The Complexity of Tensor Rank

Marcus Schaefer,Daniel Stefankovic

2024-01-10

Abstract:We show that determining the rank of a tensor over a field has the same complexity as deciding the existential theory of that field. This implies earlier NP-hardness results by Håstad~\cite{H90}. The hardness proof also implies an algebraic universality result.

Computational Complexity

What problem does this paper attempt to address?

Notions of Tensor Rank

Mandar Juvekar,Arian Nadjimzadah

2023-06-15

Abstract:Tensors, or multi-linear forms, are important objects in a variety of areas from analytics, to combinatorics, to computational complexity theory. Notions of tensor rank aim to quantify the "complexity" of these forms, and are thus also important. While there is one single definition of rank that completely captures the complexity of matrices (and thus linear transformations), there is no definitive analog for tensors. Rather, many notions of tensor rank have been defined over the years, each with their own set of uses. In this paper we survey the popular notions of tensor rank. We give a brief history of their introduction, motivating their existence, and discuss some of their applications in computer science. We also give proof sketches of recent results by Lovett, and Cohen and Moshkovitz, which prove asymptotic equivalence between three key notions of tensor rank over finite fields with at least three elements.

Computational Complexity
Asymptotic tensor rank is characterized by polynomials

Matthias Christandl,Koen Hoeberechts,Harold Nieuwboer,Péter Vrana,Jeroen Zuiddam

2024-11-24

Abstract:Asymptotic tensor rank is notoriously difficult to determine. Indeed, determining its value for the $2\times 2$ matrix multiplication tensor would determine the matrix multiplication exponent, a long-standing open problem. On the other hand, Strassen's asymptotic rank conjecture makes the bold claim that asymptotic tensor rank equals the largest dimension of the tensor and is thus as easy to compute as matrix rank. Despite tremendous interest, much is still unknown about the structural and computational properties of asymptotic rank; for instance whether it is computable. We prove that asymptotic tensor rank is "computable from above", that is, for any real number $r$ there is an (efficient) algorithm that determines, given a tensor $T$, if the asymptotic tensor rank of $T$ is at most $r$. The algorithm has a simple structure; it consists of evaluating a finite list of polynomials on the tensor. Indeed, we prove that the sublevel sets of asymptotic rank are Zariski-closed (just like matrix rank). While we do not exhibit these polynomials explicitly, their mere existence has strong implications on the structure of asymptotic rank. As one such implication, we find that the values that asymptotic tensor rank takes, on all tensors, is a well-ordered set. In other words, any non-increasing sequence of asymptotic ranks stabilizes ("discreteness from above"). In particular, for the matrix multiplication exponent (which is an asymptotic rank) there is no sequence of exponents of bilinear maps that approximates it arbitrarily closely from above without being eventually constant. In other words, any upper bound on the matrix multiplication exponent that is close enough, will "snap" to it. Previously such discreteness results were only known for finite fields or for other tensor parameters (e.g., asymptotic slice rank). We obtain them for infinite fields like the complex numbers.

Computational Complexity,Algebraic Geometry,Quantum Physics
On Comon's Conjecture over Arbitrary Fields

Baodong Zheng,Riguang Huang,Xiaoyu Song,Jinli Xu

DOI: https://doi.org/10.1016/j.laa.2019.11.010

IF: 1.307

2019-01-01

Linear Algebra and its Applications

Abstract:Comon's Conjecture asserts that for a symmetric tensor, the rank is equal to the symmetric rank. For the field of real numbers R or the field of complex numbers C, Zhang et al. proved the conjecture for the case that the rank of a symmetric tensor does not exceed the order. In this paper, we show that over arbitrary fields this conjecture is true when the rank of a tensor is less than its order. Moreover, we construct a counterexample in which the rank of a symmetric tensor whose rank equals its order.
A Note on the Ranks of 2 X 2 X 2 and 2 X 2 X 2 X 2 Tensors

