, 2010

    2010

  31. [39]

    I. T. Jolliffe.Principal Component Analysis. Springer, 175 Fifth Avenue, New York., 2nd edition, 2002

    2002

  32. [40]

    Joshi, E

    S.H. Joshi, E. Klassen, A. Srivastava, and I. Jermyn. A novel representation for riemannian analysis of elastic curves in Rn.Proc. IEEE Computer Vision and Pattern Recognition (CVPR), pages pp. 1–7, 2007

    2007

  33. [41]

    D. G. Kendall. Shape manifolds, procrustean metrics, and com- plex projective spaces.Bulletin of the London Mathematical Society, 16(2):81–121, 1984

    1984

  34. [42]

    Klassen, A

    E. Klassen, A. Srivastava, M. Mio, and S. H. Joshi. Anal- ysis of planar shapes using geodesic paths on shape spaces. IEEE transactions on pattern analysis and machine intelli- gence, 26(3):372–383, 2004

    2004

  35. [43]

    Lee.Riemannian Manifolds: An Introduction to Curvature

    J. Lee.Riemannian Manifolds: An Introduction to Curvature. Springer-Verlag, New York, 1997

    1997

  36. [44]

    J. M. Lee.An Introduction to Smooth Manifolds, 2nd. ed. Springer, New York, 2003

    2003

  37. [45]

    Lee and K

    K. Lee and K. Carlberg. Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders. arXiv preprint arXiv:1812.08373, 2018

    2018 arXiv

  38. [46]

    Lewis, R

    A. Lewis, R. Smith, and B. Williams. Gradient free active sub- space construction using morris screening elementary effects. Computers & Mathematics with Applications, 72(6):1603– 1615, 2016

    2016

  39. [47]

    Li.Sufficient dimension reduction: Methods and applica- tions with R

    B. Li.Sufficient dimension reduction: Methods and applica- tions with R. Chapman and Hall/CRC, 2018

    2018

  40. [48]

    K-C. Li. On Principal Hessian Directions for Data Visual- ization and Dimension Reduction: Another Application of Stein’s Lemma.Journal of the American Statistical Associ- ation, 87(420):1025–1039, 1992

    1992

  41. [49]

    Loudon and S

    T. Loudon and S. Pankavich. Mathematical analysis and dy- namic active subspaces for a long term model of HIV.arXiv preprint arXiv:1604.04588, 2016

    2016 arXiv

  42. [50]

    Lukaczyk, F

    T. Lukaczyk, F. Palacios, Alonso J. J., and P. G. Constantine. Active subspaces for shape optimization.10th AIAA Multidis- ciplinary Design Optimization Conference, January 2014

    2014

  43. [51]

    Mallasto and A

    A. Mallasto and A. Feragen. Probabilistic Riemannian sub- manifold learning with wrapped Gaussian process latent vari- able models. InProceedings of the 21st International Confer- ence on Artificial Intelligence and Statistics (AISTATS), 2018

    2018

  44. [52]

    W. Mio, A. Srivastava, and S. H. Joshi. On shape of plane elastic curves.International Journal of Computer Vision, 73(3):307–324, 2007

    2007

  45. [53]

    Mukherjee, Q

    S. Mukherjee, Q. Wu, D.-X. Zhou, et al. Learning gradients on manifolds.Bernoulli, 16(1):181–207, 2010

    2010

  46. [54]

    J. Nash. The imbedding problem for riemannian manifolds. Annals of mathematics, pages 20–63, 1956

    1956

  47. [55]

    Nilsson, F

    J. Nilsson, F. Sha, and M. I. Jordan. Regression on manifolds using kernel dimension reduction. InProceedings of the 24th International Conference on Machine Learning (ICML), pages 697–704, 2007

    2007

  48. [56]

    Ozertem and D

    U. Ozertem and D. Erdogmus. Locally defined principal curves and surfaces.Journal of Machine Learning Research, 12:1249– 1286, 2011

    2011

  49. [57]

