Convex analysis in normed spaces and metric projections onto convex bodies

Horst Martini; Ralph Teixeira; Vitor Balestro

REVIEW 2 major objections 4 minor 34 references

Convex analysis in normed spaces and metric projections onto convex bodies

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read With a smooth, strictly convex norm, distance to a convex body is differentiable outside the body.

desk verdict The main theorem is sound and the norm-gradient framework is genuinely useful; the proof has a fixable slip in Proposition 3.5 and needs a comparison with known distance-function subdifferential results. read the letter →

arxiv 1908.08742 v1 pith:HQAXULNR submitted 2019-08-23 math.MG math.FA

Vitor Balestro , Horst Martini , Ralph Teixeira This is my paper · ORCID

classification math.MGmath.FA MSC 41A5041A6546B2046G0552A2052A2153C2358C20

keywords BirkhofforthogonalityconvexbodydistancefunctionsdifferentiabilityLegendretransformsub-gradientmetricprojectionnormedspaces

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 establishes that in a finite-dimensional normed space, if the norm's unit ball is smooth and strictly convex, then the distance function to any convex body is differentiable at every point outside the body, and its gradient is the unit vector pointing from the nearest point in the body toward the point. This answers a question the authors say was open: regularity of the distance function follows from the geometry of the ambient norm alone, with no regularity assumed for the body. The route is a new convex-analysis toolkit for normed spaces: instead of identifying vectors with linear functionals via an inner product, the paper uses the Legendre transform built from the norm, and develops norm versions of gradients and sub-gradients. These tools also give an explicit characterization of the distance function's sub-differential on the boundary and a clean link between metric projections and the norm's own orthogonality relation. A reader should care because distance functions and metric projections are basic objects in approximation, optimization, and geometric analysis, and the paper pins down exactly when their classical Euclidean regularity survives.

What carries the argument

The engine is the Legendre transform of the norm: for each nonzero vector $x$ it is the unique linear functional whose kernel is the hyperplane Birkhoff right-orthogonal to $x$ (the supporting hyperplane of the unit ball in direction $x$), normalized by $L(x)\cdot x=\|x\|^2$. This map is a norm-preserving bijection between the space and its dual, and it replaces the standard inner product in every convex-analysis step. On top of it the paper defines the norm gradient and norm sub-gradient of a convex function via the inequality $f(y)-f(x) \ge L(\nabla f(x))\cdot(y-x)$, and proves the norm analogues of the classical facts: the sub-differential is nonempty, the one-sided derivative is the maximum of $L(w)\cdot u$ over sub-gradients $w$, and a convex function is differentiable exactly when its norm sub-differential is a singleton. That last criterion, applied to the distance function with its parallel-body sub-level sets, is what produces the main theorem.

What would settle it

Take the plane with an $\ell^p$ norm for $1<p<\infty$, which is smooth and strictly convex, and let $K$ be a square. For a point $x$ whose metric projection is a corner $v$, the theorem predicts that $d_K$ is differentiable at $x$ and that every one-sided directional derivative equals $L((x-v)/\|x-v\|)\cdot u$. Computing these derivatives directly for $u$ along the two edge directions from the corner, checking that the derivatives in opposite directions are negatives, would settle the claim; any mismatch is a counterexample.

Watch

Extended reading notes

Core claim

The central discovery is Theorem 5.1: for a convex body $K$ in $(\mathbb{R}^n, \|\cdot\|)$ with a smooth and strictly convex norm, the distance function $d_K(x)=\operatorname{dist}(x,K)$ is differentiable on $\mathbb{R}^n\setminus K$ and its norm gradient is $\nabla d_K(x)=\eta_K(x)=(x-p_K(x))/\|x-p_K(x)\|$, the normalized vector from the metric projection. Consequently $d_K$ is of class $C^1$ outside $K$, because both the distance function and the metric projection are continuous. On the boundary the picture is different: $d_K$ is not differentiable, and its norm sub-differential equals the Birkhoff normal cone of $K$ at the point, the cone of outward vectors that are left-orthogonal (in the norm's sense) to a supporting hyperplane. The proof identifies the sub-level sets $\{d_K \le c\}$ with the parallel bodies $K+cB$, uses smoothness of the unit ball to get a unique supporting hyperplane there, and shows the norm sub-differential is a singleton; a general characterization from the paper then converts that singleton into differentiability.

