Pith. sign in

REVIEW 1 major objections 4 minor 48 references

Introduction to Nonlinear Spectral Analysis

T0 review · 1 major / 4 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 The paper builds a variational theory of nonlinear eigenfunctions and shows that the late-time shape of a wide class of gradient flows is always one of them.

desk verdict Useful, honest lecture notes on nonlinear spectral analysis; the core theory is solid, but the 2D total-variation flow application rests on an admitted but unproved modification, so the notes overreach slightly in the applications. read the letter →

arxiv 2506.08754 v1 pith:EG6QEEJT submitted 2025-06-10 math.SP math.APmath.OC

Leon Bungert , Yury Korolev This is my paper
classification math.SPmath.APmath.OC MSC 47J1035P3035K5549J52
keywords nonlineareigenvalueproblemssubdifferentialgradientflowsasymptoticprofilestotalvariationflowporousmediumequationRayleighquotientpowermethod
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

These lecture notes set out a unified variational theory of nonlinear eigenvalue problems. For any convex, absolutely p-homogeneous energy functional $J$ on a Hilbert space, an eigenfunction is defined by the inclusion $\lambda\|u\|^{p-2}u \in \partial J(u)$, and the paper proves that solutions of the subgradient flow $u'(t)+\partial J(u) \ni 0$ converge along their normalized trajectory to such an eigenfunction whenever $J$ has precompact sublevel sets. The same machinery yields exact finite extinction times and convergence rates when the smallest eigenvalue $\lambda_1$ is positive. This converts a collection of nonlinear PDEs—fast diffusion, porous medium, total variation, and $L^\infty$ variational problems—into spectral statements: their late-time shapes are nonlinear eigenfunctions, for example multiples of the distance function for the $L^\infty$ flow. The notes also show that an implicit Euler discretization of the flow is a nonlinear power method that computes these eigenfunctions.

What carries the argument

The driving object is the Rayleigh quotient $\Lambda(t) = pJ(u(t))/\|u(t)-u_\infty\|^p$ evaluated along the flow. Proposition 2.70 shows that $\frac{1}{p}\Lambda'(t) + \frac{1}{\|u(t)-u_\infty\|^{2-p}}\|\zeta(t)/\|u(t)-u_\infty\|^{p-1} - \Lambda(t)w(t)\|^2 \le 0$, which forces the direction of the minimal-norm subgradient to align with the normalized trajectory. The subdifferential inclusion $\lambda\|u\|^{p-2}u \in \partial J(u)$ serves as the nonlinear eigenvalue equation, and sublevel-set precompactness of $u \mapsto \|u\| + J(u)$ turns this alignment into actual convergence to an eigenfunction. An implicit Euler discretization of the flow rewrites the iteration as a proximal power method whose fixed points are exactly the nonlinear eigenfunctions.

What would settle it

For total variation flow in two dimensions with a smooth non-eigenfunction initial datum, compute $u(t)/\|u(t)\|$ at closely spaced times; if two different subsequential limits occur, or any limit fails to solve $\lambda w = -\operatorname{div}(\nabla w/|\nabla w|)$, the sketched extension of Theorem 2.73 to TV would be false.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2.73: under the stated hypotheses, for the gradient flow of $J$ with initial datum $f$, the normalized trajectory $w(t) = (u(t)-u_\infty)/\|u(t)-u_\infty\|$ has a subsequence converging to a unit vector $w_\infty$ satisfying $\Lambda_\infty w_\infty \in \partial J(w_\infty)$, meaning the asymptotic profile is a nonlinear eigenfunction. Coupled with the order-preserving uniqueness result of Theorem 2.75, the whole trajectory converges and the limit is uniquely determined. In first-order terms, Propositions 2.63, 2.65, 2.67, and 2.68 give extinction times and sharp convergence rates governed by the first eigenvalue $\lambda_1$. The headline applications are then concrete: the $L^\infty$ flow converges to a multiple of the distance function to the boundary (Theorems 2.80 and 2.83), and fast diffusion, porous medium, and total variation flows converge along normalized trajectories to eigenfunctions (Theorems 2.88 and 2.90).

Load-bearing premise

The central proofs assume that every sequence of functions with bounded energy has a convergent subsequence (precompact sublevel sets), and this assumption fails for total variation in two dimensions, where the paper only sketches a fix.