    V. M. Panaretos, T. Pham, and Z. Yao. Principal flows.Jour- nal of the American Statistical Association, 109(505):424–436, 2014

    2014

  50. [58]

    X. Pennec. Intrinsic statistics on Riemannian manifolds: Ba- sic tools for geometric measurements.Journal of Mathematical Imaging and Vision, 25(1):127–154, 2006

    2006

  51. [59]

    X. Pennec. Barycentric subspace analysis on manifolds.The Annals of Statistics, 46(6A):2711–2746, 2018

    2018

  52. [60]

    Romor, M

    F. Romor, M. Tezzele, A. Lario, and G. Rozza. Kernel-based active subspaces with application to computational fluid dy- namics parametric problems using the discontinuous Galerkin method.International Journal for Numerical Methods in En- gineering, 124(13):2947–2976, 2023

    2023

  53. [61]

    V. H. Schulz. A riemannian view on shape optimization.The Journal of the Society for the Foundations of Computational Mathematics, 2014

    2014

  54. [62]

    Serani and M

    A. Serani and M. Diez. A survey on design-space dimension- ality reduction methods for shape optimization.Archives of Computational Methods in Engineering, 2025

    2025

  55. [63]

    C. Soto, C. Wang, Y. Huang, and X. Chen. Intrinsic Riemann- ian cross-covariance for manifold-valued random objects.arXiv preprint arXiv:2606.10212, 2026

    2026 arXiv

  56. [64]

    Srivastava and E

    A. Srivastava and E. Klassen. Monte carlo extrinsic estimators of manifold-valued parameters.IEEE Transactions on Signal Processing, 50(2):299–308, 2002

    2002

  57. [65]

    Klassen.Functional and Shape Data Analysis

    Anuj Srivastava and Eric P. Klassen.Functional and Shape Data Analysis. Springer Series in Statistics. Springer, New York, 2016

    2016

  58. [66]

    Stinson, D

    K. Stinson, D. F. Gleich, and P. G. Constantine. A randomized algorithm for enumerating zonotope vertices.arXiv preprint arXiv:1602.06620, 2016

    2016 arXiv

  59. [67]

    Walker.The Shapes of Things: A Practical Guide to Dif- ferential Geometry and the Shape Derivative

    S.W. Walker.The Shapes of Things: A Practical Guide to Dif- ferential Geometry and the Shape Derivative. SIAM, Philadel- phia, PA, 2015

    2015

  60. [68]

    Efficient calibration of a tur- bine disc heat transfer model under uncertainty.Journal of Thermophysics and Heat Transfer, pages 1–11, 2020

    Andrew White, Sankaran Mahadevan, Zachary Grey, Jason Schmucker, and Alexander Karl. Efficient calibration of a tur- bine disc heat transfer model under uncertainty.Journal of Thermophysics and Heat Transfer, pages 1–11, 2020

    2020

  61. [69]

    Q. Wu, J. Guinney, M. Maggioni, and S. Mukherjee. Learn- ing gradients: predictive models that infer geometry and sta- tistical dependence.Journal of Machine Learning Research, 11(Aug):2175–2198, 2010

    2010

  62. [70]

    Ying and Z

    C. Ying and Z. Yu. Fr´ echet sufficient dimension reduction for random objects.arXiv preprint arXiv:2007.00292, 2020

    2007 arXiv

  63. [71]

    O. Zahm, P. G. Constantine, Cl´ ementine Prieur, and Y. Mar- zouk. Gradient-based dimension reduction of multivariate vector-valued functions.arXiv preprint arXiv:1801.07922, 2018

    2018 arXiv

  64. [72]

    Zanoni, G

    A. Zanoni, G. Geraci, M. Salvador, A. L. Marsden, and D. E. Schiavazzi. Neural active manifolds: nonlinear dimensionality A RIEMANNIAN VIEW ON ACTIVE SUBSPACES 27 reduction for uncertainty quantification.Journal of Scientific Computing, 105:79, 2025. National Institute of Stan...

    2025

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.