Load-bearing premise

The load-bearing premise is that the ambient norm's unit ball is smooth and strictly convex: every boundary point has exactly one supporting hyperplane and the boundary contains no line segment, which together make the metric projection unique and the Legendre transform a bijection.

Editorial extensions

If this is right

In any finite-dimensional normed space whose unit ball is smooth and strictly convex, the distance function to every convex body is $C^1$ outside the body; no smoothness or strict convexity of the body itself is needed.
The metric projection, when it is unique (guaranteed by strict convexity), is continuous, and the formula $\nabla d_K=\eta_K$ makes the gradient an explicit, computable function of the nearest point.
At boundary points, the distance function is never differentiable, and its norm sub-differential is exactly the Birkhoff normal cone of the body.
The same Legendre-transform machinery shows that a smooth norm is itself $C^1$ away from the origin, with norm gradient $x/\|x\|$, recovering a known regularity fact from the new tools.
The norm subgradient calculus reproduces the standard Euclidean subdifferential results (max formula, cyclic monotonicity, convexity detection) with the norm's Legendre transform in place of the inner product.

Reading between the lines

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

The paper does not pursue infinite-dimensional versions; a natural next step would be to ask whether reflexivity plus smooth strict convexity suffices in Banach spaces, though the compactness arguments used here would need replacement.
Since $\nabla d_K$ is the unit vector pointing from the nearest point to $x$, the identity suggests that metric projections can be recovered by integrating the gradient flow of the distance function, which could give practical algorithms for nearest-point queries under arbitrary norms.
The geometric definition of the Legendre transform, which avoids assuming differentiability of the norm, connects directly to semi-inner products; this may give a coordinate-free gradient calculus in Finsler geometry.
A testable boundary of the result: allow the body to be nonconvex while keeping the norm smooth and strictly convex; the proof leans on convexity at nearly every step, so a counterexample there would show the phenomenon is really about convexity, not just the norm.

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.

Desk Editor's Note

read the letter

Dear [Colleague],

Quick take: Theorem 5.1 is true, and the norm-gradient/subgradient framework is a useful way to organize convex analysis in Minkowski spaces. The printed proof has a fixable translation slip in Proposition 3.5; don't let that drive the verdict.

What is actually here: the paper builds convex analysis on a smooth strictly convex norm using the Legendre transform of the norm. That gives norm gradients, norm subgradients, and a clean link to Birkhoff orthogonality. The payoff is the differentiability theorem: for a convex body K in a smooth strictly convex normed space, the distance function d_K is differentiable on R^n\K and its norm gradient is the normalized metric projection direction. The proof is coherent and the C^1 corollary follows. I also liked Theorem 4.6 and Corollary 4.4, which recover differentiability of the norm and identify the Legendre transform with the norm derivative; those are clean results.

Where it needs work: Proposition 3.5, as written, is wrong. The proof translates the supporting hyperplane to pass through x and then says h+δu supports K+δB at x. That is a translation error: if h has been translated to x, h+δu doesn't pass through x; if h is the original support at z, then h+δu is the right plane and the containment can be proved in one line with the Legendre functional. So the proposition is true, but the proof needs a corrected write-up. The smoothness of K+δB is cited from [18] rather than proved; if the citation is solid that is acceptable, but I'd ask for a precise statement. Proposition 3.2 states an iff but only proves the forward direction. It is unused in the strict-convexity regime, so this is a minor completeness issue. Proposition 4.2's "Birkhoff orthogonality is continuous" is actually immediate from continuity of the norm, so I don't count that as a gap.

The bigger referee request is novelty support. The authors claim the differentiability question under smooth plus strictly convex norm was not explicitly answered, but they don't compare with the known subdifferential characterizations of distance functions (Fitzpatrick, Giles, Zajicek). It may well be that the norm-gradient apparatus is the new part and Theorem 5.1 is a corollary of known subdifferential formulas; either way, a few paragraphs of comparison would settle it.

Who it is for: convex analysts and Minkowski geometers who want a norm-adapted calculus; they will get a working toolkit and a clean theorem. It deserves serious refereeing, not a desk reject. I'd accept and ask for the Prop 3.5 fix and the comparisons.