Editorial extensions

If this is right

  • If $J$ is absolutely one-homogeneous and the minimal subgradients satisfy the dual-unit-ball inequality, the decomposition $f = P_{N_J}f + \int_0^\infty \zeta(t)\,dt$ is a decomposition into eigenfunctions.
  • For $p<2$ with $\lambda_1>0$, the flow reaches its limit in finite time, with the explicit upper bound $T_{ex} \le \|f-u_\infty\|^{2-p}/((2-p)\lambda_1)$, and lower bounds exist in the one-homogeneous case.
  • The $L^\infty$ flow with non-negative initial data converges strongly to a multiple of the distance function to the boundary as $t$ approaches the extinction time.
  • Fast diffusion has finite extinction time and its normalized solution converges to a non-negative eigenfunction, while the porous medium and heat equations have infinite extinction time but still admit an asymptotic profile.
  • The implicit Euler scheme with step sizes $\sigma_k = c/J(w_0)$ or $c/J(w_k)$ is a well-defined nonlinear power method whose iterates converge to an eigenfunction under the same compactness and Poincaré-type assumptions.

Reading between the lines

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

  • The same argument likely extends to non-compact embeddings whenever $L^\infty$-level-set compactness substitutes for sublevel-set compactness, as the paper sketches for two-dimensional total variation; testing this on other BV-type functionals would clarify how far the theorem reaches.
  • The decomposition formula suggests a practical recipe for nonlinear spectral filters on any absolutely one-homogeneous functional: verify the dual-unit-ball inequality for minimal subgradients and then use the flow to separate scales.
  • The proximal power method's convergence proof indicates that the same discretization could serve as a numerical solver for graph $p$-Laplacian and $L^\infty$ eigenproblems, although the paper only develops the Hilbert-space setting.
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

1 major / 4 minor

Summary. These lecture notes develop a variational theory of nonlinear eigenvalue problems for absolutely p-homogeneous convex lower-semicontinuous functionals on Hilbert spaces. They cover Rayleigh quotients, ground states, nonlinear eigenfunctions, gradient flows with finite-time extinction and convergence rates, asymptotic profiles, and nonlinear power methods, with applications to L-infinity distance functions, porous medium/fast diffusion equations, and total variation flow. The abstract framework is built from convex analysis and subdifferential calculus, and many proofs are provided in detail.

Significance. If the identified gaps are repaired, these notes would be a valuable expository contribution: they assemble parameter-free analytical theorems with explicit hypotheses, connect them to concrete PDEs, and include numerical power methods. The strengths are the transparent convex-analytic proofs, the precise statements of assumptions (e.g., the precompactness condition in Theorems 2.27 and 2.73), and the benchmarking against established literature such as [4], [17], and [39]. The applications to distance functions in Section 2.6.1 are self-contained and elegant. However, the TV-flow claims and the p <= 2 uniqueness argument need work before the notes can serve as a reliable reference.

major comments (1)
  1. [§2.6.2, Theorem 2.88] Theorem 2.90, the TV-flow application, is asserted as a consequence of the abstract theory, but the text admits in §2.6.2 that the sublevel sets of u maps to ||u||_2 + TV(u) are not precompact in L^2(Omega) for Omega subset R^2 and that Theorem 2.27 does not apply. The proposed fix, namely that 'one can modify the proof of Theorem 2.73' using compactness of L-infinity level sets, is not carried out, and the suggested route is questionable: the normalized trajectory w(t) = (u(t) - u_Omega)/||u(t) - u_Omega||_2 is not bounded in L-infinity because the denominator tends to zero at the finite extinction time. A sequence in BV(Omega) with fixed L^2 norm and bounded TV need not be relatively compact in L^2 for Omega subset R^2 (for example, v_k = sqrt(k) on B(0,1/k) has ||v_k||_2 = 1 and TV(v_k) -> 0 but no strongly convergent subsequence). Consequently, the proof of Theorem 2.90 is incomplete; the authors should either supply a complete compactness argument or state the theorem with a full reference to a proof.
minor comments (4)
  1. [§1.1, Definition 1.7] The definition of the extended real line as 'R union {+infinity} := R union {±infinity}' is internally inconsistent; the first expression should presumably be 'R union {±infinity}'.
  2. [§2.4, proof of Proposition 2.53] The statement 'Since J(u(t)) is non-decreasing (Theorem 2.48)' should read 'non-increasing', as Theorem 2.48 indeed states that t maps to J(u(t)) is non-increasing.
  3. [§2.6.2, Theorem 2.90] In the statement of Theorem 2.90, after 'there exists a non-decreasing sequence (t_n) ... such that w(t_n) converges to w', the phrase 'as t -> T_ex' should be 'as n -> infinity'; the limit is taken along the sequence index.
  4. [§2.6.2, proof of Proposition 2.85] The proof refers to 'Höder's and Young's inequalities'; this should be 'Hölder's and Young's inequalities'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core asymptotic-profile and ground-state theorems are proved from stated assumptions, with self-citations used only for context or externally checkable results.

