zfista : A globally convergent fast iterative shrinkage-thresholding algorithm with a new momentum factor for single and multi-objective (convex) optimization

This code repository provides a solver for the proximal gradient method (ISTA) and its acceleration (FISTA) for both single and multi-objective optimization problems, including the experimental code for the Paper1 and Paper2.

An accelerated proximal gradient method for multiobjective optimization
Hiroki Tanabe, Ellen H. Fukuda, and Nobuo Yamashita

A globally convergent fast iterative shrinkage-thresholding algorithm with a new momentum factor for single and multi-objective convex optimization
Hiroki Tanabe, Ellen H. Fukuda, and Nobuo Yamashita

The solver can deal with the unconstrained problem written by

$$ \min_{x \in \mathbf{R}^n} \quad F(x) \coloneqq f(x) + g(x), $$

where $f$ and $g$ are scalar or vector valued function, $f$ is continuously differentiable, $g$ is closed, proper and convex. Note that FISTA also requires $f$ to be convex.

• Documentation: https://zalgo3.github.io/zfista/

Requirements

• Python 3.9 or later

Install

pip install zfista

Quickstart

from zfista import minimize_proximal_gradient
help(minimize_proximal_gradient)

For developers

Installation

pip install -e .

Examples

You can run some examples on jupyter notebooks.

pip install -e ".[examples]"
jupyter notebook

Testing

You can run all tests by

pip install tox
tox

Benchmark

You can run the benchmark by

pip install -e ".[bench]"
python runtests.py