Xu Kong,Yao-Lin Jiang

DOI: https://doi.org/10.1080/03081087.2012.740027

2013-01-01

Abstract:Using rank-1 reduction formula and the vector space spanned by the real rank-1 matrices, we present a different way to show that the maximum possible rank of the 2x2x2 tensors over the real field is 3. Following, we obtain that the maximum rank of the 2x2x2x2 tensors over the real field is less than or equal to 5 and propose another way to show that the maximum rank of the 2x2x2x2 tensors over the complex field is 4, except one special case.
Countably many asymptotic tensor ranks

Andreas Blatter,Jan Draisma,Filip Rupniewski

DOI: https://doi.org/10.1080/03081087.2024.2317906

2024-03-14

Linear and Multilinear Algebra

Abstract:In connection with recent work on gaps in the asymptotic subranks of complex tensors the question arose whether the number of nonnegative real numbers that arise as the asymptotic subrank of some complex tensor is countable. In this short note we settle this question in the affirmative, for all tensor invariants that are algebraic in the sense that they are invariant under field automorphisms of the complex numbers.

mathematics
Low-Rank Tensor Decomposition over Finite Fields

Jason Yang

2024-04-17

Abstract:We show that finding rank-$R$ decompositions of a 3D tensor, for $R\le 4$, over a fixed finite field can be done in polynomial time. However, if some cells in the tensor are allowed to have arbitrary values, then rank-2 is NP-hard over the integers modulo 2. We also explore rank-1 decomposition of a 3D tensor and of a matrix where some cells are allowed to have arbitrary values.

Computational Complexity
Geometric rank of tensors and subrank of matrix multiplication

Swastik Kopparty,Guy Moshkovitz,Jeroen Zuiddam

DOI: https://doi.org/10.48550/arXiv.2002.09472

2020-02-21

Computational Complexity

Abstract:Motivated by problems in algebraic complexity theory (e.g., matrix multiplication) and extremal combinatorics (e.g., the cap set problem and the sunflower problem), we introduce the geometric rank as a new tool in the study of tensors and hypergraphs. We prove that the geometric rank is an upper bound on the subrank of tensors and the independence number of hypergraphs. We prove that the geometric rank is smaller than the slice rank of Tao, and relate geometric rank to the analytic rank of Gowers and Wolf in an asymptotic fashion. As a first application, we use geometric rank to prove a tight upper bound on the (border) subrank of the matrix multiplication tensors, matching Strassen's well-known lower bound from 1987.
The rank of sparse random matrices

Amin Coja-Oghlan,Alperen A. Ergür,Pu Gao,Samuel Hetterich,Maurice Rolvien

2024-06-19

Abstract:We determine the rank of a random matrix over an arbitrary field with prescribed numbers of non-zero entries in each row and column. As an application we obtain a formula for the rate of low-density parity check codes. This formula vindicates a conjecture of Lelarge (2013). The proofs are based on coupling arguments and a novel random perturbation, applicable to any matrix, that diminishes the number of short linear relations.

Combinatorics,Discrete Mathematics,Probability
Real and complex rank for real symmetric tensors with low ranks

Edoardo Ballico,Alessandra Bernardi

DOI: https://doi.org/10.1155/2013/794054

2013-03-11

Abstract:We study the case of a real homogeneous polynomial $P$ whose minimal real and complex decompositions in terms of powers of linear forms are different. We prove that, if the sum of the complex and the real ranks of $P$ is at most $ 3°(P)-1$, then the difference of the two decompositions is completely determined either on a line or on a conic or two disjoint lines.

