p-ary (Non Weakly) Regular Bent Functions in Generalized class
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.
Tom: I'm Tom.
Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.
Tom: We're looking at the paper: "p-ary (Non Weakly) Regular Bent Functions in Generalized class".
Title and authors: Tom: So, we’re diving into the specifics now about this paper's title and authors: "p-ary (Non Weakly) Regular Bent Functions in Generalized class." I think that title immediately tells us what’s going on—it's not just any bent function; it’s a specific type of regular one, and it lives in a generalized family.
Jane: It sounds dense, so let me break down what that means for someone listening who isn't deep into the math. Basically, they are studying functions from F 2n p to Fp where p is an odd prime <ref:2609.18318#pg1,functions from F 2n p to Fp>. They focus on functions that are regular but not weakly regular, and these functions fit into a specific generalized class defined by partial spreads.
Lu: The authors are Alexander Kholosha and Mohit Pal from the Selmer Center at the University of Bergen. Their work builds directly on previous classifications of bent functions, specifically generalizing those proposed by P. Lisonˇek and H. Y. Lu on F 2n p to Fp using partial spreads of F 2n p, which is what this paper does <ref:2609.18318#pg1,partial spreads of F 2n p>.
Meng: So they are taking established classifications and making them more general by changing the conditions on the spread—that m must divide n—which means they are broadening the scope of where these function classes exist.
Tom: Right, and this generalization is key because it shows that these structures aren't just confined to the original cases; they have a broader applicability dictated by that divisibility rule involving m and n.
Jane: It’s about extending the boundaries of what we consider a bent function in this context, showing how different structural constraints—like the size of the spread—lead to different families of functions.
Lu: The real implication for representation learning is that it shows how algebraic properties, like being a generalized Dillon exponent, can constrain the geometric structure in a very specific way.
Meng: So if we find these constraints, it might help us prune the search space when training complex models or designing secure primitives because we know exactly what structures are possible.
Tom: That’s right; it gives us a structural filter to use when designing something that needs to be both bent and fitting within these generalized Dillon exponent definitions.
Jane: It's moving beyond just finding *if* something is bent, toward understanding *why* it is bent in terms of its underlying geometric properties.
The paper's summary: Tom: Moving on to the actual summary of this paper, the main thrust is constructing two classes of functions—the p-ary PS- and PS+ classes—from F 2n p to Fp using these partial m-spreads where m divides n <ref:2609.18318#pg1,from F 2n p to Fp>.
Jane: They are essentially defining membership in these classes based on complex relationships between a function's preimages and the subspaces of those partial spreads, which is a very intricate way to describe their structure.
Lu: The summary highlights that the construction relies on the existence of an m-spread, and as we know, that only happens when m divides n, so they are defining these classes under a necessary condition for their existence.
Meng: So it’s not just any function; it has to satisfy this geometric constraint involving the spread before it can even be considered part of one of these two families.
Tom: And they also investigate monomial functions with generalized Dillon exponents that aren't in the PS- class, and show that if a function is bent and has those exponents, it *must* belong to the m-PS- class.
Jane: That’s a very strong claim: that being bent combined with having these specific monomial forms forces membership into one of these defined classes. It simplifies the classification significantly under those conditions.
Lu: They also analyze regularity conditions by looking at how the Walsh transform coefficients behave, which ultimately leads to the conclusion that if a function is bent, it must be a regular one under certain circumstances.
Meng: So they are connecting algebraic structure—the Dillon exponents—to geometric placement within these spreads, and showing that this connection forces regularity for bent functions. That’s a dense piece of work.
Tom: It really ties together the construction methodology with the properties of generalized Dillon functions, providing a unified view on how these different mathematical concepts overlap.
Jane: And they provide concrete examples, like those ternary vectorial functions and quinary examples, to illustrate exactly what these classes look like in terms of algebraic degree and their membership status.
The paper's improvements: Tom: So what are the suggested improvements or the deeper results they found? One major result is Theorem five which gives us specific values for the Walsh transform coefficients when p is greater than three and f is bent <ref:2609.18318#pg3>.
