Conversion to SOCP limits performance of ECOS #60
Labels
fairness
Some solvers are not evaluated fairly
help wanted
New contributors needed to start working on this
As @bodono pointed out in #58 (comment),
socp_from_qp
relies on NumPy's implementation, which in turns require the matrix to be positive definite.This requirement comes from the conversion, not from the underlying conic solvers. This yields a https://github.com/qpsolvers/qpsolvers_benchmark/labels/fairness issue since failing to solve the problem is no fault of the solver itself.
The text was updated successfully, but these errors were encountered: