HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation
Reports

Tensor Ranks for the Pedestrian for Dimension Reduction and Disentangling Interactions

Alain Franc 1, 2
2 PLEIADE - from patterns to models in computational biodiversity and biotechnology
LaBRI - Laboratoire Bordelais de Recherche en Informatique, Inria Bordeaux - Sud-Ouest, BioGeCo - Biodiversité, Gènes & Communautés
Abstract : A tensor is a multi-way array that can represent, in addition to a data set, the expression of a joint law or a multivariate function. As such it contains the description of the interactions between the variables corresponding to each of the entries. The rank of a tensor extends to arrays with more than two entries the notion of rank of a matrix, bearing in mind that there are several approaches to build such an extension. When the rank is one, the variables are separated, and when it is low, the variables are weakly coupled. Many calculations are simpler on tensors of low rank. Furthermore, approximating a given tensor by a low-rank tensor makes it possible to compute some characteristics of a table, such as the partition function when it is a joint law. In this note, we present in detail an integrated and progressive approach to approximate a given tensor by a tensor of lower rank, through a systematic use of tensor algebra. The notion of tensor is rigorously defined, then elementary but useful operations on tensors are presented. After recalling several different notions for extending the rank to tensors, we show how these elementary operations can be combined to build best low rank approximation algorithms. The last chapter is devoted to applying this approach to tensors constructed as the discretisation of a multivariate function, to show that on a Cartesian grid, the rank of such tensors is expected to be low.
Complete list of metadata

https://hal.inria.fr/hal-03518107
Contributor : David James Sherman Connect in order to contact the contributor
Submitted on : Tuesday, January 18, 2022 - 12:42:53 PM
Last modification on : Tuesday, February 8, 2022 - 5:19:08 PM

Files

Franc-RR-9445.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03518107, version 1
  • ARXIV : 2201.07473

Citation

Alain Franc. Tensor Ranks for the Pedestrian for Dimension Reduction and Disentangling Interactions. [Research Report] RR-9445, Inrae - BioGeCo; Inria Bordeaux Sud-Ouest. 2021. ⟨hal-03518107⟩

Share

Metrics

Record views

59

Files downloads

109