Jane: That theorem states that if p is greater than three and a function f is bent, then m must divide n, and it’s a regular function in the m-PS- class. It also gives us an explicit formula for the Walsh coefficient at zero, Wa(zero), which is equal to p n, and relates Wa(-b) to that value <ref:2609.18318#pg3>.
Lu: The formula they give for Wa(-b) is: "Wa(-b) equals one minus the difference between p n minus one divided by p m minus one plus a term involving a sum over all elements in U omega f of x."
Meng: That gives us the concrete scaling we need. If we know these exact formulas, we can use them to verify the bent property much more efficiently than just checking all Walsh coefficients.
Tom: And they also have those results about generalized Dillon functions, showing that if f is bent and has those exponents, it belongs to the m-PS- class because of the relationship between their algebraic degree and p n minus one <ref:2609.18318#pg1>.
Jane: So the improvement here is a unified framework: you can use these structural properties—like monomial form or spread membership—to predict membership in the PS- class and regularity automatically.
Lu: It suggests that this method isn't just descriptive; it’s prescriptive, giving us tools to determine if a given function belongs to one of these classes based on its structure.
Meng: From an engineering standpoint, that means we can build better verification tools because we have a clear path: check the structure, apply the theorem, and we know exactly where it lands in the classification.
Conclusion: Tom: So to wrap up on this paper on "p-ary (Non Weakly) Regular Bent Functions in Generalized class," it’s about providing a solid mathematical framework that links monomial structures to geometric constraints imposed by partial m-spreads.
Jane: They successfully constructed two classes, the p-ary PS- and PS+ classes, and proved that if a function is bent and fits certain criteria, it falls into one of those classes under specific conditions on m.
Lu: The main implication is that we have a tool to characterize these functions by combining algebraic properties like Dillon exponents with geometric constraints imposed by the spreads.
Meng: I think the most important part for production is having those explicit formulas for the Walsh transform coefficients, as they give us measurable targets instead of just abstract mathematical statements.
Tom: Exactly. We can use those formulas to check if a function satisfies the bent property without brute-forcing all p n coefficients, which saves a ton of computation time at scale.
Jane: So, we have this unified view where algebraic form dictates geometric placement, and that’s going to be really useful for anyone working on constructing or analyzing these complex functions in the future.
University of Bergen
cs.IT, cs.DM, math.IT
Submitted: 2026-09-16
Updated: 2026-11-18
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 74/100
The gist: Generalized p-ary PS Bent Functions use m-dimensional partial spreads of F2n p to construct two classes of bent functions from F2n p to Fp…
Key concepts
- p-ary Bent Function
- A p-ary bent function is a mathematical mapping from F2n^p to Fp where all its Walsh transform coefficients satisfy $|Wf(y)|^2 = p^n$. This property defines the class of functions being studied, which are related to the structure of vector spaces and their transforms.
- Partial m-spread
- An m-spread is a specific way to partition or cover a set (like F2n^p) using subspaces. A partial m-spread is a subset of these partitions. The construction method relies on using these partial spreads where the integer 'm' must divide 'n' for the construction to work.
- Generalized Dillon Exponents
- This refers to a specific type of monomial function defined as a sum of traces of monomials, $\alpha_j x^{t(p^m-1)}$, where $m$ divides $n$. These functions are analyzed to determine if they belong to the m-PS− class and are classified as generalized Dillon bent functions.