full rationale

The main derivation chain is self-contained. Theorem 2.73 proves that any limit point w_infty of the normalized flow is an eigenfunction by combining the gradient-flow identity (2.49), the integrability estimate (2.50), subsequential compactness of sublevel sets of u -> ||u|| + J(u), and lower semicontinuity; the conclusion Lambda_infty w_infty in dJ(w_infty) is derived, not assumed. Theorem 2.80 proves that ground states of the L-infinity functional are multiples of the distance function directly from the Lipschitz estimate |u(x)| <= ||grad u||_infty d_Omega(x), with no circular input. The porous-medium and fast-diffusion results use the external compact-embedding condition from [39] and the abstract theory; Theorem 2.88 is an application, not a renamed input. The total-variation application in Section 2.6.2 is explicitly attributed to the external paper [4] ('first proved in [4]'), and the paper itself flags that Theorem 2.27 does not apply in R^2 because the BV -> L^2 embedding is noncompact; the proposed modification of Theorem 2.73 is an admitted proof gap rather than a circular reduction. Self-citations such as [19], [20], [23], and [24] are used for context or for results that are either proved in the notes or published independently; none is load-bearing in a way that reduces a claim to its own input. There is no fitted parameter renamed as a prediction, no definitional equivalence between an input and a claimed output, and no imported uniqueness theorem that forces the authors' choice by appeal to their own prior work.

Assumptions & free parameters 2 free parameters · 9 assumptions · 0 invented entities

The notes build on standard convex analysis and Brezis's subgradient-flow theory, then impose structural conditions on J: absolute p-homogeneity (p >= 1 by Proposition 2.15), sublevel-set precompactness (Theorems 2.27, 2.73, 3.5), and the Poincare-type bound lambda_1 > 0 (2.34). The precompactness condition is the most fragile: it fails for total variation in R^2, and the paper's patch (L-infinity bounds) is asserted rather than proved. The comparison principle (2.53) needed for uniqueness of profiles is stronger than the positivity preservation cited for the porous medium flow. No constants are fitted to data; the numbers chosen by hand (c in the power method, q in the profile-uniqueness proof) are universal 'for any' parameters that do not affect the statements.