Referee Report

2 major / 4 minor

Summary. The paper develops a convex-analytic framework for finite-dimensional normed spaces in which the usual identification of the space with its dual is replaced by the Legendre transform associated with the norm. On this basis it defines norm gradients and norm sub-gradients, relates them to Birkhoff orthogonality, and proves a main theorem (Theorem 5.1): if the ambient norm is smooth and strictly convex and K is a convex body, then the distance function d_K is differentiable on R^n\K, and its norm gradient is the normalized outer normal direction from the metric projection, ∇d_K(x)=η_K(x). The paper also proves auxiliary results on uniqueness and continuity of metric projections, continuity and self-duality of the Legendre transform, a norm-subgradient characterization of differentiability, and a description of the boundary subdifferential of d_K as the Birkhoff normal cone.

Significance. If the proof is completed, Theorem 5.1 settles a natural question: under only smoothness and strict convexity of the norm, the distance to an arbitrary convex body is differentiable outside the body, with an explicit formula for the gradient. The Legendre-transform framework is elegant and gives a genuinely norm-dependent replacement for classical Euclidean gradient and subgradient theory, with connections to Birkhoff orthogonality that appear to be new. The paper is largely self-contained, uses no fitted parameters or data, and makes a concrete, falsifiable prediction about the norm gradient. Several auxiliary results, such as the self-duality of the Legendre transform and the norm-cyclic-monotonicity version of Rockafellar's theorem, are of independent interest. The main caveat is that the proof of the central theorem currently depends on two geometric facts that are not adequately established in the manuscript.

major comments (2)

[§3, Proposition 3.5, and its use in §5, Theorem 5.1] The proof of Proposition 3.5 is not valid as written. After the sentence 'we assume that h is translated to pass through x', the subsequent containment K+δB ⊆ (h+δu_K(z))^- treats h as the supporting hyperplane at z rather than at x; with h translated through x, the hyperplane h+δu_K(z) does not pass through x=z+δu_K(z) and therefore cannot be the supporting hyperplane asserted. This is not a cosmetic issue, because Theorem 5.1 uses Proposition 3.5 to conclude that an arbitrary norm subgradient v is a multiple of η_K(x). The statement is repairable: take h to be the supporting hyperplane of K at z with u_K(z) ⊣_B h, and let φ=L(u_K(z)); then for y∈K and b∈B one has φ(y)+δφ(b) ≤ δ, showing that the translate h+δu_K(z) supports K+δB at x. The manuscript should be corrected accordingly.
[§3, parallel sets, and §5, Theorem 5.1] Theorem 5.1 relies on the uniqueness of the supporting hyperplane of K+cB at x: if K+cB had several supporting hyperplanes at x, Theorem 4.4 would only give that a subgradient is orthogonal to one of them, and the conclusion ∂d_K(x)⊆span{η_K(x)} would not follow. The manuscript asserts smoothness of K+δB with only a citation to [18] and gives no proof or precise statement. Since this is load-bearing, please add a self-contained proof, for instance using normal cones: for a boundary point x=a+b of K+B with b∈∂B, the normal cone N_{K+B}(x) is contained in N_B(b), and smoothness of B makes N_B(b) a ray, so uniqueness of the supporting hyperplane follows. Alternatively, quote the exact theorem from [18].

minor comments (4)

[§3, Proposition 3.2] The proposition states an 'if and only if', but the proof establishes only the direction 'non-unique metric projection implies parallel boundary segments'; the converse is asserted without proof. The converse is not used in the proof of Theorem 5.1, but the statement should either be proved in full or explicitly weakened.
[§4, Corollary 4.2] In the displayed formula for f'_+(x,-u), the symbol L(w) should be L(v); the variable w is undefined in that display.
[§4, Proposition 4.2] The proof uses the assertion that Birkhoff orthogonality is a continuous relation without proof. This follows in one line from the definition by passing to the limit in ||x_n+t z_n|| ≥ ||x_n||, and the manuscript should include that observation.
[Throughout Sections 3–5] The notation h is used sometimes for a hyperplane through the origin and sometimes for a translate (through z, through x, or shifted by δu). Distinguishing the linear subspace from its affine translates would prevent the kind of confusion that occurs in the proof of Proposition 3.5.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: distance-function differentiability is derived from norm convex analysis; cited works are background and flagged gaps are rigor issues, not circular reductions.

