Skip to main content

Functional factorization for matrices and tensors

Project description

FunFact

CI Coverage PyPI version Documentation Status License

FunFact is a library for computing the functional factorization of algebraic tensors, a.k.a. multidimensional arrays. A functional factorization, in our context, is a generalization of the (linear) factorization of tensors. By generalization, we meant to replace the standard inner/outer product between the factor tensors with nonlinear operations.

For example, a rank-1 matrix can be factored into the outer product between a column vector and a row vector: $$ M \approx \mathbf{u} \mathbf{v}^\mathsf{T}, $$ where $M$ is an $n \times m$ matrix, $\mathbf{u}$ is a $n$-dimensional column vector, and $\mathbf{v}$ is a $m$-dimensional row vector. This can be equivalently represented in indexed notation as $$ M_{ij} \approx \mathbf{u}_i \mathbf{v}_j. $$ Moreover, if we replace the standard multiplication operation between $\mathbf{u}_i$ and $\mathbf{v}j$ by an RBF function $\kappa(x, y) = \exp\left[-(x - y)^2\right]$, we then obtain an RBF approximation of $M$ such that: $$ M{ij} \approx \kappa(\mathbf{u}_i, \mathbf{v}_j). $$

Given the rich expressivity of nonlinear operators and functional forms, we expect that a proper functional factorization of a tensor can yield representations that are more compact than what is possible withtin the existing linear framework. However, there is (obviously) no free lunch. The challenges to obtain the functional factorization of a tensor is twofold and involves

  • Finding the most appropriate functional form given a specific piece of data,
  • Finding the component tensors given the functional form for a specific data.

The two points above are exactly what we aim to facilitate using FunFact.

Copyright

FunFact Copyright (c) 2021, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved.

If you have questions about your rights to use or distribute this software, please contact Berkeley Lab's Intellectual Property Office at IPO@lbl.gov.

NOTICE. This Software was developed under funding from the U.S. Department of Energy and the U.S. Government consequently retains certain rights. As such, the U.S. Government has been granted for itself and others acting on its behalf a paid-up, nonexclusive, irrevocable, worldwide license in the Software to reproduce, distribute copies to the public, prepare derivative works, and perform publicly and display publicly, and to permit others to do so.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

funfact-0.7a2.tar.gz (41.2 kB view details)

Uploaded Source

File details

Details for the file funfact-0.7a2.tar.gz.

File metadata

  • Download URL: funfact-0.7a2.tar.gz
  • Upload date:
  • Size: 41.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.6.0 importlib_metadata/4.8.2 pkginfo/1.7.1 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.61.2 CPython/3.8.10

File hashes

Hashes for funfact-0.7a2.tar.gz
Algorithm Hash digest
SHA256 04783e02a085877724f9b43ebd6dd5aa6744f705e29d961a355ae46103d17370
MD5 edf2f7fdb90f61cdaf08e538d52c13db
BLAKE2b-256 5d4215f67ad44302ddcca1c46e1aa86dc5f3524f320827c00b47b26c9b0c034b

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page