Pith. sign in

REVIEW 5 major objections 5 minor 12 references

Model-Free Kernel Conformal Depth Measures Algorithm for Uncertainty Quantification in Regression Models in Separable Hilbert Spaces

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

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

Pith's one-line read A model-free conformal recipe turns kernel depth bands into prediction regions for functional responses.

desk verdict Useful synthesis for conformal UQ in Hilbert-space-valued regression, but the advertised conditional consistency result leans on an unproved sup-norm assumption the paper itself flags as delicate. read the letter →

arxiv 2506.08325 v1 pith:APGQ5GQT submitted 2025-06-10 stat.ML cs.LGstat.ME

Marcos Matabuena , Rahul Ghosal , Pavlo Mozharovskyi , Oscar Hernan Madrid Padilla , Jukka-Pekka Onnela This is my paper
classification stat.MLcs.LGstat.ME MSC 62G0562G1562G2062R10
keywords conformalpredictiondatadepthkernelmeanembeddingregionsfunctionalseparableHilbertspacestoleranceuncertaintyquantification
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper aims to give a general recipe for prediction regions in regression when both predictor and response live in separable Hilbert spaces, such as functions or probability distributions, where standard ordering tools do not exist. It proposes ranking each candidate response by a conditional depth measure built from conditional kernel mean embeddings, then applying conformal calibration to the ranked values. The central claims are that the conformalized regions satisfy marginal non-asymptotic coverage $P(Y \in \hat C_\alpha(X)) \geq 1-\alpha$, and that under uniform consistency of the embedding estimator the regions converge conditionally to the oracle level sets. A concrete motivation is personalized physical activity recommendations: given age, sex, and BMI, the method outputs a band of plausible activity-distribution profiles. If these claims hold, the framework fills a gap where few competing uncertainty-quantification methods exist for functional responses.

What carries the argument

The central object is the conditional kernel mean embedding $\mu_{Y|X}(x)=\mathbb{E}[\phi(Y)\mid X=x]$, estimated by the ridge-regularized operator $\hat C = \Phi(K_X+n\lambda I)^{-1}\Psi^\top$, which yields a depth value for each response through the kernel feature map. The paper identifies this embedding with an $h$-integrated depth band, so thresholding it gives nested level sets. The theoretical engine also includes the Li\u2013Liu depth-spacing theorem, which makes the probability masses of depth spacings behave like uniform spacings and gives tolerance regions with Beta-distributed coverage. Conformal calibration on held-out splits supplies the quantile thresholds, and a bootstrap variant exploits the asymptotic Gaussianity of the estimators to produce tolerance regions in probability.

What would settle it

Construct a functional-response model with a kernel for which the conditional kernel mean embedding estimators fail the Donsker property that Remark 3 concedes may be absent, then check whether the integrated symmetric difference between the estimated and oracle $\alpha$-level regions converges to zero; a concrete counterexample would settle the conditional-consistency claim.

Watch

Extended reading notes

Core claim

The paper's central discovery is that kernel mean embeddings can serve as conditional depth measures, and that the resulting depth bands can be conformalized to produce prediction regions for responses in separable Hilbert spaces. For a candidate response $y$ at predictor $x$, the method evaluates the estimated conditional kernel mean embedding $\hat\mu_{Y|X}(x)$ and treats the inner product $D_k(y;\hat P_{Y|X}) = \langle \phi(y), \hat\mu_{Y|X}(x)\rangle$ as a ranking score. Conformal calibration converts these scores into regions of the form $\hat C_\alpha(x)=\{y:\hat g(x,y) \geq \hat q_{1-\alpha}\}$ that, by exchangeability, obey $P(Y \in \hat C_\alpha(X)) \geq 1-\alpha$ regardless of model specification. Under Assumption 3, Theorem 4 establishes the stronger asymptotic statement that the integrated symmetric difference between estimated and oracle conditional regions is $o_P(1)$, and in homoscedastic settings the method needs only train/test splits. Heteroscedastic versions use GAMLSS to model how the depth distribution varies with $x$.

Load-bearing premise

The load-bearing premise is Assumption 3(2): the estimated conditional kernel mean embedding must converge uniformly over the whole response space in sup norm, $\|\hat\mu_{Y|X}-\mu_{Y|X}\|_\infty = o_P(1)$, and the paper does not prove this uniform convergence, which can fail for infinite-dimensional responses.

