def test_relative_parameters(self):
crit = criterion(iterations_max = 1000, xtol = 0.1)
state = {'iteration' : 1001, 'old_parameters' : numpy.ones((1,)), 'new_parameters' : numpy.zeros((1,))}
assert(crit(state))
state = {'iteration' : 5, 'old_parameters' : numpy.ones((1,)), 'new_parameters' : numpy.zeros((1,))}
assert(not crit(state))
state = {'iteration' : 5, 'old_parameters' : numpy.ones((1,)), 'new_parameters' : numpy.ones((1,)) * 0.9}
assert(crit(state))
def test_relative_gradient(self):
crit = criterion(iterations_max = 1000, gtol = 0.1)
state = {'iteration' : 1001, 'function' : Function(1.), 'new_parameters' : numpy.zeros((1,))}
assert(crit(state))
state = {'iteration' : 5, 'function' : Function(1.), 'new_parameters' : numpy.zeros((1,))}
assert(not crit(state))
state = {'iteration' : 5, 'function' : Function(0.09), 'new_parameters' : numpy.zeros((1,))}
assert(crit(state))
def test_relative_value(self):
crit = criterion(iterations_max = 1000, ftol = 0.1)
state = {'iteration' : 1001, 'old_value' : 1., 'new_value' : 0.}
assert(crit(state))
state = {'iteration' : 5, 'old_value' : 1., 'new_value' : 0.}
assert(not crit(state))
state = {'iteration' : 5, 'old_value' : 1., 'new_value' : 0.9}
assert(crit(state))
def test_swp_frgradient_relative():
startPoint = numpy.zeros(2, numpy.float)
optimi = optimizer.StandardOptimizer(
function=Quadratic(),
step=step.FRConjugateGradientStep(),
criterion=criterion.criterion(ftol=0.000001,
iterations_max=1000,
gtol=0.0001),
x0=startPoint,
line_search=line_search.StrongWolfePowellRule())
assert_almost_equal(optimi.optimize(),
numpy.array([1, 3], dtype=numpy.float))
def test_simple_marquardt():
startPoint = numpy.zeros(2, numpy.float)
optimi = optimizer.StandardOptimizer(
function=Quadratic(),
step=step.MarquardtStep(),
criterion=criterion.criterion(gtol=0.00001, iterations_max=200),
x0=startPoint,
line_search=line_search.BacktrackingSearch())
opt = optimi.optimize()
assert_almost_equal(optimi.optimize(),
numpy.array([1, 3], dtype=numpy.float),
decimal=5)
def test_wpr_cwgradient():
startPoint = numpy.empty(2, numpy.float)
startPoint[0] = -1.01
startPoint[-1] = 1.01
optimi = optimizer.StandardOptimizer(
function=Rosenbrock(2),
step=step.CWConjugateGradientStep(),
criterion=criterion.criterion(iterations_max=1000,
ftol=0.00000001,
gtol=0.0001),
x0=startPoint,
line_search=line_search.WolfePowellRule())
assert_array_almost_equal(optimi.optimize(), numpy.ones(2, numpy.float))
def test_simple_marquardt(): startPoint = numpy.zeros(2, numpy.float) optimi = optimizer.StandardOptimizer(function = Quadratic(), step = step.MarquardtStep(), criterion = criterion.criterion(gtol=0.00001, iterations_max=200), x0 = startPoint, line_search = line_search.BacktrackingSearch()) opt = optimi.optimize() assert_almost_equal(optimi.optimize(), numpy.array([1, 3], dtype = numpy.float), decimal = 5)
def test_swp_dypgradient_relative(): startPoint = numpy.zeros(2, numpy.float) optimi = optimizer.StandardOptimizer(function = Quadratic(), step = step.DYConjugateGradientStep(), criterion = criterion.criterion(ftol=0.000001, iterations_max=1000, gtol = 0.0001), x0 = startPoint, line_search = line_search.StrongWolfePowellRule()) assert_almost_equal(optimi.optimize(), numpy.array([1, 3], dtype = numpy.float))
def test_swpr_dygradient(): startPoint = numpy.empty(2, numpy.float) startPoint[0] = -1.01 startPoint[-1] = 1.01 optimi = optimizer.StandardOptimizer(function = Rosenbrock(2), step = step.DYConjugateGradientStep(), criterion = criterion.criterion(iterations_max = 1000, ftol = 0.00000001, gtol = 0.0001), x0 = startPoint, line_search = line_search.StrongWolfePowellRule()) assert_array_almost_equal(optimi.optimize(), numpy.ones(2, numpy.float), decimal = 4)