full rationale

The derivation chain is self-contained. The norm-gradient and norm-subgradient formalism is introduced by definitions (Definition 4.1 and inequality (4.5)) that do not presuppose the differentiability of distance functions, and Corollary 4.2 characterizes differentiability by a singleton subdifferential using Lemma 4.2 and Theorem 4.2. Theorem 5.1 then combines metric-projection uniqueness from strict convexity (Corollary 3.1), the Birkhoff-orthogonality properties of projections (Theorem 3.1, Propositions 3.3 and 3.5), and the subdifferential machinery to conclude ∂d_K(x) = {η_K(x)}. No equation in the proof is equivalent to the conclusion by construction, and no fitted or data-dependent parameter is introduced. The self-citations present, such as the survey [2] and [20], are background references for standard facts about Birkhoff orthogonality and strict convexity; they do not carry the load of Theorem 5.1. The skeptical concerns about Proposition 3.5's proof as written and the cited smoothness of parallel bodies ([18]) are correctness or rigor gaps, not circularity: even if those arguments need repair, their statements are not inputs equivalent to the target theorem. Hence the circularity score is 0.

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

The central claim rests on standard convex analysis and on the geometric assumption of a smooth strictly convex norm. There are no fitted parameters. The norm gradient and norm subgradient are definitions, not postulated entities. The proof of Theorem 5.1 uses uniqueness of the metric projection, smoothness of parallel bodies, and the subdifferential characterization of differentiability; all of these depend on the norm assumption. The proof of Proposition 4.2 also quietly assumes continuity of Birkhoff orthogonality, which is plausible but not demonstrated in the text.

assumptions (4)

domain assumption The norm ||·|| on R^n is smooth and strictly convex, so its unit ball has a unique supporting hyperplane at each boundary point and its boundary contains no line segment.
Adopted at the start of Section 4 and used in the proofs of Theorems 4.1 through 5.2; it ensures the Legendre transform is a bijection, metric projections are unique, and parallel bodies K+δB are smooth.
standard math Standard convex analysis facts from Schneider [28, Chapter 1] are taken as background.
Several proofs in Sections 2 and 3 reference [28] rather than reproducing the arguments, including Proposition 2.1, Lemma 2.1, and the separation theorem used in Theorem 3.1.
domain assumption Birkhoff orthogonality is continuous in the sense used in the proof of Proposition 4.2.
In the first step of the proof of Proposition 4.2, the authors assert that Birkhoff orthogonality is a continuous relation and use it to conclude α_n → α; no proof or citation is given at that point.
standard math The hyperplane separation theorem for disjoint convex bodies.
Used in Theorem 3.1 to separate the interiors of K and B(x,ρ) and to obtain the supporting hyperplane at p_K(x).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Convex analysis in normed spaces and metric projections onto convex bodies." pith.science (2026). https://pith.science/paper/HQAXULNR

@misc{pith2026190808742,
  author       = {Pith},
  title        = {Pith review of: Convex analysis in normed spaces and metric projections onto convex bodies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HQAXULNR}},
  note         = {Machine review of arXiv:1908.08742}
}

read the original abstract

We investigate metric projections and distance functions referring to convex bodies in finite-dimensional normed spaces. For this purpose we identify the vector space with its dual space by using, instead of the usual identification via the standard inner product, the Legendre transform associated with the given norm. This approach yields re-interpretations of various properties of convex functions, and new relations between such functions and geometric properties of the studied norm are also derived.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 34 canonical work pages

[18]

S. G. Krantz and H. R. Parks, On the vector sum of two convex sets in space. Canad. J. Math. 43(2), pp. 347–355, 1991

work page 1991
[1]

A. R. Alimov, I. G. Tsar’kov, Connectedness and other geometr ic properties of suns and Chebyshev sets (Russian). Fundam. Prikl. Mat. 19(4), pp. 21–91, 2014; translation in J. Math. Sci. (N.Y.) 217(6), pp. 683–730, 2016

work page 2014
[2]

Alonso, H

J. Alonso, H. Martini, S. Wu, On Birkhoﬀ orthogonality and isoscele s orthogonality in normed linear spaces. Aequationes Math. 83, 153-189, 2012

work page 2012
[3]

J. C. ´Alvarez-Paiva and A. C. Thompson, Volumes in normed and Finsler spa ces, in: A Sampler of Riemann-Finsler Geometry (eds.: D. Bao, R. Bryant, S. S. Chern, and Z. Shen), pp. 1–49. Cambridge University Press, Cambridge, 2004

work page 2004
[4]

Asplund, Diﬀerentiability of the metric projection in ﬁnite-dimen sional Euclidean space

