search

MadNLP / DynamicNLPModels.jl

NLPModels for dynamic optimization
MIT License
11 stars 1 forks source link

Add support for sparse `Q,R,A,B` #6

Closed sshin23 closed 2 years ago

sshin23 commented 2 years ago

Currently we do not take into account the sparsity in Q,R,A,B. The hessian/Jacobian should respect the sparsity of Q,R,A,B given by the user. A specialized _build_H and _build_J function should be added via multiple dispatch

dlcole3 commented 2 years ago

Should I assume that Q and R are both sparse in the specialized _build_H function? Or should I create multiple functions so that Q can be spares while R is dense or vice versa?

sshin23 commented 2 years ago

@dlcole3 should make it all sparse. We have enforced they have the same type in LQDynamicData

sshin23 commented 2 years ago

closing it as the current approach preserves the sparsity. Thanks @dlcole3 for the corretion