free parameters (2)
  • power method step-size constant c = any value in (0,1)
    In Proposition 3.3 and Theorem 3.5 the proximal power-method step sizes are sigma_k = c/J(w0) or c/J(wk) with 0 < c < 1. The convergence theorem holds for every such c, so the value is irrelevant; it is a universal algorithmic constant, not fitted to data.
  • auxiliary homogeneity q = any q > 2p
    In the proof of Theorem 2.75 for p <= 2 the notes reparametrize the flow to a q-homogeneous functional with r = q/p > 2; any q > 2p works, so this is a universal proof parameter, not a fitted value.
assumptions (9)
  • standard math Brezis's existence-uniqueness-regularity theorem for the subgradient flow (Theorem 2.48)
    Invoked throughout Section 2.4 to define the flows (2.23), obtain the energy identity (2.25), and justify the Bochner integral representation (2.26); cited to [13].
  • domain assumption Precompactness of sublevel sets of u -> ||u|| + J(u)
    Assumed in Theorems 2.27, 2.73, and 3.5 (ground states, asymptotic profiles, power-method convergence). Fails for the total variation functional in R^2 (Section 2.6.2), and the proposed L-infinity-level-set patch is asserted, not proved.
  • domain assumption Poincare-type inequality (2.34): ||u - P_NJ u||^p <= C J(u), equivalent to lambda_1 > 0
    Drives finite extinction (Corollary 2.64), the convergence-rate bounds (Propositions 2.63, 2.67, 2.68), and power-method convergence (Theorem 3.5). Verified for classical Poincare and Sobolev inequalities in Examples 2.59-2.60.
  • domain assumption Comparison principle (2.53) for the semigroup S(t): f >= g implies S(t)[f] >= S(t)[g]
    Assumed in Theorem 2.75 for uniqueness of the asymptotic profile and used in Theorems 2.83, 2.88. The notes cite only positivity preservation ([39]) for the porous medium / fast diffusion flow, which is weaker than the full comparison needed.
  • domain assumption L-infinity a priori bound for the total variation flow, ||u(t)||_Linf <= C
    Invoked in Section 2.6.2 to rescue Theorem 2.73 for the 2D TV flow after noting the precompactness assumption fails; the bound is cited to [4] but the modified argument is not written out.
  • domain assumption Compact embedding L^{m+1}(Omega) into H^{-1}(Omega) for m > (n-2)/(n+2)
    Used in Theorem 2.87 to obtain a positive first eigenvalue and ground states for the porous medium / fast diffusion Rayleigh quotient (2.68); cited to [39].
  • standard math Crandall-Benilan regularization estimate for homogeneous flows (Lemma 2.74)
    Yields the monotone rescaled trajectory t^{1/(p-2)} u(t) used in Theorem 2.75 for p > 2; cited to [27].
  • standard math Kantorovich-Banach space assumption: monotone norm-bounded sequences converge; Hilbert spaces qualify (Remark 2.76)
    Upgrades subsequential profile convergence to actual convergence of w(t) in Theorems 2.83 and 2.88; characterization cited to [41].
  • standard math Standard functional analysis: Riesz representation, Banach-Alaoglu, Rellich-Kondrachov, Fatou's lemma, density of smooth functions
    Background invoked throughout Chapter 1 and in the existence proofs (Theorems 2.27, 2.87) and the compactness arguments of Chapter 3; stated as assumed background on pages 7-13.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Introduction to Nonlinear Spectral Analysis." pith.science (2026). https://pith.science/paper/EG6QEEJT

@misc{pith2026250608754,
  author       = {Pith},
  title        = {Pith review of: Introduction to Nonlinear Spectral Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EG6QEEJT}},
  note         = {Machine review of arXiv:2506.08754}
}
abstract

These notes are meant as an introduction to the theory of nonlinear spectral theory. We will discuss the variational form of nonlninear eigenvalue problems and the corresponding non-linear Euler--Lagrange equations, as well as connections with gradient flows. For the latter ones, we will give precise conditions for finite time extinction and discuss convergence rates. We will use this theory to study asymptotic behaviour of nonlinear PDEs and present applications in $L^\infty$ variational problems. Finally we will discuss numerical methods for solving gradient flows and computing nonlinear eigenfunctions based on a nonlinear power method. Our main tools are convex analysis and calculus of variations, necessary background on which will be provided. It is expected that the reader is familiar with Hilbert spaces; familiarity with Banach spaces is beneficial but not strictly necessary. The notes are based on the lectures taught by the authors at the universities of Bonn and Cambridge in 2022.

Figures

Figures reproduced from arXiv: 2506.08754 by the authors.

Figure 2.1
Figure 2.1. Visualisation of the subdifferential. Linear approximations of the functional have [PITH_FULL_IMAGE:figures/full_fig_p023_2_1.png] view at source ↗
Figure 2.2
Figure 2.2. Four different dual unit balls KJ with all eigenvectors. Theorem 2.36 implies that, if they exist, maximal elements in the dual unit ball are eigenfunctions. Corollary 2.40. Let ζ ∈ arg max{∥η∥ : η ∈ KJ }. Then it holds ζ ∈ ∂J (ζ). Proof. By the maximality of ζ it follows ⟨ζ, η⟩ ≤ ∥ζ∥ ∥η∥ ≤ ∥ζ∥ 2 ∀η ∈ KJ and hence Theorem 2.36 implies the assertion [PITH_FULL_IMAGE:figures/full_fig_p033_2_2.png] view at source ↗
Figure 2.3
Figure 2.3. Left: (2.12) is met since S is supporting, Right: here (2.12) is violated, i.e., S is not supporting and ζ is no eigenvector The next example concerns the L 1 -norm, regarded as absolutely one-homogeneous func￾tional on L 2 (Ω). Proposition 2.43 (L 1 -norm). Let H := L 2 (Ω) and J (u) := (R Ω |u| dx, if u ∈ L 1 (Ω) ∩ L 2 (Ω), +∞, else [PITH_FULL_IMAGE:figures/full_fig_p034_2_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 45 canonical work pages