Algebraic Geometry
Discreteness of asymptotic tensor ranks

Jop Briët,Matthias Christandl,Itai Leigh,Amir Shpilka,Jeroen Zuiddam

2024-09-24

Abstract:Tensor parameters that are amortized or regularized over large tensor powers, often called "asymptotic" tensor parameters, play a central role in several areas including algebraic complexity theory (constructing fast matrix multiplication algorithms), quantum information (entanglement cost and distillable entanglement), and additive combinatorics (bounds on cap sets, sunflower-free sets, etc.). Examples are the asymptotic tensor rank, asymptotic slice rank and asymptotic subrank. Recent works (Costa-Dalai, Blatter-Draisma-Rupniewski, Christandl-Gesmundo-Zuiddam) have investigated notions of discreteness (no accumulation points) or "gaps" in the values of such tensor parameters. We prove a general discreteness theorem for asymptotic tensor parameters of order-three tensors and use this to prove that (1) over any finite field (and in fact any finite set of coefficients in any field), the asymptotic subrank and the asymptotic slice rank have no accumulation points, and (2) over the complex numbers, the asymptotic slice rank has no accumulation points. Central to our approach are two new general lower bounds on the asymptotic subrank of tensors, which measures how much a tensor can be diagonalized. The first lower bound says that the asymptotic subrank of any concise three-tensor is at least the cube-root of the smallest dimension. The second lower bound says that any concise three-tensor that is "narrow enough" (has one dimension much smaller than the other two) has maximal asymptotic subrank. Our proofs rely on new lower bounds on the maximum rank in matrix subspaces that are obtained by slicing a three-tensor in the three different directions. We prove that for any concise tensor, the product of any two such maximum ranks must be large, and as a consequence there are always two distinct directions with large max-rank.

Computational Complexity,Combinatorics,Quantum Physics
Depth-first search for tensor rank and border rank over finite fields

Jason Yang

2024-11-22

Abstract:We present an $O^*\left(|\mathbb{F}|^{(R-n_*)\left(\sum_d n_d\right)+n_*}\right)$-time algorithm for determining whether a tensor of shape $n_0\times\dots\times n_{D-1}$ over a finite field $\mathbb{F}$ has rank $\le R$, where $n_*:=\max_d n_d$; we assume without loss of generality that $\forall d:n_d\le R$. We also extend this problem to its border rank analog, i.e., determining tensor rank over rings of the form $\mathbb{F}[x]/(x^H)$, and give an $O^*\left(|\mathbb{F}|^{H\sum_{1\le r\le R} \sum_d \min(r,n_d)}\right)$-time algorithm. Both of our algorithms use polynomial space.

Computational Complexity
MAXIMAL RANK OF m x n x 2 TENSORS OVER ARBITRARY FIELDS

Xiaoyu Song,Baodong Zheng,Riguang Huang

DOI: https://doi.org/10.1137/21M1431424

IF: 1.908

2022-01-01

SIAM Journal on Matrix Analysis and Applications

Abstract:In this paper, we completely characterize the maximum rank of m x n x 2 tensors over an arbitrary field. We note that the maximal tensor rank may be different over different fields, and we find the maximum rank of a given m x n x 2 tensor is consistent over any field except F-2. We discuss the field in two cases: one is in F-2, and the other is in any field except F-2.
A Functorial Version of Chevalley's Theorem on Constructible Sets

Andreas Blatter

2024-06-13

Abstract:To determine whether an $n\times n$-matrix has rank at most $r$ it suffices to check that the $(r+1)\times (r+1)$-minors have rank at most $r$. In other words, to describe the set of $n\times n$-matrices with the property of having rank at most $r$, we only need the description of the corresponding subset of $(r+1)\times (r+1)$-matrices. We will generalize this observation to a large class of subsets of tensor spaces. A description of certain subsets of a high-dimensional tensor space can always be pulled back from a description of the corresponding subset in a fixed lower-dimensional tensor space.

