Structured coordinate systems in L p(R)
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.
Ada: We're looking at the paper: "Structured coordinate systems in L p(R)".
Emmy: Here's the gist of the paper. The paper constructs Schauder bases of L p(R) under specific structural constraints, namely that elements are non-negative functions or translates of a single function.
Ada: The construction proceeds in several stages depending on the value of p.
Emmy: For every 1 at most p < infinity, non-negative Schauder bases are constructed.
Ada: Duality allows for the construction of bases with positive biorthogonal functionals, maintaining the same bounds for all 1 at most p < infinity.
Paper summary: Ada: So we're looking at this paper called "Structured coordinate systems in L p(R)," and what it claims is they can build Schauder bases for L p of the real numbers under these specific structural rules. It’s about functions that are either non-negative or just translates of a single function.
Emmy: Right, Ada, so the main thesis seems to be that they construct these bases for every p between one and infinity. They can even get them to have basis constants arbitrarily close to one when the elements are non-negative >
Kolya: And they use duality to get bases with positive biorthogonal functionals which maintain those same bounds across all p values, so it’s not just a thing for one specific exponent >
Ada: Exactly. The paper says they build non-negative Schauder bases for every p between one and infinity, even though these bases aren't unconditional > It says their basis constants can be made arbitrarily close to one >
Emmy: That’s interesting because usually, when you constrain things like that, the constant shoots up really fast. But they manage to keep it near one > And they also get bases with positive biorthogonal functionals that match those bounds across all p values >
Kolya: So for the listener who just cares about how well we can represent functions in a certain space, this means we have a way to find systems where the representation is very efficient, even when you're dealing with non-negative functions > But I wonder if these bases are useful outside of this theoretical setting >
Ada: That’s the next thing to consider. The paper also tackles different p values separately. For p equals two, they found a basis of exponentials where the basis constant is less than one plus epsilon for any epsilon you pick > And they said frequencies can be rational and separated quite far apart outside the initial part of the sequence >
Emmy: That's a bit specific, Ada. It sounds like for p equals two, we have some sort of nice structure involving Fourier transforms and these rational frequencies that stay well-behaved > But how do they handle the larger values of p, say when p is greater than two >
Paper summary: Kolya: For those larger values, the construction uses a generator that extends to an entire function. They choose the translation parameters from any set you prescribe that is unbounded > And they manage to keep derivative estimates controlled with a factor of two in the increments >
Ada: That sounds like a bit of a technical lift for p greater than two. The paper says the resulting basis constant is at most K times one plus rho over one minus rho times two > Which involves approximating some function using translated copies of a reference function >
Emmy: So for these higher p values, they are relying on this generator that’s a limit of entire functions and these specific translation choices to keep the structure going > It sounds like the complexity increases as p gets larger there >
Kolya: And for p equals one, they use a lemma called Lemma six point two to construct a non-negative basis where the partial sum projections satisfy bounds related to some theta value > They also achieve positive biorthogonal functionals by running the construction with p equals infinity and q equals one >
Ada: So for p equals one, the construction seems to rely on these specific projection bounds and then they use a technique involving duality from the p infinity case to get those positive functionals > The coefficient functions turn out to be non-negative step functions that are compactly supported >
Emmy: That sounds like a very concrete description of what those coefficient functions look like, which is helpful for visualizing the structure behind it > But the paper also proves some strict lower bounds on these bases, which suggests there are limits to how small the basis constants can actually get >
Kolya: That’s important because if you're looking for a basis where the constant is as close to one as possible, Proposition seven point one says for p between one and infinity, the first coordinate projection of a non-negative basis cannot be contractive > This means you can't get arbitrarily close to one in that range >
Ada: So the infimum of these basis constants is actually one for every p between one and infinity, but you can't reach it in that range > And for p equals one, the paper points to work by Dor and Odell showing that this nonattainment follows from the description of monotone bases in L one spaces >