  1. [4]

    Andreu, V

    F. Andreu, V. Caselles, J. I. Díaz and J. M. Mazón. ‘Some qualitative properties for the total variation flow’.Journal of Functional Analysis 188.2 (2002), pp. 516–547 (cit. on pp. 5, 59)

    work page 2002

  2. [17]

    R. E. Bruck. ‘Asymptotic convergence of nonlinear contraction semigroups in Hilbert space’. Journal of Functional Analysis18.1 (1975), pp. 15–26.issn: 0022-1236 (cit. on p. 39)

    work page 1975

  3. [39]

    Littig and J

    S. Littig and J. Voigt. ‘Porous medium equation and fast diffusion equation as gradient systems’. Czechoslovak Mathematical Journal65.4 (2015), pp. 869–889 (cit. on pp. 56, 57)

    work page 2015

  4. [1]

    D. R. Adams and L. I. Hedberg. Function spaces and potential theory. Vol. 314. Springer Science & Business Media, 1999 (cit. on p. 36)

    work page 1999

  5. [2]

    C. D. Aliprantis and K. Border.Infinite Dimensional Analysis: A Hitchhiker’s Guide. Springer, 2006 (cit. on pp. 22, 27)

    work page 2006

  6. [3]

    Ambrosio, N

    L. Ambrosio, N. Fusco and D. Pallara.Functions of bounded variation and free dis- continuity problems. Courier Corporation, 2000 (cit. on p. 59)

    work page 2000

  7. [5]

    Aujol, G

    J.-F. Aujol, G. Gilboa and N. Papadakis. ‘Theoretical analysis of flows estimating eigenfunctions of one-homogeneous functionals’.SIAM Journal on Imaging Sciences 11.2 (2018), pp. 1416–1440 (cit. on p. 5)

    work page 2018

  8. [6]

    Barthélemy

    L. Barthélemy. ‘Invariance d’un convexe fermé par un semi-groupe associé à une forme non-linéaire’. Abstract and Applied Analysis. Vol. 1. 3. Hindawi. 1996, pp. 237–262 (cit. on p. 39)

    work page 1996

Show all 48 references
  1. [7]

    H. H. Bauschke and P. L. Combettes.Convex Analysis and Monotone Operator Theory in Hilbert Spaces. Springer, 2011 (cit. on p. 21)

    2011

  2. [8]

    Benning and M

    M. Benning and M. Burger. ‘Ground states and singular vectors of convex variational regularization methods’.Methods and Applications of Analysis20.4 (2013), pp. 295– 334 (cit. on p. 5)

    2013

  3. [9]

    Benning, M

    M. Benning, M. Möller, R. Z. Nossek, M. Burger, D. Cremers, G. Gilboa and C.-B. Schönlieb. ‘Nonlinear spectral image fusion’.International Conference on Scale Space and Variational Methods in Computer Vision. Springer. 2017, pp. 41–53 (cit. on p. 5)

    2017

  4. [10]

    Bozorgnia, L

    F. Bozorgnia, L. Bungert and D. Tenbrinck. ‘The infinity Laplacian eigenvalue prob- lem: reformulation and a numerical scheme’.Journal of Scientific Computing 98.2 (2024), p. 40 (cit. on p. 5)

    2024

  5. [11]

    Bredies and M

    K. Bredies and M. Holler. A pointwise characterization of the subdifferential of the total variation functional. 2016. arXiv:1609.08918 (cit. on p. 32)

    2016 arXiv

  6. [12]

    H. Brézis. ‘Propriétés régularisantes de certains semi-groupes non linéaires’. Israel Journal of Mathematics9.4 (1971), pp. 513–534 (cit. on p. 39)

    1971

  7. [13]

    H. Brézis. Operateurs maximaux monotones et semi-groupes de contractions dans les espaces de Hilbert. Vol. 5. Elsevier, 1973 (cit. on pp. 38, 39). 67 68 BIBLIOGRAPHY

    1973

  8. [14]