Editorial extensions

If this is right

  • Prediction regions for functional or distributional responses come with a finite-sample marginal coverage guarantee $P(Y \in \hat C_\alpha(X)) \geq 1-\alpha$ that does not depend on the correctness of any regression model.
  • In homoscedastic settings the procedure reduces to a two-split algorithm and still attains the same conformal guarantee.
  • Under Assumption 3, the estimated regions converge to the oracle conditional level sets in integrated symmetric difference, so conditional coverage approaches $1-\alpha$ as sample size grows.
  • The bootstrap variant returns tolerance regions in probability, meaning the random region contains at least $1-\alpha$ of the conditional mass with prescribed confidence $\gamma$.
  • The depth-plus-conformal pattern extends beyond functional data to random objects in separable Hilbert spaces, including distributions equipped with Wasserstein-type kernels.

Reading between the lines

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

  • If the sup-norm consistency assumption highlighted in Remark 3 fails for infinite-dimensional responses, the conditional-consistency theorem no longer follows, but the marginal conformal guarantee would survive; practical users should then read the method as a marginal rather than conditional promise.
  • The method's stated speed advantage over density-level-set estimators comes from kernel smoothing: a natural next benchmark is comparing region volume or sharpness against CQR/HPD-style regions in a functional-to-functional setting.
  • Because conformal guarantees here rely on exchangeability, applying the algorithm to correlated wearable-device time series would require weighted or split-conformal adaptations; the paper lists dependent data as future work.
  • Using a 2-Wasserstein Gaussian kernel makes the response ranking interpretable in quantile space, so the same construction could directly produce clinical reference bands for biomarker distributions.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes prediction and tolerance regions for regression with responses in separable Hilbert spaces, based on conditional kernel mean embeddings interpreted as depth measures. The main algorithms combine a plug-in estimate of the conditional depth distribution with conformal calibration to obtain marginal coverage guarantees, and a bootstrap variant for tolerance regions in probability. The methods are applied to Euclidean, functional, and distributional (2-Wasserstein) responses, with simulation studies and an NHANES accelerometer case study. The paper claims non-asymptotic marginal conformal guarantees, asymptotic conditional consistency (Theorem 4), and fast convergence rates, but the theoretical apparatus is incomplete and at least one central definition appears to invert the intended coverage level.

Significance. If the main claims were rigorously established, this would be a useful contribution: it extends distributional conformal prediction to infinite-dimensional responses, connects depth bands with kernel mean embeddings, and provides a practical implementation in a health application. The simulation comparison in the Euclidean conditional-coverage setting and the functional-data illustration are valuable and suggest the approach works in practice. However, the advertised asymptotic results are currently conditional on an unproved and delicate sup-norm consistency assumption, and the formal coverage statements are not correctly specified. The paper does not provide machine-checked proofs or code; the conformal argument is standard and likely salvageable, but the present manuscript does not yet support its headline claims.

major comments (5)
  1. [§2.3, Eq. (5); Algorithms 1 and 3] As written, the population region has coverage α, not 1−α. Since g(x,Y) = P(D_k(Y;P_{Y|X}) ≤ D_k(Y;P_{Y|X}) | X) is the probability integral transform of the conditional depth, it is Uniform(0,1) under continuity, so P(Y ∈ {y : g(x,y) ≥ q_{1−α}(x)} | X=x) = 1 − q_{1−α}(x) = α when q_{1−α}(x)=1−α. The same inversion appears in Algorithm 1's threshold bD_{⌈n+1−α(n+1)⌉} and in Algorithm 3's use of the empirical (1−α) quantile of depth values. To obtain 1−α coverage the threshold should be the α-quantile of the depth distribution, or the inequality should be reversed. This needs to be corrected and propagated through Theorem 4 and the simulation and application sections.
  2. [§2.3, Assumption 3 and Theorem 4] Theorem 4 is the paper's advertised conditional-consistency result, but it is stated without proof and rests entirely on Assumption 3(2), which requires sup-norm consistency of the conditional kernel mean embedding estimator. Standard kernel-ridge guarantees for conditional mean embeddings are in RKHS norm or L2(P), neither of which implies sup-norm convergence for infinite-dimensional responses; such uniformity requires Glivenko–Cantelli/Donsker conditions on {k_Y(y,·): y∈Y}, and Remark 3 concedes these may fail for infinite-dimensional Y. The theorem also needs extra conditions, for example continuity and level-set stability of g, convergence of the estimated quantile, and uniformity over x. Without a proof or a reference establishing Assumption 3(2), the conditional-consistency claim in the abstract and Section 1.3 is unsupported.
  3. [§2.4, Propositions 5 and 6] Propositions 5 and 6 are stated without proofs. Proposition 6 is the only result supporting the heteroscedastic Algorithm 2's conditional consistency, yet it asserts sup-norm consistency and √n-weak convergence of \widehat{g} under a GAMLSS fit, without specifying how the GAMLSS model is identified with the conditional CDF g(x,y) or how the Donsker condition is verified. The Donsker assumption also appears to conflict with Remark 3 for infinite-dimensional responses. Please provide complete proofs or precise references, and state the regularity conditions on the GAMLSS parameterization explicitly.
  4. [§2.3, Algorithm 2 and Contribution 2] The advertised 'non-asymptotic guarantees of the type P(Y ∈ \widehat{C}_α(X)) ≥ 1−α' are not formalized anywhere: there is no theorem stating the coverage property, the exact quantile index, or the tie-breaking rule, and the roles of Dcalibration and Dtest in Algorithm 2 are not justified by an exchangeability argument. The standard split-conformal argument is available once the fitted depth and \widehat{g} are fixed, but it must be stated carefully because \widehat{g} itself is estimated on Dcalibration and the quantile is computed on Dtest. Please add a formal proposition for the marginal conformal guarantee.
  5. [Abstract and §1.3] The abstract and Section 1.3 promise 'faster convergence rates' and 'fast convergence rates' in certain homoscedastic settings, but the manuscript gives no theorem quantifying rates for \widehat{C}_α; Theorem 4 is only an o_p(1) consistency statement. Please either add a rate result with its assumptions or remove the rate claim from the contributions.
minor comments (5)
  1. [Throughout] Typos and leftover editing notes should be cleaned: 'popululation bases' in §1.1, 'continuty hyphotesis' in Assumption 3, 'hypthesis' in Remark 3, 'bellow' in §1.3, the duplicate 'Matabuena and Petersen (2023)Matabuena and Petersen (2023)' in §4.2, and the editorial note 'why not put more plots in each row? we are wasting a lot of space...' in the Figure 3 caption.
  2. [§1 and §2.4] The paper has duplicate equation numbers: Eq. (1) is used for the regression model and again for the GAMLSS additive predictor, and Eq. (2) is used for both the conditional-coverage target and the empirical-process display; please renumber them.
  3. [Definition 3] Property (vi) says the depth is 'continuous in P' but does not specify a topology or metric on P; please make this formal.
  4. [Algorithm 1 and Corollary 2] Algorithm 1's threshold should be reconciled with Corollary 2: for an α-expectation tolerance region the coverage result suggests r_n = ⌈(n+1)α⌉, which is not what the displayed formula computes; this is related to Major Comment 1.
  5. [Assumption 3(3)] The population quantile condition ('By definition, under continuity hypothesis, q_{1−α}=1−α') assumes a continuous conditional depth distribution; the paper should address atoms and ties, which also matter for the conformal quantile.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the conformal coverage guarantee is external exchangeability, and the asymptotic result is a plug-in level-set consistency theorem resting on a strong, unproved regularity assumption rather than on a fitted target.

