MathProgComplex module is a tool for polynomial optimization problems with complex variables. These problems consist in optimizing a generic complex multivariate polynomial function, subject to some complex polynomial equality and inequality constraints.
MathProgComplex module enables:
The modules need to be accessible from julia's path to be loaded with the
This can be done by running
push!(LOAD_PATH, "location/of/modules"), which will update the path for the current session.
The path can be updated at every start of a Julia session by adding the command to the
.juliarc.jl file, which should be located (or created) at the location given by
MathProgComplex environment provides a structure and methods for working with complex polynomial optimization problems subject to polynomial constraints. It is based on a polynomial environment that allows to work on polynomial objects with natural operations (+, -, *, conj, |.|^2).
Variable: it is a structure with a name (a string) and a type (Complex, Real or Bool).
using MathProgComplex a = Variable("a", Complex) b = Variable("b", Real) c = Variable("c", Bool)
Polynomial can be constructed by calling the respective constructors or with algebraic operations (+, -, *, conj, |.|).
Exponent is a product of
expo1 = a*b expo2 = conj(a)^3*b^5 expo3 = abs2(a) # =a*conj(a)
Polynomial is a sum of
Exponents times complex numbers.
p = 3*expo1 + (4+2im)*expo2 +2im*expo3
Point type holds the variables at which polynomials can be evaluated.
- isconst, isone - evaluate - abs2, conj - is_homogeneous: tests if p(exp(iϕ)X) = p(X) ∀X∈R^n, Φ∈R (phaseinvariant equation) - cplx2real: convert the provided object to a tuple of real and imaginary part, expressed with real and imaginary part variables.
using MathProgComplex a = Variable("a", Complex) b = Variable("b", Real) p = a*conj(a) + b + 2 print(p) # 2 + b + conj(a) * a pt = Point([a, b], [1+2im, 1+im]) print(pt) # a 1 + 2im # b 1 pt = pt - Point([a], ) print(pt) # a 0 + 2im # b 1 val = evaluate(p, pt) # 7.0 + 0.0im p_real, p_imag = cplx2real(p) pt_r = cplx2real(pt) val_real = evaluate(p_real, pt_r) # 7.0 val_imag = evaluate(p_imag, pt_r) # 0
Constraint structure holds a
Polynomial and complex upper and lower bounds.
Problem is made up of:
Variables of the problem (updated internally),
Here is a full example, more can be found in the
using MathProgComplex a = Variable("a", Complex) b = Variable("b", Real) p_obj = abs2(a) + abs2(b) + 2 p_cstr1 = 3a + b + 2 p_cstr2 = abs2(b) + 5a*b + 2 pb = Problem() set_objective!(pb, p_obj) add_constraint!(pb, "Cstr 1", p_cstr1 << (3+5im)) # 2 + (3.0)*a + b < 3 + 5im add_constraint!(pb, "Cstr 2", (2-im) << p_cstr2 << (3+7im)) # 2 - 1im < 2 + (5.0)*a * b + b^2 < 3 + 7im add_constraint!(pb, "Cstr 3", p_cstr2 == 0) # 2 + (5.0)*a * b + b^2 = 0 print(pb) # ▶ variables: a b # ▶ objective: 2 + conj(a) * a + b^2 # ▶ constraints: # → Cstr 1: 2 + (3.0)*a + b < 3 + 5im # → Cstr 2: 2 - 1im < 2 + (5.0)*a * b + b^2 < 3 + 7im # → Cstr 3: 2 + (5.0)*a * b + b^2 = 0 pt_sol = Point([a, b], [1, 1+2im]) print("slack by constraint at given point:\n", get_slacks(pb, pt_sol)) # slack by constraint at given point: # Cstr 1 -3.0 + 5.0im # Cstr 2 -5.0 + 1.0im # Cstr 3 -8.0 - 0.0im
The polynomial optimization problems can be converted into JuMP models or be exported into formatted text files to be used in another language.
m, JuMPvar = get_JuMP_cartesian_model(pb, solver) solve(m)
.dat text format used in this module allows storing polynomial optimization problems in inary, Real or Complex variables, along with a scalar value for each variable (a point).
export_to_dat(pb, amplexportpath, point) run_knitro(amplexportpath, amplscriptpath)
6 days ago