Skip to content

Toolbox that provides functions and data structures to generate and handle ultrametric matrices.

License

Notifications You must be signed in to change notification settings

aoertel/ultrametric-matrix-tools

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

65 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Ultrametric matrix tools

DOI

This toolbox provides functions and data structures to construct and handle ultrametric matrices in Rust and Python. The aim of the project is to provide efficient tools for ultrametric matrices and ultrametric trees. Currently, the project has the following features.

Features:

  • Generate a random ultrametric matrix
  • Test ultramtic matrix property
  • Construct the ultrametric tree from the ultrametric matrix
  • Get properties of ultrametric tree
  • Fast multiplication of ultrametric matrix with vector

The implementation is written in Rust and can be cross-compiled to Python.

Table of Contents

Quickstart

Quickstart Rust

Add the following to the Cargo.toml file:

[dependencies]
# TODO: replace the * by the latest version.
ultrametric_matrix_tools = "*"

An example of the usage of is:

use ultrametric_matrix_tools::na::{DMatrix, DVector};
use ultrametric_matrix_tools::UltrametricTree;

fn main() {
    let matrix = DMatrix::from_vec(
        4,
        4,
        vec![
            0.0, 1.0, 3.0, 1.0, 1.0, 3.0, 1.0, 1.0, 3.0, 1.0, 5.0, 1.0, 1.0, 1.0, 1.0, 1.0,
        ],
    );
    let vector = DVector::from_vec(vec![4.0, 2.0, 7.0, 5.0]);

    let tree = UltrametricTree::from_matrix(&matrix);
    let product = tree * vector;
}

More examples can be found in ./examples/.

Quickstart Python

You can install the current release by running:

pip install ultrametric_matrix_tools

An example of the construction of the ultrametric tree and multiplication with it is:

from ultrametric_matrix_tools import UltrametricTree
import numpy as np

matrix = np.array([[0.0, 1.0, 3.0, 1.0], [1.0, 3.0, 1.0, 1.0], [
                  3.0, 1.0, 5.0, 1.0], [1.0, 1.0, 1.0, 1.0]])
vector = np.array([4.0, 2.0, 7.0, 5.0])

tree = UltrametricTree(matrix)
product = tree.mult(vector)

More examples can be found in ./examples/.

Build

Build Rust Library

The Rust library is build by running:

cargo build --release

The compiled Rust library is located in ./target/release/ and can be copied from there.

Build Python Module

The Python module is build from the Rust code using the PyO3. To build the Python module, you need to install Cargo and run:

cargo build --release

The compiled Python module is located in ./target/release/ and can be copied from there.

To export the Python wheels from a Linux host system run the following commands:

Linux (requires docker):

docker run --rm -v $(pwd):/io konstin2/maturin build --release

Windows (requires mingw32-python and mingw64-python):

make python_package_windows

Currently, cross-compiling to macOS is not supported.

Examples

Rust Example

You can try out the Rust examples, you need to install Cargo. You can try out the Python examples located in ./examples/ by running the following command:

cargo run --release --example [example_name]

E.g. to run the multiplication example run:

cargo run --release --example multiplication

Python Example

To run the Python examples, you need to install Cargo. You can try out the Python examples located in ./examples/ by running the following command:

make python_example name=[example_name]

E.g. to run the multiplication example run:

make python_example name=multiplication

Alternatively, if you have the Python package already installed via pip, then you can run the examples directly:

python [example_name].py

License

This project is under the Apache-2.0 license.

Benchmarks

The benchmarks use criterion for cargo, which can be installed by running:

cargo install cargo-criterion

The benchmarks can be found in ./benches and are run by:

cargo criterion --bench [benchmark_name]