    Briani, A

    A. Briani, A. Chambolle, M. Novaga and G. Orlandi. ‘On the gradient flow of a one- homogeneous functional’.Confluentes Mathematici 3.04 (2011), pp. 617–635 (cit. on p. 36)

    2011

  9. [15]

    Brokman, M

    J. Brokman, M. Burger and G. Gilboa. ‘Spectral total-variation processing of shapes – theory and applications’.ACM Transactions on Graphics43.2 (2024), pp. 1–20 (cit. on p. 5)

    2024

  10. [16]

    Brokman and G

    J. Brokman and G. Gilboa. ‘Nonlinear spectral processing of shapes via zero-homo- geneous flows’.International Conference on Scale Space and Variational Methods in Computer Vision. Springer. 2021, pp. 40–51 (cit. on p. 5)

    2021

  11. [18]

    Bühler and M

    T. Bühler and M. Hein. ‘Spectral clustering based on the graph p-Laplacian’.Proceed- ings of the 26th Annual International Conference on Machine Learning. 2009, pp. 81– 88 (cit. on p. 5)

    2009

  12. [19]

    Bungert and M

    L. Bungert and M. Burger. ‘Asymptotic profiles of nonlinear homogeneous evolution equations of gradient flow type’.Journal of Evolution Equations20.3 (2019), pp. 1061– 1092 (cit. on pp. 5, 26)

    2019

  13. [20]

    Bungert and M

    L. Bungert and M. Burger. ‘Gradient flows and nonlinear power methods for the computation of nonlinear eigenfunctions’.Handbook of Numerical Analysis. Vol. 23. Elsevier, 2022, pp. 427–465 (cit. on p. 6)

    2022

  14. [21]

    Bungert, M

    L. Bungert, M. Burger, A. Chambolle and M. Novaga. ‘Nonlinear spectral decomposi- tions by gradient flows of one-homogeneous functionals’.Analysis & PDE14.3 (2021), pp. 823–860 (cit. on p. 5)

    2021

  15. [22]

    Bungert, M

    L. Bungert, M. Burger and D. Tenbrinck. ‘Computing nonlinear eigenfunctions via gradient flow extinction’. International Conference on Scale Space and Variational Methods in Computer Vision. Springer. 2019, pp. 291–302 (cit. on p. 5)

    2019

  16. [23]

    Bungert and Y

    L. Bungert and Y. Korolev. ‘Eigenvalue problems inL∞: optimality conditions, du- ality, and relations with optimal transport’.Communications of the American Math- ematical Society2.08 (2022), pp. 345–373 (cit. on pp. 5, 6)

    2022

  17. [24]

    Bungert, Y

    L. Bungert, Y. Korolev and M. Burger. ‘Structural analysis of an L-infinity variational problem and relations to distance functions’.Pure and Applied Analysis2.3 (2020), pp. 703–738 (cit. on p. 5)

    2020

  18. [25]

    Burger, G

    M. Burger, G. Gilboa, M. Moeller, L. Eckardt and D. Cremers. ‘Spectral decompos- itions using one-homogeneous functionals’. SIAM Journal on Imaging Sciences 9.3 (2016), pp. 1374–1408 (cit. on p. 5)

    2016

  19. [26]

    G.-Q. Chen, W. P. Ziemer and M. Torres. ‘Gauss-Green theorem for weakly differenti- able vector fields, sets of finite perimeter, and balance laws’.Communications on Pure and Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences 62.2 (2009), p...

    2009

  20. [27]

    M. G. Crandall and P. Benilan. Regularizing effects of homogeneous evolution equa- tions. Tech. rep. Mathematics Research Center, University of Wisconsin–Madison, 1980 (cit. on p. 49). BIBLIOGRAPHY 69

    1980

  21. [28]

    Deidda, F

    P. Deidda, F. Tudisco and D. Zhang. Nonlinear spectral graph theory. 2025. arXiv: 2504.03566 (cit. on p. 5)

    2025 arXiv

  22. [29]

    Ekeland and R

    I. Ekeland and R. Témam. Convex Analysis and Variational Problems. 1976 (cit. on pp. 21–24)

    1976

  23. [30]

    Elmoataz, M

    A. Elmoataz, M. Toutain and D. Tenbrinck. ‘On the p-Laplacian and∞-Laplacian on graphs with applications in image and data processing’.SIAM Journal on Imaging Sciences 8.4 (2015), pp. 2412–2451 (cit. on p. 5)