Terminology
Summary
Generalized p-ary PS Bent Functions use m-dimensional partial spreads of F2n p to construct two classes of bent functions from F2n p to Fp, generalizing the classes proposed by Lisonˇek and Lu
Construction Methodology
The paper proposes two classes of bent functions from F2n p to Fp using partial m-spreads of F2n p where m divides n. These constructions generalize the classes PS− and PS+ bent functions proposed by P. p. 1 In [7], Lisonˇek and Lu proposed two classes of bent functions from F
2n
p
to
Fp using partial n-spread of F
2n
p
. These classes generalize the classes PS− and
PS+ of the Boolean bent functions discovered by Dillon [3]. Our construction generalizes the
classes of p-ary PS− and PS+ bent functions proposed by P. Lisonˇek
and H. We use m-dimensional partial spreads of F
2n
p , where p is
an odd prime, n is a positive integer and m divides n, to construct two
classes of bent functions from F
2n
p to Fp. Our construction generalizes the
classes of p-ary PS− and PS+ bent functions proposed by P.Where the paper says it
Where the paper says it
Where the paper says it
Bent Function Properties
A p-ary bent function f is defined as one where all its Walsh transform coefficients satisfy Wf (y) 2 = p n. A p-ary bent function is called weakly regular if a specific condition on its Walsh transform holds involving a complex sign u. p. 1 As defined in [6], f is a p-ary bent function if all its Walsh transform coefficients satisfy |Wf (y)|
2 = p
n. A p-ary bent function f(x) is called weakly
regular if we can fix a complex u (called sign) having unit magnitude such that
p
−n/2Wf (y) = uωf
∗
(y)
for every y ∈ Fpn (then necessarily u = ±1 or u = ±i)
and f
∗
: Fpn 7→ Fp is called the dual of f. A p-ary bent function f(x) is called weakly
regular if we can fix a complex u (called sign) having unit magnitude such that
p
−n/2Wf (y) = uωf
∗
(y)
for every y ∈ Fpn (then necessarily u = ±1 or u = ±i)
and f
∗
: Fpn 7→ Fp is called the dual of f. If u = 1, then f is regular (here
necessarily n is even or p ≡ 1 (mod 4)).Where the paper says it
Where the paper says it
Classification of Classes
The paper introduces two classes of bent functions from F2n p to Fp using partial m-spreads of F2n p with m dividing n. These are referred to as p-ary PS− and PS+ classes. p. 0 We use m-dimensional partial spreads of F
2n
p , where p is
an odd prime, n is a positive integer and m divides n, to construct two
classes of bent functions from F
2n
p to Fp. Our construction generalizes the
classes of p-ary PS− and PS+ bent functions proposed by P.Where the paper says it
Generalized Dillon Exponents
The paper investigates monomial functions with a generalized Dillon exponent that are not in the PS− class. A p-ary function f is said to consist of generalized Dillon exponents if it is a sum of traces of monomials αjx t(pm−1) where m divides n, 0 < t < p n−1 (pm−1) and αt ∈ Fp2n. p. 1 In
Sections 3 and 4, we propose two classes of bent functions from F
2n
p
to Fp using
partial m-spreads of F
2n
p with m dividing n. In the final Section 5, we study
the problem if there exist monomial bent functions with a generalized Dillon
exponent that are not in the PS− class. In
Sections 3 and 4, we propose two classes of bent functions from F
2n
p
to Fp using
partial m-spreads of F
2n
p with m dividing n. In the final Section 5, we study
the problem if there exist monomial bent functions with a generalized Dillon
exponent that are not in the PS− class.Where the paper says it
Where the paper says it
Regularity Conditions