full rationale

The central non-asymptotic claim is the standard split-conformal guarantee: after fitting the conditional kernel mean embedding on the training split and the conditional distribution function on the calibration split, the scores computed on the held-out split are exchangeable with a future score, so the empirical (1-alpha) quantile yields marginal coverage. This guarantee does not depend on the consistency of the depth estimator, which is a plug-in trained on a separate split. The asymptotic conditional consistency claim in Theorem 4 is a direct level-set consistency statement: if Assumption 3(2) supplies sup-norm consistency of the estimated conditional mean embedding, then the estimated depth level sets converge to the oracle level sets defined through the true embedding. This is an implication, not a reduction of the conclusion to the input. Assumption 3(2) is substantial and is not proved in the paper; Remark 3 even concedes that Donsker-type conditions may fail for kernel mean embeddings in infinite-dimensional response spaces. That is a genuine correctness and rigor risk, but an unproved hypothesis is not a circular step. The many self-citations in the paper, such as Matabuena and Petersen (2023) and Lugosi and Matabuena (2024), are used for the accelerometer representation and for contextual claims about the literature, not to supply the exchangeability argument or the consistency theorem. The bootstrap tolerance calibration in Algorithm 4 is a standard bootstrap quantile-calibration heuristic rather than a fitted parameter being relabeled as a prediction. No step in the paper reduces, by its own equations, to the quantity it is claimed to predict.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical or mathematical entities are introduced; kernel mean embeddings and depth are existing constructs. The free parameters are algorithmic tuning choices rather than fitted constants for the target coverage.

free parameters (4)
  • kernel bandwidth sigma = unspecified
    Used in the Wasserstein Gaussian kernel in Eq. (7) and in the kernel mean embedding; choice affects the depth ranking.
  • regularization parameter lambda = unspecified
    In kernel ridge regression for the conditional mean embedding in Section 4.3; controls bias-variance tradeoff.
  • GAMLSS smoothing/basis parameters = unspecified
    For the heteroscedastic conditional distribution model in Section 2.4; not specified in the main text.
  • number of bootstrap samples B = unspecified
    In Algorithm 4; should be large for stable estimates.
assumptions (5)
  • domain assumption X and Y are separable Hilbert spaces with universal kernels
    Section 1.2; the entire framework is built on this.
  • standard math Li-Liu theorem on depth spacings
    Used to justify tolerance regions in Section 2.1; requires affine-invariant depth.
  • ad hoc to paper Conditional kernel mean embedding estimator is sup-norm consistent (Assumption 3.2)
    This is a strong assumption that is not proved; it is load-bearing for Theorem 4.
  • ad hoc to paper GAMLSS model correctly specifies the conditional distribution of depth values
    The heteroscedastic algorithm relies on this parametric model for g(x,y) in Section 2.4.
  • standard math Exchangeability of calibration scores
    Underlies the split conformal guarantee; standard and reasonable with i.i.d. data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Model-Free Kernel Conformal Depth Measures Algorithm for Uncertainty Quantification in Regression Models in Separable Hilbert Spaces." pith.science (2026). https://pith.science/paper/APGQ5GQT

@misc{pith2026250608325,
  author       = {Pith},
  title        = {Pith review of: Model-Free Kernel Conformal Depth Measures Algorithm for Uncertainty Quantification in Regression Models in Separable Hilbert Spaces},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/APGQ5GQT}},
  note         = {Machine review of arXiv:2506.08325}
}
read the original abstract

Depth measures are powerful tools for defining level sets in emerging, non--standard, and complex random objects such as high-dimensional multivariate data, functional data, and random graphs. Despite their favorable theoretical properties, the integration of depth measures into regression modeling to provide prediction regions remains a largely underexplored area of research. To address this gap, we propose a novel, model-free uncertainty quantification algorithm based on conditional depth measures--specifically, conditional kernel mean embeddings and an integrated depth measure. These new algorithms can be used to define prediction and tolerance regions when predictors and responses are defined in separable Hilbert spaces. The use of kernel mean embeddings ensures faster convergence rates in prediction region estimation. To enhance the practical utility of the algorithms with finite samples, we also introduce a conformal prediction variant that provides marginal, non-asymptotic guarantees for the derived prediction regions. Additionally, we establish both conditional and unconditional consistency results, as well as fast convergence rates in certain homoscedastic settings. We evaluate the finite--sample performance of our model in extensive simulation studies involving various types of functional data and traditional Euclidean scenarios. Finally, we demonstrate the practical relevance of our approach through a digital health application related to physical activity, aiming to provide personalized recommendations

Figures

Figures reproduced from arXiv: 2506.08325 by the authors.