Paper summary: Emmy: It seems like the overall message is that while we can construct these highly structured bases, there are fundamental limits on how good they can be when p is small, specifically near one > What this means for us as listeners is that if you're trying to build a system where the representation efficiency is maximized under these structural rules, you have to accept some gap between what’s possible and what’s achievable >
Kolya: So we have these constructions for every p between one and infinity, but they all end up having this lower bound of one that you can't touch in that range > It suggests that the structure itself imposes a kind of unavoidable cost on the efficiency of the basis representation > We haven't seen this type of universal lower bound before in this context >
Ada: Exactly. And for p greater than two, they show a way to select translation parameters from any unbounded set, which is quite powerful for defining coordinate systems > The basis constant there is bounded by K times one plus rho over one minus rho times two >
Emmy: So the key difference between the p values seems to be whether we can control those translation parameters freely or if we are stuck with more constrained choices when p is small > It’s about whether you have freedom in choosing your basis vectors or if you are just accepting a fixed pattern >
Kolya: For someone who only listens to this, it means that depending on the space you're working in, say L two versus L one versus some higher p space, the mathematical machinery we use to build these bases changes fundamentally > It’s not just about finding *a* basis; it’s about finding a basis that fits your structural needs >
Ada: So to wrap up on the paper "Structured coordinate systems in L p(R)," it confirms we can construct these specific types of bases for every p between one and infinity, and we know where the absolute minimum efficiency lies across all those spaces > It’s about understanding what structural constraints force certain limitations on our coordinate systems.
Conclusion: Ada: So we've been looking at this paper called "Structured coordinate systems in L p(R)," and what it boils down to is they’re building specific types of bases for functions that are either non-negative or just translates of one function, across every possible value of p from one up to infinity.
Emmy: It seems like the authors are really focused on how those structural constraints—like only using non-negative functions—force a certain kind of mathematical architecture when you're trying to create these coordinate systems for L p spaces.
Kolya: What I’m hearing is that even with these restrictions, they can still maintain bases where the constant, the number that tells you how well the basis works, stays very close to one across all those different p values.
Ada: That's what they claim. They show constructions for every case—for p less than two, for p equal to two with exponentials, and then for all p greater than two using these generators that extend to entire functions.
Emmy: The real implication is that the choice of structure you impose on the functions themselves dictates exactly what kind of basis system you can build in a space. It’s not just about finding *any* basis; it's about finding one that respects those underlying properties.
Kolya: And I’m interested in those lower bounds they prove for non-negative bases. They show that for p between one and infinity, the first projection of such a basis can't be contractive. That suggests there’s a fundamental limit to how efficient these representations can get when you start with non-negative functions.
Ada: Exactly, Kolya. The infimum of those basis constants is actually one for every p in that range, but you just can't hit it. It means the structure itself imposes a kind of unavoidable cost on representation efficiency in that regime.
Emmy: So, when you think about this paper, don’t just see it as a construction paper for bases; see it as a study of how structural properties—non-negativity, or being translates—limit the performance characteristics of coordinate systems in L p spaces.
Kolya: It sets up some really interesting questions about where these limits lie when you move into those higher p values, especially with those translation parameters they can pick from any unbounded set. That’s where I think the next piece of this puzzle needs to be explored.
math.FA
Submitted: 2026-10-13
Updated: 2026-11-12
Code: https://github.com/lukasliehr/Structured-Coordinate-Systems
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 83/100
The gist: The paper constructs Schauder bases of L p(R) under specific structural constraints, namely that elements are non-negative functions or translates of a single function.
Key concepts
- Schauder basis
- A Schauder basis is a set of functions in a function space (like $L^p(\mathbb{R})$) that can be used to represent any function in that space as an infinite linear combination of these basis elements. The paper focuses on finding such bases with specific structural properties, like non-negativity.
