PANOC is a Newton-type accelerated proximal gradient method for nonsmooth optimization: this repository contains its generic implementation in Julia.
Deprecated: an up-to-date implementation of the same algorithm is available as part of ProximalAlgorithms.jl.
From the Julia REPL, hit ]
to enter the package manager, then
pkg> add https://github.com/kul-forbes/PANOC.jl
PANOC solves optimization problems of the form
minimize f(Ax) + g(x)
where x
is the decision variable, while
f
is a smooth function,g
is a function with easily computable proximal operator, both of which can be taken from ProximalOperators.jl;A
is a linear mapping, e.g. a matrix or an object from linear operator packages such as AbstractOperators.jl, LinearMaps.jl, or LinearOperators.jl.
The above problem is solved calling the panoc
function:
julia> using PANOC
julia> x_opt, it = panoc(f, A, g, x0)
where x0
is the starting point of the iterations.
This returns the optimal point found, and the number of iterations it took to find it.
The full list of options is described in the docstring, accessible with
julia> ?panoc
If you use this package for your publications, please consider including the following BibTeX entries in the references
@inproceedings{stella2017simple,
author = {Stella, Lorenzo and Themelis, Andreas and Sopasakis, Pantelis and Patrinos, Panagiotis},
title = {A simple and efficient algorithm for nonlinear model predictive control},
booktitle = {56th IEEE Conference on Decision and Control (CDC)},
year = {2017},
pages = {1939-1944},
doi = {10.1109/CDC.2017.8263933},
url = {https://doi.org/10.1109/CDC.2017.8263933}
}
@misc{stella2018panoc,
author = {Stella, Lorenzo},
title = {{PANOC}.jl: {N}ewton-type accelerated proximal gradient method in Julia},
howpublished = {\url{https://github.com/kul-forbes/PANOC.jl}},
year = {2018}
}
Stella, Themelis, Sopasakis, Patrinos, A simple and efficient algorithm for nonlinear model predictive control, 56th IEEE Conference on Decision and Control (2017).