def test_convergence(self):
gnCriterion = GradientNormCriterion(10**-3)
step_opt = ArmijoBacktracking(1, 0.5)
opt = DFP(gnCriterion, step_opt)
res = opt.optimize(self.f, np.array([-1, -2]))
self.assertInRange(res.x, np.array([1, 1]), 10**-2)
def test_convergence(self):
gnCriterion = GradientNormCriterion(10**-3)
step_opt = ArmijoBacktracking(1, 0.5)
unc_opt = BFGS(gnCriterion, step_opt)
opt = Penalty(unc_opt)
res = opt.optimize(self.f,
np.array([-6, 9]),
eq_constraints=[lambda x: x[0] + x[1] - 2],
ineq_constraints=[lambda x: -x[0]])
self.assertInRange(res.x, np.array([1, 1]), 10**-3)
def test_convergence(self):
gnCriterion = GradientNormCriterion(10**-3)
step_opt = ArmijoBacktracking(1, 0.5)
unc_opt = BFGS(gnCriterion, step_opt)
opt = AugmentedLagrangian(unc_opt)
res = opt.optimize(self.f,
np.array([4, 8]),
eq_constraints=[lambda x: x[0] - 1],
ineq_constraints=[lambda x: x[0] + x[1] - 2])
self.assertInRange(res.x, np.array([1, 0]), 10**-3)
def test_convergence(self):
step_opt = ArmijoBacktracking(1, 0.5)
arg_criterion = IterationNumberCriterion(500)
unc_opt = HookeJeeves(arg_criterion,
HookeJeeves.generate_pertubation_vector(2, 0.1))
opt = Interior(unc_opt, epsilon=10**-4)
res = opt.optimize(self.f,
np.array([-1, -1]),
eq_constraints=[lambda x: x[0] - 1],
ineq_constraints=[lambda x: x[0] + x[1] - 2])
self.assertInRange(res.x, np.array([1, 0]), 10**-3)
def setUp(self):
self.__x0 = np.array([0, 0])
self.__gnCriterion = GradientNormCriterion(10**-3)
self.__step_opt = ArmijoBacktracking(1, 0.5)
self.__n = 3
self.__opt = np.array([3, 1])