def test_single_variable_quadratic(self):
for initial_step in [0.001, 0.01, 10]:
obj, param, optimum = problems.build_single_variable_quadratic()
opt = AdaptiveGradientDescent(obj, param, initial_step)
opt.max_iters = 60
opt.optimize()
np.testing.assert_almost_equal(opt.param.to_vector(), optimum)
def test_single_variable_quadratic(self):
for method in self.methods:
obj, param, optimum = problems.build_single_variable_quadratic()
opt = ScipyOptimizer(obj, param, method)
opt.optimize()
np.testing.assert_almost_equal(opt.param.to_vector(),
optimum,
decimal=4)
def test_single_variable_quadratic(self):
obj, param, optimum = problems.build_single_variable_quadratic()
opt = GradientDescent(obj, param, 0.05)
opt.max_iters = 500
opt.optimize()
np.testing.assert_almost_equal(opt.param.encode(), optimum)
def test_single_variable_quadratic(self):
obj, param, optimum = problems.build_single_variable_quadratic()
opt = Nag(obj, param, 0.1)
opt.max_iters = 60
opt.optimize()
np.testing.assert_almost_equal(opt.param.encode(), optimum, decimal=4)
def test_single_variable_quadratic(self):
obj, param, optimum = problems.build_single_variable_quadratic()
opt = Adagrad(obj, param, 1)
opt.max_iters = 40
opt.optimize()
np.testing.assert_almost_equal(opt.param.encode(), optimum)