Closed htadashi closed 7 months ago
You can do
con_ref = @constraint(model, p in SOSCone())
gram_matrix(con_ref) - p
# or
sos_decomposition(con_ref, tol) - p
The difference is that gram_matrix
gives you the plain gram matrix while sos_decomposition
additionally discards the negligible eigenvalues with tol
.
Thanks for clarifying! :)
Hi @blegat, is there (or any plans to add) a function in SumOfSquares.jl that checks the difference between the SOS decomposition found by the solver and the original polynomial (similar to YALMIP's checkset command)?