The analysis of the Walsh transform coefficients reveals conditions for regularity based on the structure of subspaces contained in the preimages of function values. For regular bent functions, specific relationships between Wf(0), Wf(a), and ζt p are established. p. 1 If u = 1, then f is regular (here
necessarily n is even or p ≡ 1 (mod 4)). In [7], Lisonˇek and Lu proposed two classes of bent functions from F
2n
p
to
Fp using partial n-spread of F
2n
p
. As defined in [6], f is a p-ary bent function if all its Walsh transform coefficients satisfy |Wf (y)|
2 = p
n. A p-ary bent function f(x) is called weakly
regular if we can fix a complex u (called sign) having unit magnitude such that
p
−n/2Wf (y) = uωf
∗
(y)
for every y ∈ Fpn (then necessarily u = ±1 or u = ±i)
and f
∗
: Fpn 7→ Fp is called the dual of f.Where the paper says it
Where the paper says it
Dual Function Analysis
The dual of an n-PS− function is also n-PS−. The dual subspaces of T ∈ S have the same dimension n and also form a partial n-spread. p. 6 5], the dual
of an n-PS− function is also n-PS−. It follows since the dual subspaces of T ∈ S
have the same dimension n and also form a partial n-spread. It follows since the dual subspaces of T ∈ S
have the same dimension n and also form a partial n-spread. The general case
of m-PS− functions is different.Where the paper says it
Where the paper says it
Example Construction
Example 3 demonstrates that for F5 4, the function f(x) = Tr4(η 2x squared ± x 6) is a 1-PS− quinary bent function of algebraic degree four, so it is not from the PS− class. Example 3 also shows that component functions of F3 8 are all bent functions also from the 2-PS− but not from the PS− class. p. 7 Now assume n is even. Using the same argument as in Example 1 we conclude
that component functions of F are all bent functions also from the 2-PS− but
not from the PS− class.Where the paper says it
Example Construction (Dillon)
Example 2 provides a ternary vectorial function F: F3 2n 7→ F3n given by F(x) = Tr2n n a(x 2(3k+1+1) + x 3k+1+5 + x 8), which is a bent function of algebraic degree four. Component functions of F are all regular bent functions from the completed class M. p. 6 Ternary
vectorial function F : F3
2n 7→ F3
k given by
F(x) = Tr2n
k
a1x
2(3k+1) + a2x
(3k+1)2
is a bent function of algebraic degree four. Moreover, component functions of F
are all regular bent functions from the completed class M. Ternary
vectorial function F : F3
2n 7→ F3
k given by
F(x) = Tr2n
k
a1x
2(3k+1) + a2x
(3k+1)2
is a bent function of algebraic degree four. Moreover, component functions of F
are all regular bent functions from the completed class M.Where the paper says it
Where the paper says it
Example 1 shows that for a ternary vectorial function F: F3 2n 7→ F3n, f(x) = Tr2n k a1x 2(3k+1) + a2x(3k+1) squared, it is a bent function of algebraic degree four. Component functions of F are all regular bent functions from the completed class M.
Example 3 shows that for F5 4, the function f(x) = Tr4(η 14x 28 ± x 36), it is a 1-PS− quinary bent function of algebraic degree four, so it is not from the PS− class.
Theorem 5
If p n > 3 and f is bent, then m divides n and f is a regular bent function from m-PS− class, Wa(0) = p n that gives (−1)k−1Kk−1(N(a)) = (p n − 1)/(p m − 1). Then also Wa(-1) = 1 − (p n − 1)/(p m − 1) + p m Xx∈U ω f(x).
Theorem 4
For any a ∈ F q k, Xx∈F qk χ Trk(axq−1) = (−1)k−1 (q − 1)Kk-1(N(a)). If the sum is over leaders of the multiplicative cosets of F q k then Xx∈F qk /F qk χ Trk(axq−1) = (−1)k-1Kk-1(N(a)).
Theorem 2
Let p be an odd prime and m, n be positive integers such that m 3. Assume S is a partial m-spread of F2n p and f: F2n p 7→ Fp is such that for each T ∈ S, f is constant on T∗. p. 3 Let p be an odd prime and m, n be positive integers such that
m < 2n and p
