tensr
Note: the word tensor used by this library refers to the n-dimensional array representation of a mathematical tensor.
A prototype for a linear algebra library that aims at supporting:
- Multi-demensional array ("tensor") arithmetic (including scalars, vectors, matrices and higher dimensional structures),
- Small (array based) tensors,
- Dense dynamically allocated tensors,
- Sparse tensors,
- Lazy arithmetic,
- Block (sparse or dense) matrices,
- Special matrices (block diagonal, lower triangular, upper triangular, etc.),
- Flat data layout in all tensor types for faster processing,
- Arithmetic between compatible tensors.
Some subset of the features above has already been implemented.
The quality criteria for this library (in no particular order) are
- performance
- simplicity (usage as well as implementation)
- safety
Meeting these criteria is work in progress.
Higher level algorithms like decompositions are currently outside the scope of this project and are expected to be provided by third-party crates.
License
This repository is licensed under either of
- Apache License, Version 2.0, (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
- MIT License (LICENSE-MIT or https://opensource.org/licenses/MIT)
at your option.