- Basis constant
- The basis constant measures how efficiently the basis can approximate functions in the space. A smaller basis constant means the set of functions is more efficient at representing all possible functions in $L^p(\mathbb{R})$. The paper proves that for all $1 \le p < \infty$, this constant can be made arbitrarily close to one.
- Non-negative basis
- This refers to a Schauder basis where every element in the set is a non-negative function (i.e., its values are always greater than or equal to zero). The construction methods ensure that these bases maintain this non-negativity while controlling the efficiency of the approximation.
- Positive biorthogonal functionals
- This concept relates to finding dual elements for the basis functions. A positive biorthogonal functional is a linear functional that maps a function in $L^p$ to a scalar, and its corresponding dual element has positive values. The paper shows how to construct bases that possess these properties, particularly in $L^1$ by relating it to constructions in $L^\infty$.
Terminology
Summary
The paper constructs Schauder bases of L p(R) under specific structural constraints, namely that elements are non-negative functions or translates of a single function. p. 0 We construct Schauder bases of L
p
(R) with significant structural
constraints on their elements. First, we construct non-negative Schauder bases
of L
p
(R) for every 1 ≤ p < ∞. , φN
therefore provide the additional functions needed to span the refined step-function
space. Each takes the value −θ on part of its support, so we must modify the
construction to obtain non-negative basis functions.Where the paper says it
Where the paper says it
For every 1 at most p < infinity, non-negative Schauder bases are constructed. p. 1 Indeed, the proof for L
1
(R) made crucial use of the fact that non-negative bumps
translated to infinity do not converge weakly to zero, whereas the proof for L
2
(R)
crucially used that every closed subspace of a Hilbert space is 1-complemented. Our first main theorem constructs a non-negative Schauder basis for L
p
(R) for
every 1 ≤ p < ∞. Although such bases cannot be unconditional,
their basis constants can be made arbitrarily close to one. By duality, we also
obtain bases with positive biorthogonal functionals and the same bounds for all
1 ≤ p < ∞. Outline of the non-negative construction
We construct the basis inductively by adding finitely many functions at a time,
starting with f1 = 1[0,1). At each stage, the functions already constructed form a
non-negative basis of a finite-dimensional space of step functions. However, even for p > 2, the whole lattice aZ + b cannot be ordered
as a Schauder basis of translates [24, Proposition 3. When p = 2, our construction allows for rational parameters whose pairwise
separations tend to infinity; see Section 17.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
For p > 2, the construction utilizes a generator that is a limit of entire functions, with translation parameters chosen from any prescribed unbounded subset S of R. p. 0 The generating function f
can be chosen to be the restriction to R of an entire function. For p > 2, the
translation parameters can be chosen from any prescribed unbounded subset of
R, a conclusion that fails for p = 2. The Schauder basis of translates (Tλn f) in Theorem 1. 3 can be
chosen with the following additional regularity and structural properties:
(i) If p = 2 then for every ε > 0 one can choose f which extends to an entire
function and (λn) so that the basis constant satisfies K((Tλn f)) < 1 + ε. Similarly, Gabor systems are built from time-frequency
shifts [3, 19] of a single function. In both cases, the rigid structure of the basis
elements allows one to study properties of the whole system via its generator.Where the paper says it
Where the paper says it
Where the paper says it
For p=1, a non-negative basis with basis constant at most 1+ epsilon is constructed using Lemma 6. p. 1 Thus, both non-negative bases and bases of translates are necessarily
conditional. In 2015, Johnson and Schechtman [12] constructed a non-negative Schauder basis
of L
1
(R). We
call the γn the coefficient functions. For a basis B = (fn), we define its partial sum
projections and basis constant by
Pnf =
Xn
j=1
aj (f)fj
, K(B) = sup
n
∥Pn∥ . There is a Schauder basis (fn) of L
p
(R)
with K((fn)) < 1 + ε whose coefficient functions γn are non-negative. Moreover,
no basis with positive biorthogonal functionals has basis constant one.Where the paper says it
Where the paper says it
Where the paper says it
The paper proves strict lower bounds for non-negative bases: for 1 < p < infinity, the first coordinate projection of a non-negative basis cannot be contractive (Proposition 7. p. 2 Such bases are related by duality. Our first main result constructs non-negative bases of L
p
(R) with basis constants
arbitrarily close to the optimal lower bound of one for every 1 ≤ p < ∞. First, we construct non-negative Schauder bases
of L
p
(R) for every 1 ≤ p < ∞. Although such bases cannot be unconditional,
their basis constants can be made arbitrarily close to one.Where the paper says it
Where the paper says it
The construction for positive biorthogonal functionals in L 1 is also established by running the construction with p= infinity and q=1, yielding a basis whose first coordinate projection cannot be contractive (Proposition 7. p. 3 However, even for p > 2, the whole lattice aZ + b cannot be ordered
as a Schauder basis of translates [24, Proposition 3. When p = 2, our construction allows for rational parameters whose pairwise
separations tend to infinity; see Section 17.Where the paper says it
For p > 2, the construction involves a generator extending to an entire function and translation parameters in the prescribed unbounded set S. p. 0 The generating function f
can be chosen to be the restriction to R of an entire function. For p > 2, the
translation parameters can be chosen from any prescribed unbounded subset of
R, a conclusion that fails for p = 2. Key words and phrases. Schauder bases, non-negative functions, translates, weighted exponential bases, basis constant, L
p
spaces. In applications, it
is often the case that only non-negative functions have physical meaning, so the
ability to read off the admissibility of a signal directly from its coefficients is
desirable. However, constructing non-negative bases is not easy, as non-negativity significantly alters the role of cancellation.Where the paper says it
Where the paper says it
Where the paper says it
The construction for p=2 uses the Fourier transform to yield a real generator with an even weight on the line. p. 3 Moreover, for every
finite p, the Fourier transform of f, interpreted as a distribution when necessary,
must have support equal to R. Interestingly, for 1 < p < 2, Lev and Olevskii [13,
Corollary 2] constructed two functions in L
1
(R) ∩ L
p
(R) whose Fourier transforms
have the same zero set, but whose families of all real translates have different
completeness properties in L
p
(R). Historically, developments on systems of translates have run in parallel with
developments on non-negative coordinate systems. A non-negative basis (fn)
∞
n=1
allows one to infer that a function f := P∞
n=1 anfn is pointwise a. 5) that cn ≥ 0. Moreover, since (µn) is decreasing, we have
µn =
Xn
k=1
ckµn−k ≥ µn
Xn
k=1
ck,
which proves the bound on the sum.Where the paper says it
Where the paper says it
Where the paper says it
The final result demonstrates that for every 2
0, and unbounded set S R, there exists a real-valued function g in L p(R) extending to an entire function, with distinct parameters (lambda n) S, such that (T lambda n g) forms a Schauder basis of L p(R) with a basis constant less than 12 + epsilon.
The constructions confirm non-negative bases for 1 at most p 0 is arbitrary, the infimum of the basis constants is one for every 1 at most p < infinity. p. 0 First, we construct non-negative Schauder bases
of L
p
(R) for every 1 ≤ p < ∞. Although such bases cannot be unconditional,
their basis constants can be made arbitrarily close to one. However, even for p > 2, the whole lattice aZ + b cannot be ordered
as a Schauder basis of translates [24, Proposition 3. When p = 2, our construction allows for rational parameters whose pairwise
separations tend to infinity; see Section 17. STRUCTURED COORDINATE SYSTEMS IN L
p
(R) 9
contractive, and the projection that stops between them has norm 1 + 2θ. Thus,
later stages preserve all earlier partial sum norms, while the same refinement and
density arguments as in the reflexive case give a basis. Nevertheless, depending on the choice
of p, there may exist unconditional Schauder frames of translates. Indeed, the first
author, Odell, Schlumprecht, and Zsák [10, Theorem 3.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The construction for positive biorthogonal functionals in L 1 is achieved by running the construction with p= infinity and q=1, using vectors in (3. p. 8 The fact that a non-negative
basis in L
1
(R) cannot be monotone is a consequence of [9, Theorem 3. To obtain a basis with positive biorthogonal functionals in L
1
, we run the
construction presented in Sections 3 to 5 with p = ∞ and prove norm convergence
for the coefficient functions in L
1
. 3 applies with the parameters λn chosen from any prescribed unbounded
subset of R. However, even for p > 2, the whole lattice aZ + b cannot be ordered
as a Schauder basis of translates [24, Proposition 3. Let 1 ≤ p < ∞ and ε > 0. There is a Schauder basis (fn) of L
p
(R)
with K((fn)) < 1 + ε whose coefficient functions γn are non-negative. , there must exist rearrangements
of the sequence which do not form a basis. Nevertheless, depending on the choice
of p, there may exist unconditional Schauder frames of translates. The columns sj have uniformly bounded
ℓp norms, whereas each of the first N columns has coordinate sum at least ΛN . Since ΛN → ∞, this choice of zN gives the required pairings and makes ωN tend
to zero. STRUCTURED COORDINATE SYSTEMS IN L
p
(R) 9
contractive, and the projection that stops between them has norm 1 + 2θ. Thus,
later stages preserve all earlier partial sum norms, while the same refinement and
density arguments as in the reflexive case give a basis.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Non-negative basis construction for L p(R) spaces can be achieved with basis constants arbitrarily close to one for every 1 <= p < infinity, improving coordinate system efficiency by minimizing the required norm of partial sum projections to less than 1 + epsilon.
-
The ability to select translation parameters from any prescribed unbounded subset of R for p > 2 allows the AI system to generate highly flexible, structured coordinate systems for L p(R), which is crucial for modeling complex, spatially varying data structures where specific frequency distributions are required.
-
For p = 2, the construction of a Schauder basis of translates can yield a generator that extends to an entire function and uses rational parameters with zero upper uniform density, enabling the AI to achieve high-resolution spatial localization and spectral analysis in Hilbert spaces.
-
The iterative construction process allows for the extension of any non-negative basis of step functions to a full Schauder basis of L p(R), where every coordinate projection has a norm at most K0 = 1 + delta + eta, providing robust error bounds for approximation algorithms.
-
The weighted Fourier construction for p = 2 enables the generation of an even real generator with basis constant less than 1 + epsilon using a Gaussian weight bounded above by a positive Gaussian, which facilitates the creation of stable and well-behaved spectral bases in weighted L 2 spaces.
-
For p > 2, the construction of a Schauder basis of translates allows for the selection of translation parameters from any prescribed unbounded set S, enabling coordinate systems tailored to specific, arbitrarily spread spatial frequencies.
Abstract
We construct Schauder bases of L p(R) with significant structural constraints on their elements. For every 1 at most p< infinity, we obtain bases of non-negative functions with basis constants arbitrarily close to one, as well as bases with positive biorthogonal functionals. For every 2 at most p < infinity, we construct Schauder bases consisting of translates of a single function, which can be chosen to extend to an entire function. When p>2, the translation parameters can be chosen from any prescribed unbounded subset of R, a conclusion that fails for p=2. Finally, for p>2, the entire function generating the basis of translates can also be chosen non-negative.
Related papers
- Local-to-Global Isometries in Standard-Norm Generalized Gyrovector Spaces
- Banach Gelfand Triples with Kernel Theorems via Equivalence Bimodules
- The Local Embedding Problem for Hardy Spaces of Dirichlet Series
- The Strong Naor--Schechtman Convolution Inequality on the Product of Cyclic Groups
- Gelfand numbers of Schatten class embeddings in the quasi-Banach range
- A half-line of Schur idempotent norms