I’ve been working on an SDP model with (among other variables) a hermitian variable W of dimensions n x n. Although it would be nice to have support for complex variables in JuMP, a simple workaround is to declare 2 matrix variables; Wr for the real part and Wi for the imaginary part. n = 111 # Ex…

Improving performance for SDPs with "matrix trace constraints"

Specific Domains Optimization (Mathematical)

martincornejo April 7, 2020, 2:56pm 2

Declaring the matrices Mr[k] and Mi[k] as sparse matrices seems to significantly imporve performance (~10 fold), also: JuMP does exploit sparsity!

Further performance improvement recommendations are welcome!

Topic		Replies	Views
JuMP: Hermitian matrix variable and complex numbers in model Optimization (Mathematical) question , complex-numbers	2	1211	February 6, 2020
How to efficiently generate constraints Optimization (Mathematical)	4	928	January 10, 2019
Advice on optimizing large SDP model generation with JuMP + ComplexOptInterface Optimization (Mathematical) jump , performance	6	555	June 24, 2022
SDP trace minimization with affine constraints: help to formulate the problem Optimization (Mathematical)	7	1557	April 10, 2019
JuMP Mosek sparsity scaling Optimization (Mathematical)	6	657	March 9, 2021

Improving performance for SDPs with "matrix trace constraints"

Related topics