Algebraic Geometry
Partition and Analytic Rank are Equivalent over Large Fields

Alex Cohen,Guy Moshkovitz

DOI: https://doi.org/10.1215/00127094-2022-0086

2023-11-28

Abstract:We prove that the partition rank and the analytic rank of tensors are equal up to a constant, over finite fields of any characteristic and any large enough cardinality depending on the analytic rank. Moreover, we show that a plausible improvement of our field cardinality requirement would imply that the ranks are equal up to 1+o(1) in the exponent over every finite field. At the core of the proof is a technique for lifting decompositions of multilinear polynomials in an open subset of an algebraic variety, and a technique for finding a large subvariety that retains all rational points such that at least one of these points satisfies a finite-field analogue of genericity with respect to it. Proving the equivalence between these two ranks, ideally over fixed finite fields, is a central question in additive combinatorics, and was reiterated by multiple authors. As a corollary we prove, allowing the field to depend on the value of the norm, the Polynomial Gowers Inverse Conjecture in the d vs. d-1 case.

Combinatorics,Computational Complexity,Algebraic Geometry
On the algebraic proof complexity of Tensor Isomorphism

Nicola Galesi,Joshua A. Grochow,Toniann Pitassi,Adrian She

2023-05-31

Abstract:The Tensor Isomorphism problem (TI) has recently emerged as having connections to multiple areas of research within complexity and beyond, but the current best upper bound is essentially the brute force algorithm. Being an algebraic problem, TI (or rather, proving that two tensors are non-isomorphic) lends itself very naturally to algebraic and semi-algebraic proof systems, such as the Polynomial Calculus (PC) and Sum of Squares (SoS). For its combinatorial cousin Graph Isomorphism, essentially optimal lower bounds are known for approaches based on PC and SoS (Berkholz & Grohe, SODA '17). Our main results are an $\Omega(n)$ lower bound on PC degree or SoS degree for Tensor Isomorphism, and a nontrivial upper bound for testing isomorphism of tensors of bounded rank. We also show that PC cannot perform basic linear algebra in sub-linear degree, such as comparing the rank of two matrices, or deriving $BA=I$ from $AB=I$. As linear algebra is a key tool for understanding tensors, we introduce a strictly stronger proof system, PC+Inv, which allows as derivation rules all substitution instances of the implication $AB=I \rightarrow BA=I$. We conjecture that even PC+Inv cannot solve TI in polynomial time either, but leave open getting lower bounds on PC+Inv for any system of equations, let alone those for TI. We also highlight many other open questions about proof complexity approaches to TI.

Computational Complexity,Data Structures and Algorithms,Logic in Computer Science
Maximal Rank of ${m}\times{n}\times 2$ Tensors over Arbitrary Fields

Xiaoyu Song,Baodong Zheng,Riguang Huang

DOI: https://doi.org/10.1137/21m1431424

IF: 1.908

2022-01-01

SIAM Journal on Matrix Analysis and Applications

Abstract:In this paper, we completely characterize the maximum rank of $m\times n\times 2$ tensors over an arbitrary field. We note that the maximal tensor rank may be different over different fields, and we find the maximum rank of a given $m\times n\times 2$ tensor is consistent over any field except $\mathbb{F}_2$. We discuss the field in two cases: one is in $\mathbb{F}_2$, and the other is in any field except $\mathbb{F}_2$.
A Gap in the Subrank of Tensors

Matthias Christandl,Fulvio Gesmundo,Jeroen Zuiddam

DOI: https://doi.org/10.1137/22M1543276

2023-11-24

Abstract:The subrank of tensors is a measure of how much a tensor can be ''diagonalized''. This parameter was introduced by Strassen to study fast matrix multiplication algorithms in algebraic complexity theory and is closely related to many central tensor parameters (e.g. slice rank, partition rank, analytic rank, geometric rank, G-stable rank) and problems in combinatorics, computer science and quantum information theory. Strassen (J. Reine Angew. Math., 1988) proved that there is a gap in the subrank when taking large powers under the tensor product: either the subrank of all powers is at most one, or it grows as a power of a constant strictly larger than one. In this paper, we precisely determine this constant for tensors of any order. Additionally, for tensors of order three, we prove that there is a second gap in the possible rates of growth. Our results strengthen the recent work of Costa and Dalai (J. Comb. Theory, Ser. A, 2021), who proved a similar gap for the slice rank. Our theorem on the subrank has wider applications by implying such gaps not only for the slice rank, but for any ``normalized monotone''. In order to prove the main result, we characterize when a tensor has a very structured tensor (the W-tensor) in its orbit closure. Our methods include degenerations in Grassmanians, which may be of independent interest.

