How can I add a constraint in JuMP model with ProxSDP optimizer such that the solution is invertible?
Do you mean creating an SDP variable which is invertible?
You can do
@variable(model, Q[1:n, 1:n], PSD)
using LinearAlgebra
eps = 1e-5
X = Q + eps * I
X is then positive definite hence invertible
1 Like