SearcharxivSearch

arXiv subjects

Toshio Sakata

Publications and source records attributed to Toshio Sakata.

14 recordsLinked to original sources

Maximal and typical nonnegative ranks of nonnegative tensors

Let $N_1, \ldots, N_d$ be positive integers with $N_1\leq\cdots\leq N_d$. Set $N=N_1\cdots N_{d-1}$. We show in this paper that an integer $r$ is a typical nonnegative rank of nonnegative tensors of format $N_1\times\cdots\times N_d$ if and only if $r\leq N$ and $r$ is greater than or equals to the generic rank of tensors over $\mathbb{C}$ of format $N_1\times\cdots\times N_d$. We also show that the maximal nonnegative rank of nonnegative tensors of format $N_1\times\cdots\times N_d$ is $N$.

math.RA

Typical ranks of semi-tall real 3-tensors

Let $m$, $n$ and $p$ be integers with $3\leq m\leq n$ and $(m-1)(n-1)+1\leq p\leq (m-1)m$. We showed in previous papers that if $p\geq (m-1)(n-1)+2$, then typical ranks of $p\times n\times m$-tensors over the real number field are $p$ and $p+1$ if and only if there exists a nonsingular bilinear map $\mathbb{R}^m\times \mathbb{R}^n\to\mathbb{R}^{mn-p}$. We also showed that the "if" part also valid in the case where $p=(m-1)(n-1)+1$. In this paper, we consider the case where $p=(m-1)(n-1)+1$ and show that the typical ranks of $p\times n\times m$-tensors over the real number field are $p$ and $p+1$ in several cases including the case where there is no nonsingular bilinear map $\mathbb{R}^m\times \mathbb{R}^n\to\mathbb{R}^{mn-p}$. In particular, we show that the "only if" part of the above mentioned fact does not valid for the case $p=(m-1)(n-1)+1$.

math.RA

Typical ranks for 3-tensors, nonsingular bilinear maps and determinantal ideals

Let $m,n\geq 3$, $(m-1)(n-1)+2\leq p\leq mn$, and $u=mn-p$. The set $\mathbb{R}^{u\times n\times m}$ of all real tensors with size $u\times n\times m$ is one to one corresponding to the set of bilinear maps $\mathbb{R}^m\times \mathbb{R}^n\to \mathbb{R}^u$. We show that $\mathbb{R}^{m\times n\times p}$ has plural typical ranks $p$ and $p+1$ if and only if there exists a nonsingular bilinear map $\mathbb{R}^m\times\mathbb{R}^n\to\mathbb{R}^{u}$. We show that there is a dense open subset $\mathscr{O}$ of $\mathbb{R}^{u\times n\times m}$ such that for any $Y\in\mathscr{O}$, the ideal of maximal minors of a matrix defined by $Y$ in a certain way is a prime ideal and the real radical of that is the irrelevant maximal ideal if that is not a real prime ideal. Further, we show that there is a dense open subset $\mathscr{T}$ of $\mathbb{R}^{ n\times p \times m}$ and continuous surjective open maps $ν\colon\mathscr{O}\to\mathbb{R}^{u\times p}$ and $σ\colon\mathscr{T}\to\mathbb{R}^{u\times p}$, where $\mathbb{R}^{u \times p}$ is the set of $u\times p$ matrices with entries in $\mathbb{R}$, such that if $ν(Y)=σ(T)$, then $\mathrm{rank} T=p$ if and only if the ideal of maximal minors of the matrix defined by $Y$ is a real prime ideal.

math.RA

Holonomic Decent Minimization Method for Restricted Maximum Likelihood Estimation

Recently, the school of Takemura and Takayama have developed a quite interesting minimization method called holonomic gradient descent method (HGD). It works by a mixed use of Pfaffian differential equation satisfied by an objective holonomic function and an iterative optimization method. They successfully applied the method to several maximum likelihood estimation (MLE) problems, which have been intractable in the past. On the other hand, in statistical models, it is not rare that parameters are constrained and therefore the MLE with constraints has been surely one of fundamental topics in statistics. In this paper we develop HGD with constraints for MLE .

stat.CO

Rank of tensors with size 2 x ... x 2

We study an upper bound of ranks of $n$-tensors with size $2\times\cdots\times2$ over the complex and real number field. We characterize a $2\times 2\times 2$ tensor with rank 3 by using the Cayley's hyperdeterminant and some function. Then we see another proof of Brylinski's result that the maximal rank of $2\times2\times2\times2$ complex tensors is 4. We state supporting evidence of the claim that 5 is a typical rank of $2\times2\times2\times2$ real tensors. Recall that Kong and Jiang show that the maximal rank of $2\times2\times2\times2$ real tensors is less than or equal to 5. The maximal rank of $2\times2\times2\times2$ complex (resp. real) tensors gives an upper bound of the maximal rank of $2\times\cdots\times 2$ complex (resp. real) tensors.