    2015

  24. [31]

    Feld, J.-F

    T. Feld, J.-F. Aujol, G. Gilboa and N. Papadakis. ‘Rayleigh quotient minimization for absolutely one-homogeneous functionals’.Inverse Problems35.6 (2019), p. 064003 (cit. on p. 5)

    2019

  25. [32]

    Fumero, M

    M. Fumero, M. Möller and E. Rodolà. ‘Nonlinear spectral geometry processing via the TV transform’.ACM Transactions on Graphics (TOG)39.6 (2020), pp. 1–16 (cit. on p. 5)

    2020

  26. [33]

    G. Gilboa. ‘A spectral approach to total variation’.International Conference on Scale Space and Variational Methods in Computer Vision. Springer. 2013, pp. 36–47 (cit. on p. 5)

    2013

  27. [34]

    G. Gilboa. ‘A total variation spectral framework for scale and texture analysis’.SIAM Journal on Imaging Sciences7.4 (2014), pp. 1937–1961 (cit. on p. 5)

    2014

  28. [35]

    G. Gilboa. Nonlinear Eigenproblems in Image Processing and Computer Vision . Springer, 2018 (cit. on p. 5)

    2018

  29. [36]

    Gilboa, M

    G. Gilboa, M. Moeller and M. Burger. ‘Nonlinear spectral analysis via one- homogeneous functionals: overview and future prospects’. Journal of Mathematical Imaging and Vision56 (2016), pp. 300–319 (cit. on p. 5)

    2016

  30. [37]

    Kawohl and P

    B. Kawohl and P. Lindqvist. ‘Positive eigenfunctions for thep-Laplace operator revis- ited’. Analysis 26.4 (2006), pp. 545–550 (cit. on pp. 16, 17)

    2006

  31. [38]

    Kawohl and F

    B. Kawohl and F. Schuricht. ‘Dirichlet problems for the1-Laplace operator, including the eigenvalue problem’.Communications in Contemporary Mathematics9.04 (2007), pp. 515–543 (cit. on p. 59)

    2007

  32. [40]

    J. M. Mazón, M. Solera and J. Toledo. ‘The total variation flow in metric random walk spaces’. Calculus of Variations and Partial Differential Equations59 (2020), pp. 1–64 (cit. on p. 5)

    2020

  33. [41]

    Meyer-Nieberg

    P. Meyer-Nieberg. Banach Lattices. Springer, 1991 (cit. on p. 51)

    1991

  34. [42]

    Moeller, J

    M. Moeller, J. Diebold, G. Gilboa and D. Cremers. ‘Learning nonlinear spectral filters for color image reconstruction’.Proceedings of the IEEE International Conference on Computer Vision. 2015, pp. 289–297 (cit. on p. 5)

    2015

  35. [43]

    R. T. Rockafellar. ‘Generalised directional derivatives and subgradients of nonconvex functions’. Canadian Journal of Mathematics XXXII.2 (1980), pp. 257–280 (cit. on p. 30)

    1980

  36. [44]

    L. I. Rudin, S. Osher and E. Fatemi. ‘Nonlinear total variation based noise removal algorithms’. Physica D: Nonlinear Phenomena 60.1-4 (1992), pp. 259–268 (cit. on p. 5). 70 BIBLIOGRAPHY

    1992

  37. [45]

    G. Savaré. ‘Regularity results for elliptic equations in Lipschitz domains’.Journal of Functional Analysis152.1 (1998), pp. 176–201 (cit. on p. 15)

    1998

  38. [46]

    N. S. Trudinger. ‘On Harnack type inequalities and their application to quasilinear elliptic equations’. Communications on Pure and Applied Mathematics20.4 (1967), pp. 721–747 (cit. on p. 16)

    1967

  39. [47]

    Tudisco and D

    F. Tudisco and D. Zhang.Nonlinear spectral duality. 2022. arXiv:2209.06241 (cit. on p. 5)

    2022 arXiv

  40. [48]

    Varvaruca

    E. Varvaruca. ‘Exact rates of convergence ast → ∞for solutions of nonlinear evolution equations’. Journal of Evolution Equations4.4 (2004), pp. 543–565 (cit. on p. 5)

    2004

Pith tools

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