allow upper bound for kappa to boost accuracy of A

This commit is contained in:
kalmar 2017-03-13 11:22:51 +01:00
parent 26f4900e82
commit 261e4395f5
1 changed files with 4 additions and 2 deletions

View File

@ -47,7 +47,7 @@ function laplacian_coeff(S, basis)
end
function create_SDP_problem(matrix_constraints, Δ::GroupAlgebraElement)
function create_SDP_problem(matrix_constraints, Δ::GroupAlgebraElement; upper_bound=Inf)
N = size(Δ.product_matrix,1)
const Δ² = Δ*Δ
@assert length(Δ) == length(matrix_constraints)
@ -56,7 +56,9 @@ function create_SDP_problem(matrix_constraints, Δ::GroupAlgebraElement)
JuMP.@SDconstraint(m, A >= 0)
JuMP.@constraint(m, sum(A[i] for i in eachindex(A)) == 0)
JuMP.@variable(m, κ >= 0.0)
JuMP.@constraint(m, κ <= 0.26)
if upper_bound < Inf
JuMP.@constraint(m, κ <= upper_bound)
end
JuMP.@objective(m, Max, κ)
for (pairs, δ², δ) in zip(matrix_constraints, Δ².coefficients, Δ.coefficients)