Algebraic Geometry,Computational Complexity,Combinatorics,Quantum Physics
On the tensor rank of multiplication in any extension of $\F_2$

Stéphane Ballet,Julia Pieltant

DOI: https://doi.org/10.48550/arXiv.1003.1864

2010-03-09

Algebraic Geometry

Abstract:In this paper, we obtain new bounds for the tensor rank of multiplication in any extension of $\F_2$. In particular, it also enables us to obtain the best known asymptotic bound. In this aim, we use the generalized algorithm of type Chudnovsky with derivative evaluations on places of degree one, two and four applied on the descent over $\F_2$ of a Garcia-Stichtenoth tower of algebraic function fields defined over $\F_{2^4}$.
Strength, partition rank and algebraic closure

Benjamin Baily,Amichai Lampert

2024-12-06

Abstract:Let $F$ be a tensor of degree $d$ over a field $K.$ Write $prk(F)$ for the partition rank of $F$ and $\overline{prk} (F) := prk(F\otimes_K \overline{K}),$ where $\overline{K}$ is an algebraic closure of $K.$ We prove that $prk(F) \ll_{d,K} \overline{prk} (F),$ as long as $K$ is both infinite and of finite absolute transcendence degree, partially resolving a conjecture of Adiprasito, Kazhdan and Ziegler. For finite fields we prove a result which, together with results of Moshkovitz-Zhu, implies $prk(F) \ll_d \overline{prk} (F) \log(\overline{prk} (F)+1).$ This improves the state of the art bound for partition rank in terms of analytic rank. By a standard argument, we deduce similar bounds for strength of degree $d$ polynomials, as long as $char(K) > d.$

Number Theory,Algebraic Geometry
Rank Decomposition and Symmetric Rank Decomposition over Arbitrary Fields

Riguang Huang,Baodong Zheng,Xiaoyu Song

DOI: https://doi.org/10.1080/03081087.2021.1935436

2021-01-01

Linear and Multilinear Algebra

Abstract:We give a necessary and sufficient condition for symmetric tensors to have symmetric rank decompositions and give some further results on the Comon's Conjecture (i.e. the rank and symmetric rank agree for a symmetric tensor). In particular, as an application, we prove that if an m-order 2-dimensional symmetric tensor has a symmetric rank decomposition, then its symmetric rank equals its rank or its rank plus one.

The Complexity of Tensor Rank

Notions of Tensor Rank

Asymptotic tensor rank is characterized by polynomials

On Comon's Conjecture over Arbitrary Fields

A Note on the Ranks of 2 X 2 X 2 and 2 X 2 X 2 X 2 Tensors

Countably many asymptotic tensor ranks

Low-Rank Tensor Decomposition over Finite Fields

Geometric rank of tensors and subrank of matrix multiplication

The rank of sparse random matrices

Real and complex rank for real symmetric tensors with low ranks

Discreteness of asymptotic tensor ranks

Depth-first search for tensor rank and border rank over finite fields

MAXIMAL RANK OF m x n x 2 TENSORS OVER ARBITRARY FIELDS

A Functorial Version of Chevalley's Theorem on Constructible Sets

Partition and Analytic Rank are Equivalent over Large Fields

On the algebraic proof complexity of Tensor Isomorphism

Maximal Rank of ${m}\times{n}\times 2$ Tensors over Arbitrary Fields

A Gap in the Subrank of Tensors

On the tensor rank of multiplication in any extension of $\F_2$

Strength, partition rank and algebraic closure

Rank Decomposition and Symmetric Rank Decomposition over Arbitrary Fields