Figure 1
Figure 1. Physical activity predictive regions for the individuals a) and b) [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Estimated marginal coverage, functional to functional regression ex [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. (a) Women, BMI= 20 kg/m2 ; (b) Women, BMI= 25 kg/m2 ; (c) Women, BMI= 30 kg/m2 . Expected Quantile of Physical Activity Across Different Ages and Body Masses in Healthy Women of the American Population for α = 0.5.why not put more plots in each row? we are wasting a lot of space... 5 Discussion This paper introduces a novel framework for uncertainty quantification in ran￾dom statistical objects defined in separable … view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

12 extracted references · 9 canonical work pages

  1. [8]

    URL https://doi.org/10

    doi: 10.1080/01621459.2018.1543123. URL https://doi.org/10. 1080/01621459.2018.1543123. K. Muandet, K. Fukumizu, B. Sriperumbudur, B. Sch¨ olkopf, et al. Kernel mean embedding of distributions: A review and beyond. Foundations and Trends in Machine Learning , 10(1-2):1–141,

    arXiv 2018

  2. [10]

    B. K. Sriperumbudur, A. Gretton, K. Fukumizu, G. Lanckriet, and B. Sch¨ olkopf. Injective hilbert space embeddings of probability measures. In 21st annual conference on learning theory (COLT 2008), pages 111–122. Omnipress,

    work page 2008

  3. [1997]

    J. W. Tukey. Mathematics and the picturing of data. In Proceedings of the International Congress of Mathematicians, Vancouver, 1975 , volume 2, pages 523–531,

    work page 1975

  4. [2004]

    Causal survival embeddings: non-parametric counterfactual inference under censoring

    C. Garc ´ ıa-Meixide and M. Matabuena. Causal survival embeddings: non- parametric counterfactual inference under censoring. arXiv preprint arXiv:2306.11704,

    work page Pith review arXiv

  5. [2007]

    J. Virta. Spatial depth for data in metric spaces. arXiv preprint arXiv:2306.09740,

    arXiv

  6. [2012]

    Dombry, T

    C. Dombry, T. Modeste, and R. Pic. Stone’s theorem for distributional regres- sion in wasserstein distance. arXiv preprint arXiv:2302.00975 ,

    work page arXiv

  7. [2016]

    URL https://doi.org/10.1214/ 15-AOS1363

    doi: 10.1214/15-AOS1363. URL https://doi.org/10.1214/ 15-AOS1363. K. L. Piercy, R. P. Troiano, R. M. Ballard, S. A. Carlson, J. E. Fulton, D. A. Galuska, S. M. George, and R. D. Olson. The physical activity guidelines for americans. Jama, 320(19):2020–2028,

    doi:10.1214/15-aos1363

  8. [2020]

    K. Bolin. Physical inactivity: productivity losses and healthcare costs 2002 and 2016 in sweden. BMJ Open Sport & Exercise Medicine , 4(1):e000451,

    work page 2002

Show all 12 references
  1. [2021]

    URL https://EconPapers.repec.org/RePEc:spr:alstar:v:105:y:2021: i:2:d:10.1007_s10182-021-00391-y . R. F. Barber, E. J. Candes, A. Ramdas, and R. J. Tibshirani. Conformal prediction beyond exchangeability. The Annals of Statistics , 51(2):816–845,

    2021

  2. [2022]

    Matabuena, J

    M. Matabuena, J. C. Vidal, O. H. M. Padilla, and J.-P. Onnela. knn algorithm for conditional mean and variance estimation with automated uncertainty quantification and variable selection. arXiv preprint arXiv:2402.01635 ,

  3. [2023]

    doi: 10.1093/jrsssc/qlad007

    ISSN 0035-9254. doi: 10.1093/jrsssc/qlad007. URL https://doi.org/10.1093/jrsssc/qlad007. M. Matabuena, P. R. Hayes, and L. Puente-Maestu. Prediction of maximal oxy- gen uptake from submaximal exercise testing in chronic respiratory patients. new perspectives. Archivos de Bronc...

    doi

  4. [2024]

    Matabuena and C

    M. Matabuena and C. M. Crainiceanu. Multilevel functional distributional mod- els with application to continuous glucose monitoring in diabetes clinical tri- als. arXiv preprint arXiv:2403.10514 ,

    arXiv

Pith tools

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