n > 3. Assume S is a partial m-spread of F
2n
p and f : F
2n
p
7→
Fp is such that for each T ∈ S, f is constant on T
∗
. Assume S is a partial m-spread of F
2n
p and f : F
2n
p
7→
Fp is such that for each T ∈ S, f is constant on T
∗
. Moreover, suppose that
f
−1
(0) = F
2n
p \
S
T ∈S T
∗
. Assume S is a partial m-spread of F
2n
p and f : F
2n
p
7→
Fp is such that for each T ∈ S, f is constant on T
∗
. Moreover, suppose that
f
−1
(0) = F
2n
p \
S
T ∈S T
∗
. Moreover, suppose that
f
−1
(0) = F
2n
p \
S
T ∈S T
∗
. Then f is bent if and only if m | n and f
−1
(j) ∪ {0}
is a union of exactly p
n−1
(p
n−1)
pm−1
elements of S for every j ∈ F
∗
p and for every
nonzero a ∈ F
2n
p
, a
⊥ contains either
N subspaces T ⊆ f
−1
(j) for every j ∈ F
∗
p or (1)
(
N + p
n−m subspaces T ⊆ f
−1
(t) for some t ∈ F
∗
p and
N subspaces T ⊆ f
−1
(j) for remaining j ∈ F
∗
p \ {t} ,
(2)
where N =
p
n−1
(p
n−m−1)
pm−1
.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Corollary 2
Let p be a prime and n be a positive integer such that p n > 3. Assume S is a partial n-spread of F2n p and f: F2n p 7→ Fp is such that for each T ∈ S, f is constant on T∗. p. 3 Let p be an odd prime and m, n be positive integers such that
m < 2n and p
n > 3. Assume S is a partial m-spread of F
2n
p and f : F
2n
p
7→
Fp is such that for each T ∈ S, f is constant on T
∗
. Let p be an odd prime and m, n be positive integers such that
m < 2n and p
n > 3. Assume S is a partial m-spread of F
2n
p and f : F
2n
p
7→
Fp is such that for each T ∈ S, f is constant on T
∗
. Moreover, suppose that
f
−1
(0) = F
2n
p \
S
T ∈S T
∗
. Then f is bent if and only if m | n and f
−1
(j) ∪ {0}
is a union of exactly p
n−1
(p
n−1)
pm−1
elements of S for every j ∈ F
∗
p and for every
nonzero a ∈ F
2n
p
, a
⊥ contains either
N subspaces T ⊆ f
−1
(j) for every j ∈ F
∗
p or (1)
(
N + p
n−m subspaces T ⊆ f
−1
(t) for some t ∈ F
∗
p and
N subspaces T ⊆ f
−1
(j) for remaining j ∈ F
∗
p \ {t} ,
(2)
where N =
p
n−1
(p
n−m−1)
pm−1
. of S for every j ∈ F
∗
p
. Moreover, if f is bent then for every nonzero a ∈ F
2n
p
, a
⊥
contains either
none of the subspaces from S or (6)
only one subspace T ∈ S with T ⊆ f
−1
(t) for some t ∈ F
∗
p
(7)
and these are the only cases possible.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Corollary 3
Let p be an odd prime and n be positive integers such that p n > 3. Assume S is a partial n-spread of F2n p and f: F2n p 7→ Fp is such that for each T ∈ S, f is constant on T∗. Moreover, suppose that f−1(0) = F2n p S T ∈S T and f(0) = z ≠ 0. p. 3 Assume S is a partial m-spread of F
2n
p and f : F
2n
p
7→
Fp is such that for each T ∈ S, f is constant on T
∗
. Moreover, suppose that
f
−1
(0) = F
2n
p \
S
T ∈S T
∗
. Moreover, suppose that
f
−1
(0) = F
2n
p \
S
T ∈S T
∗
. Then f is bent if and only if m | n and f
−1
(j) ∪ {0}
is a union of exactly p
n−1
(p
n−1)
pm−1
elements of S for every j ∈ F
∗
p and for every
nonzero a ∈ F
2n
p
, a
⊥ contains either
N subspaces T ⊆ f
−1
(j) for every j ∈ F
∗
p or (1)
(
N + p
n−m subspaces T ⊆ f
−1
(t) for some t ∈ F
∗
p and
N subspaces T ⊆ f
−1
(j) for remaining j ∈ F
∗
p \ {t} ,
(2)
where N =
p
n−1
(p
n−m−1)
pm−1
. of S for every j ∈ F
∗
p
. Moreover, if f is bent then for every nonzero a ∈ F
2n
p
, a
⊥
contains either
none of the subspaces from S or (6)
only one subspace T ∈ S with T ⊆ f
−1
(t) for some t ∈ F
∗
p
(7)
and these are the only cases possible.Where the paper says it
Where the paper says it
Where the paper says it
Definition 1
The class of (regular) bent functions satisfying conditions of Theorem 2 is called p-ary m-PS− class.
Definition 2
The class of (regular) bent functions satisfying conditions of Theorem 3 is called p-ary m-PS+ class.
Definition 5
Take function f: Fp 2n 7→ Fp given by a sum of traces of monomials αjx t(pm−1) where m divides n, 0 < t < p n−1 (pm−1) and αt ∈ Fp2n. Then f is said to consist of generalized Dillon exponents. p. 1 In
Sections 3 and 4, we propose two classes of bent functions from F
2n
p
to Fp using
partial m-spreads of F
2n
p with m dividing n. In the final Section 5, we study
the problem if there exist monomial bent functions with a generalized Dillon
exponent that are not in the PS− class.Where the paper says it
Definition 2 (Generalized Dillon)
If f is bent then it belongs to the m-PS− class and we call it generalized Dillon bent function.
Theorem 5 (Monomial Bent)
Let p be an odd prime and n, m, t be positive integers such that m divides 2n and gcd(t,(p 2n − 1)/(p m − 1)) = 1. For any a ∈ F p2n, define p-ary function f: Fp 2n 7→ Fp as f(x) = Tr2n axt(pm−1). p. 6 Take any a1 being a nonsquare in F3
2n and define
a2 = ±I
k
a
(3k+1)/2
1
(−1)k
a
(3k−1)(32k+1)/4
1 + a
−(3k−1)(32k+1)/4
1
,
where I is a primitive 4th root of unity in F3
2n and the sign is arbitrary. Ternary
vectorial function F : F3
2n 7→ F3
k given by
F(x) = Tr2n
k
a1x
2(3k+1) + a2x
(3k+1)2
is a bent function of algebraic degree four. They are functions from F
n
2
to F2 for n even, whose
Hamming distance to the set of all affine functions is maximum. Equivalently,
absolute square of their Walsh transform coefficients is constant. Note that p − 1 divides |Dj | when j ̸= 0 and does not divide |D0|. First, assume that f is bent.Where the paper says it
Where the paper says it
Where the paper says it
If p n > 3 and f is bent, then by Theorem 2, m divides n and f is a regular bent function from m-PS− class, Wa(0) = p n that gives (−1) k-1Kk-1(N(a)) = (p n − 1)/(p m − 1). Then also Wa(-b) = 1 − (p n - 1)/(p m - 1) + p m Xx∈U ω f(x).
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
-
Generalization of Bent Functions Construction: The system can construct two classes of bent functions from F2n p to Fp using partial m-spreads of F2n p where m divides n, generalizing the classes PS− and PS+ bent functions proposed by Lisonˇek and Lu.
-
Characterization of Regular Bent Functions: The system can determine if a function is regular or not by checking conditions derived from Theorem 2, which states that
If f is bent then it is a regular bent functions.
-
Verification of Algebraic Degree Bounds: The system can verify if a function has an algebraic degree less than (p minus 1)m for some m <= 2n, which implies the function cannot be constant on any m-dimensional subspace U of F2n p.
-
Walsh Transform Analysis: The system can calculate the Walsh transform coefficients Wa(0) and Wa(-b) for generalized Dillon monomial functions, using formulas like
Wa(0) = 1 + (−1)k−1 (p m − 1)Kk−1(N(a))
. -
Class Membership Testing: The system can test if a generalized Dillon monomial p-ary function is in the m-PS− class by checking if
(p minus 1)k−1Kk−1(N(a)) = (p n − 1)/(p m − 1)
when f is bent.
Abstract
We use m-dimensional partial spreads of 2n, where p is an odd prime, n is a positive integer and m divides n, to construct two classes of bent functions from 2n to. Our construction generalizes the classes of p-ary- and+ regular bent functions proposed by P. Lison ek and H. Y. Lu (Des. Codes Cryptogr. 73 (2014), 209--216). Unlike the original class, such bent functions can be either regular or non weakly regular.
Related papers
- The Gaussian Multiuser Channel Capacity Region
- The mathematical landscape of partial information decomposition: A comprehensive review of properties and measures
- Electromagnetic Signal and Information Theory: A Continuous-Aperture Array Perspective
- Analytically Characterized Optimal Power Control for Signal-Level-Integrated Sensing, Computing and Communication in Federated Learning
- Information Spectrum Methods for Multiple-Access Channels--Part I: epsilon-Capacity Regions in the Mixed Gaussian Case
- Seed Recycling: Explicit Codes for Mixed Synchronization Errors Matching the Leading Term of the Gilbert--Varshamov Bound