Skip to content

GPU-aware interior point solver for convex optimization problems

Notifications You must be signed in to change notification settings

fdeguire03/InteriorPoint-GPU

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

60 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

InteriorPoint-GPU

GPU-aware interior point solver for convex optimization problems

Please note that this package is still under development and contains a few bugs. Most notably, the solvers occasionally fail to converge on some problems, despite confirmed feasibility. We have noticed this issue most commonly on heavily constrained (hundreds of equality and inequality constraints) QPs and SOCPs. Likely caused by numerical errors in our backtracking search / linear solves or by suboptimal hyperparameter values (maximum number of iterations, initial t, mu, etc. are all set nominally instead of dynamically tuned to each problem instance).

See demo.ipynb for overview of usage. Init method of LPSolver also describes all of the input parameters.

Currently implemented for LPs, QPs, and SOCPs. Dedicated solver using ADMM for LASSO problems in parallel.

Implemented as part of the final project for EE364B Convex Optimization II Spring 2024 at Stanford University.

About

GPU-aware interior point solver for convex optimization problems

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published