E. Asplund, Diﬀerentiability of the metric projection in ﬁnite-dimen sional Euclidean space. Proc. Amer. Math. Soc. 38, pp. 218–219, 1973

work page 1973
[5]

Borwein, S

M. Borwein, S. P. Fitzpatrick and J. R. Giles, The diﬀerentiability of real functions on normed linear space using generalized subgradients. J. Math. Anal. Appl. 128(2), pp. 512–534, 1987

work page 1987
[6]

Borwein and A

M. Borwein and A. S. Lewis, Convex Analysis and Nonlinear Optimization: Theory and Examples , Second Edition. CMS Books in Mathematics 3, Canadian Mathematical Society, 2006

work page 2006
[7]

A. L. Brown, Suns in normed linear spaces which are ﬁnite-dimensio nal. Math. Ann. 279(1), pp. 87–101, 1987

work page 1987

Show all 34 references

[8]

Bunt, Bijdrage tot de theorie der konvekse puntverzameling en

L. Bunt, Bijdrage tot de theorie der konvekse puntverzameling en. Thesis, University of Groningen, Amsterdam, 1934

1934
[9]

F. H. Clarke, Optimization and Nonsmooth Analysis . Classics in Applied Mathematics 5, Society for Industrial and Applied Mathematics (SIAM), Philadelph ia, PA, 1990

1990
[10]

Correa, D

R. Correa, D. Salas and L. Thibault, Smoothness of the metric p rojection onto nonconvex bodies in Hilbert spaces. J. Math. Anal. Appl. 457(2), pp. 1307–1332, 2018. 26

2018
[11]

Deutsch, The convexity of Chebyshev sets in Hilbert space, in Topics in Polynomials of One and Several Variables and Applications

F. Deutsch, The convexity of Chebyshev sets in Hilbert space, in Topics in Polynomials of One and Several Variables and Applications . World Scientiﬁc Publishing Co., Inc., NJ, pp. 143–150, 1993

1993
[12]

Fitzpatrick, Metric projections and the diﬀerentiability of dis tance functions

S. Fitzpatrick, Metric projections and the diﬀerentiability of dis tance functions. Bull. Aust. Math. Soc. 22, pp. 291–312, 1980

1980
[13]

J. R. Giles, A distance function property implying diﬀerentiability. Bull. Aust. Math. Soc. 39(1), pp. 59–70, 1989

1989
[14]

J. R. Giles, Diﬀerentiability of distance functions and a proximinal property inducing convexity. Proc. Amer. Math. Soc. 104(2), pp. 458–464, 1988

1988
[15]

Hetzelt, On suns and cosuns in ﬁnite-dimensional normed rea l vector spaces

L. Hetzelt, On suns and cosuns in ﬁnite-dimensional normed rea l vector spaces. Acta Math. Hungar. 45(1–2), pp. 53–68, 1985

1985
[16]

´A. G. Horv´ ath, Z. L´ angi and M. Spirova, Semi-inner products and the concept of semi- polarity. Results Math. 71(1–2), pp. 127–144, 2017

2017
[17]

L. A. Karlovitz, The construction and application of contractiv e retractions in 2- dimensional normed linear spaces. Indiana Univ. Math. J. 22(5), pp. 473–481, 1972

1972
[19]

Li and L

Y. Li and L. Nirenberg, Regularity of the distance function to t he boundary. Rend. Accad. Naz. Sci. XL Mem. Mat. Appl. (5) 29(1), pp. 257–264, 2005

2005
[20]

Martini, K

H. Martini, K. J. Swanepoel and G. Weiss, The geometry of Minko wski spaces – a survey. Part I. Expositiones Math. 19, pp. 97–142, 2001

2001
[21]

B. S. Mordukhovich and N. M. Nam, Convex Analysis with Applications to Optimization and Location Problems . Springer, to appear
[22]

Noll, Directional diﬀerentiability of the metric projection in Hilbe rt space

D. Noll, Directional diﬀerentiability of the metric projection in Hilbe rt space. Paciﬁc J. Math. 170(2), pp. 567–592, 1995

1995
[23]

J. P. Penot and R. Ratsimahalo, Characterizations of metric pr ojections in Banach spaces and applications. Abstr. Appl. Anal. 3(1–2), pp. 85–103, 1998

1998
[24]

R. A. Poliquin, R. T. Rockafellar and L. Thibault, Local diﬀerentia bility of distance functions. Trans. Amer. Math. Soc. 352(11), pp. 5231–5249, 2000

2000
[25]

