linox: Linear Operators in JAX

linox is a Python package that provides a collection of linear operators for JAX, enabling efficient and flexible linear algebra operations with lazy evaluation. This package is designed as an JAX alternative to probnum.linops, but it is currently still under development having less and more instable features. It has no dependencies other than JAX and plum for multiple dispatch.

Features

  • Lazy Evaluation: All operators support lazy evaluation, allowing for efficient computation of complex linear transformations
  • JAX Integration: Built on top of JAX, providing automatic differentiation, parallelization, JIT compilation, and GPU/TPU support
  • Composable Operators: Operators can be combined to form complex linear transformations

Linear Operators

Basic Operators

  • Matrix: General matrix operator
  • Identity: Identity matrix operator
  • Diagonal: Diagonal matrix operator
  • Scalar: Scalar multiple of identity
  • Zero: Zero matrix operator
  • Ones: Matrix of ones operator

Block Operators

  • BlockMatrix: General block matrix operator
  • BlockMatrix2x2: 2x2 block matrix operator
  • BlockDiagonal: Block diagonal matrix operator

Low Rank Operators

  • LowRank: General low rank operator
  • SymmetricLowRank: Symmetric low rank operator
  • IsotropicScalingPlusSymmetricLowRank: Isotropic scaling plus symmetric low rank
  • PositiveDiagonalPlusSymmetricLowRank: Positive diagonal plus symmetric low rank

Special Operators

  • Kronecker: Kronecker product operator
  • Permutation: Permutation matrix operator
  • EigenD: Eigenvalue decomposition operator

Benefits of JAX Integration

  • Automatic Differentiation: Compute gradients automatically through operator compositions
  • JIT Compilation: Speed up computations with just-in-time compilation
  • Vectorization: Efficient batch processing of linear operations via e.g. jax.vmap
  • GPU/TPU Support: Run computations on accelerators without code changes
  • Functional Programming: Pure functions enable better optimization and parallelization

Quick Example

import jax
import jax.numpy as jnp
from linox import Matrix, Diagonal, BlockMatrix

# Create operators
A = Matrix(jnp.array([[1, 2], [3, 4]], dtype=jnp.float32))
D = Diagonal(jnp.array([1, 2], dtype=jnp.float32))

# Compose operators
B = BlockMatrix([[A, D], [D, A]])

# Apply to vector
x = jnp.ones((4,), dtype=jnp.float32)
y = B @ x  # Lazy evaluation

# Parallelize over batch of vectors
x_batched = jnp.ones((10, 4), dtype=jnp.float32)
y_batched = jax.vmap(B)(x_batched)

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (0.0.3)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for lenardrommel from gravatar.com lenardrommel

Credits

Author: Tobias Weber

Requires

Python >=3.10

Provides Extra

dev test

Tags

Linear operators Operators Matrix-free linear algebra

Classifiers

Report project as malware

Download files

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

Source Distribution

linox-0.0.1.tar.gz (36.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

linox-0.0.1-py3-none-any.whl (27.3 kB view details)

Uploaded Python 3

File details

Details for the file linox-0.0.1.tar.gz.

File metadata

  • Download URL: linox-0.0.1.tar.gz
  • Upload date:
  • Size: 36.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for linox-0.0.1.tar.gz
Algorithm Hash digest
SHA256 de873f43764556d8f4d15d9eae8eaf7bcbbbb1a9b1dcfd7721e3dd4837307f2d
MD5 53c0199dbcd66b722ddc0edbf1052fdb
BLAKE2b-256 a0ddd3985761cc3d2e32ce711fec989a0c04739b5ddacbfa1f2c334e20f6d187

See more details on using hashes here.

Provenance

The following attestation bundles were made for linox-0.0.1.tar.gz:

Publisher: publish.yml on lenardrommel/linox

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file linox-0.0.1-py3-none-any.whl.

File metadata

  • Download URL: linox-0.0.1-py3-none-any.whl
  • Upload date:
  • Size: 27.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for linox-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 601c2ef618913174c76a4cac22930d02f8b91a6d8d268aae1152d8fb84c172a5
MD5 61430c5c5def4aaeb8f3a5ab9bde0f7e
BLAKE2b-256 9c1c7296fd8513fd2f6a7a69d5b6fb9489c40f3d771d650a8273b38808338f86

See more details on using hashes here.

Provenance

The following attestation bundles were made for linox-0.0.1-py3-none-any.whl:

Publisher: publish.yml on lenardrommel/linox

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Release history Release notifications | RSS feed

0.0.3

2 files

0.0.2

2 files

This release

0.0.1 This release

2 files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page