math.RA

Upper bound of typical ranks of m x n x ((m-1)n-1) tensors over the real number field

Let $3\leq m\leq n$. We study typical ranks of $m\times n\times ((m-1)n-1)$ tensors over the real number field. The number $(m-1)n-1$ is a minimal typical rank of $m\times n\times ((m-1)n-1)$ tensors over the real number field. We show that a typical rank of $m\times n\times ((m-1)n-1)$ tensors over the real number field is less than or equal to $(m-1)n$ and in particular, $m\times n\times ((m-1)n-1)$ tensors over the real number field has two typical ranks $(m-1)n-1, (m-1)n$ if $m\leq ρ(n)$, where $ρ$ is the Hurwitz-Radon function defined as $ρ(n)=2^b+8c$ for nonnegative integers $a,b,c$ such that $n=(2a+1)2^{b+4c}$ and $0\leq b<4$.

math.RA

Typical ranks of certain 3-tensors and absolutely full column rank tensors

In this paper, we study typical ranks of 3-tensors and show that there are plural typical ranks for m\times n\times p tensors over R in the following cases: (1) 3\leq m\leq ρ(n) and (m-1)(n-1)+1\leq p\leq (m-1)n, where ρ is the Hurwitz-Radon function, (2) m=3, n\equiv 3\pmod 4 and p=2n-1, (3) m=4, n\equiv 2\pmod 4, n\geq 6 and p=3n-2, (4) m=6, n\equiv 4\pmod 8, n\geq 12 and p=5n-4. (5) m=10, n\equiv 24\pmod{32} and p=9n-8.

math.RA

Typical rank of $m\times n\times (m-1)n$ tensors with $3\leq m\leq n$ over the real number field

Tensor type data are used recently in various application fields, and then a typical rank is important. Let $3\leq m\leq n$. We study typical ranks of $m\times n\times (m-1)n$ tensors over the real number field. Let $ρ$ be the Hurwitz-Radon function defined as $ρ(n)=2^b+8c$ for nonnegative integers $a,b,c$ such that $n=(2a+1)2^{b+4c}$ and $0\leq b<4$. If $m \leq ρ(n)$, then the set of $m\times n\times (m-1)n$ tensors has two typical ranks $(m-1)n,(m-1)n+1$. In this paper, we show that the converse is also true: if $m > ρ(n)$, then the set of $m\times n\times (m-1)n$ tensors has only one typical rank $(m-1)n$.

math.RA

Tests of Non-Equivalence among Absolutely Nonsingular Tensors through Geometric Invariants

4x4x3 absolutely nonsingular tensors are characterized by their determinant polynomial. Non-quivalence among absolutely nonsingular tensors with respect to a class of linear transformations, which do not chage the tensor rank,is studied. It is shown theoretically that affine geometric invariants of the constant surface of a determinant polynomial is useful to discriminate non-equivalence among absolutely nonsingular tensors. Also numerical caluculations are presented and these invariants are shown to be useful indeed. For the caluculation of invarinats by 20-spherical design is also commented. We showed that an algebraic problem in tensor data analysis can be attacked by an affine geometric method.

stat.OT

Perfect type of n-tensors

In various application fields, tensor type data are used recently and then a typical rank is important. Although there may be more than one typical ranks over the real number field, a generic rank over the complex number field is the minimum number of them. The set of $n$-tensors of type $p_1\times p_2\times\cdots\times p_n$ is called perfect, if it has a typical rank $\max(p_1,\ldots,p_n)$. In this paper, we determine perfect types of $n$-tensor.

math.RA

Tensor rank problem in statistical high-dimensional data and quantum information theory:their comparisons on the methods and the results

Quantum communication is concerned with the complexity of entanglement of a state and statistical data analysis is concerned with the complexity of a model. A common key word for both is "rank". In this paper we will show that both community is tracing the same target and that the methods used are slightly different. Two different methods, the range criterion method from quantum communication and the determinant polynomial method, are shown as an examples.

quant-ph

About the maximal rank of 3-tensors over the real and the complex number field

High dimensional array data, tensor data, is becoming important in recent days. Then maximal rank of tensors is important in theory and applications. In this paper we consider the maximal rank of 3 tensors. It can be attacked from various viewpoints, however, we trace the method of Atkinson-Stephens(1979) and Atkinson-Lloyd(1980). They treated the problem in the complex field, and we will present various bounds over the real field by proving several lemmas and propositions, which is real counterparts of their results.

math.RA

A simple estimation of the maximal rank of tensors with two slices by row and column operations, symmetrization and induction

The determination of the maximal ranks of a set of a given type of tensors is a basic problem both in theory and application. In statistical applications, the maximal rank is related to the number of necessary parameters to be built in a tensor model. Based on this classical theorem by Bosch we will show the tight bound for 2 x n x n tensors by simple row and column operations, symmetrization and mathematical induction, which has been given by several authors based on eigenvalue theories.

math.RA