D. Sain, K. Paul and A. Mal, On approximate Birkhoﬀ-James orth ogonality and normal cones in a normed space. J. Convex. Anal. 26(1), pp. 341–351, 2019

2019
[26]

Sangeeta and T. D. Narang, On suns in linear metric spaces and c onvex metric spaces. East J. Approx. 17(2), pp. 127–139, 2011

2011
[27]

Safdari, The distance function from the boundary of a doma in with corners

M. Safdari, The distance function from the boundary of a doma in with corners. Nonlinear Anal. 181, pp. 294–310, 2019

2019
[28]

Schneider, Convex Bodies: The Brunn-Minkowski Theory

R. Schneider, Convex Bodies: The Brunn-Minkowski Theory . Encyclopedia of Mathe- matics and its Applications 151, Cambridge University Press, Cambridge, 2014

2014
[29]

A. C. Thompson, Minkowski Geometry. Encyclopedia of Mathematics and Its Applica- tions 63, Cambridge University Press, Cambridge, 1996. 27

1996
[30]

L. P. Vlasov, Approximatively convex sets in uniformly smooth sp aces. Mat. Zametki 1, pp. 443–450, 1967

1967
[31]

L. P. Vlasov, On almost convex sets in Banach spaces (Russian) . Dokl. Akad. Nauk SSSR 163, pp. 18–21, 1965

1965
[32]

Zaj ´ ıˇ cek, Diﬀerentiability of the distance function and points of multivaluedness of the metric projection in Banach space

L. Zaj ´ ıˇ cek, Diﬀerentiability of the distance function and points of multivaluedness of the metric projection in Banach space. Czechoslovak Math. J. 33(2), pp. 292–308, 1983

1983
[33]

Zaj ´ ıˇ cek, On the Fr´ echet diﬀerentiability of distance functions

L. Zaj ´ ıˇ cek, On the Fr´ echet diﬀerentiability of distance functions. Proceedings of the 12th winter school on abstract analysis (Srn ´ ı, 1984).Rend. Circ. Mat. Palermo 2, Suppl. No. 5, pp. 161–165, 1984

1984
[34]

Wu, A Chebyshev set and its distance function

Z. Wu, A Chebyshev set and its distance function. J. Approx. Theory 119(2), pp. 181– 192, 2002. 28

2002

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [18]

S. G. Krantz and H. R. Parks, On the vector sum of two convex sets in space. Canad. J. Math. 43(2), pp. 347–355, 1991

work page 1991

[2] [1]

A. R. Alimov, I. G. Tsar’kov, Connectedness and other geometr ic properties of suns and Chebyshev sets (Russian). Fundam. Prikl. Mat. 19(4), pp. 21–91, 2014; translation in J. Math. Sci. (N.Y.) 217(6), pp. 683–730, 2016

work page 2014

[3] [2]

Alonso, H

J. Alonso, H. Martini, S. Wu, On Birkhoﬀ orthogonality and isoscele s orthogonality in normed linear spaces. Aequationes Math. 83, 153-189, 2012

work page 2012

[4] [3]

J. C. ´Alvarez-Paiva and A. C. Thompson, Volumes in normed and Finsler spa ces, in: A Sampler of Riemann-Finsler Geometry (eds.: D. Bao, R. Bryant, S. S. Chern, and Z. Shen), pp. 1–49. Cambridge University Press, Cambridge, 2004

work page 2004

[5] [4]

Asplund, Diﬀerentiability of the metric projection in ﬁnite-dimen sional Euclidean space

E. Asplund, Diﬀerentiability of the metric projection in ﬁnite-dimen sional Euclidean space. Proc. Amer. Math. Soc. 38, pp. 218–219, 1973

work page 1973

[6] [5]

Borwein, S

M. Borwein, S. P. Fitzpatrick and J. R. Giles, The diﬀerentiability of real functions on normed linear space using generalized subgradients. J. Math. Anal. Appl. 128(2), pp. 512–534, 1987

work page 1987

[7] [6]

Borwein and A

M. Borwein and A. S. Lewis, Convex Analysis and Nonlinear Optimization: Theory and Examples , Second Edition. CMS Books in Mathematics 3, Canadian Mathematical Society, 2006

work page 2006

[8] [7]

A. L. Brown, Suns in normed linear spaces which are ﬁnite-dimensio nal. Math. Ann. 279(1), pp. 87–101, 1987

work page 1987