def test_correct_training_data_length(python, input_size):
data_input = np.arange(input_size).reshape((-1, 1))
training_data = ImplicitTrainingData(data_input, data_input) \
if python \
else bingocpp.ImplicitTrainingData(data_input, data_input)
assert len(training_data) == input_size
def test_correct_partial_calculation_in_training_data_2(python):
data_input = np.arange(20, dtype=float).reshape((20, 1)) * 2.0
data_input = np.vstack((data_input, [np.nan], data_input))
training_data = ImplicitTrainingData(data_input) \
if python \
else bingocpp.ImplicitTrainingData(data_input)
expected_derivative = np.full((26, 1), 2.0)
np.testing.assert_array_almost_equal(training_data.dx_dt,
expected_derivative)
def test_evaluate_fitness_vector_implicit():
print("-----test_evaluate_fitness_vector_implicit-----")
x_t = snake_walk()
y = (x_t[:, 0] + x_t[:, 1])
x = np.hstack((x_t, y.reshape([-1, 1])))
py_training_data = ImplicitTrainingData(x)
py_implicit_regressor = ImplicitRegression()
c_training_data = bingocpp.ImplicitTrainingData(x)
c_implicit_regressor = bingocpp.ImplicitRegression()
py_manip = AGraphCpp.AGraphCppManipulator(3, 64, nloads=2)
py_manip.add_node_type(2)
py_manip.add_node_type(3)
py_manip.add_node_type(4)
c_manip = bingocpp.AcyclicGraphManipulator(3, 64, nloads=2)
c_manip.add_node_type(2)
c_manip.add_node_type(3)
c_manip.add_node_type(4)
temp = np.array([[1, 0, 0], [0, 0, 0], [0, 1, 1], [3, 2, 2], [4, 2, 3],
[4, 4, 0], [3, 2, 0], [3, 0, 0], [3, 4, 2], [3, 1, 4],
[2, 6, 0], [3, 8, 7], [2, 3, 11], [3, 7, 7], [3, 2, 7],
[2, 9, 5], [4, 4, 14], [3, 9, 1], [3, 4, 5], [0, 1, 1],
[4, 8, 4], [1, -1, -1], [3, 20, 5],
[2, 9, 17], [1, -1, -1], [4, 23, 4], [4, 25, 19],
[2, 26, 14], [4, 22, 10], [0, 0, 0], [2, 0, 17],
[2, 29, 16], [4, 23, 14], [4, 3, 22], [0, 2, 2],
[2, 31, 27], [2, 35, 28], [2, 25, 29], [2, 36, 28],
[3, 8, 29], [3, 4, 24], [2, 14, 9], [4, 25, 9], [0, 2, 2],
[3, 26, 10], [3, 12, 6], [0, 1, 1], [4, 42,
26], [3, 41, 6],
[3, 13, 1], [3, 42, 36], [3, 15, 34], [2, 14, 23],
[2, 13, 12], [0, 2, 2], [2, 28, 45], [3, 1, 12],
[3, 15, 30], [3, 34, 38], [3, 43, 50], [2, 31, 9],
[2, 54, 46], [1, -1, -1], [4, 60, 29]])
py_1 = py_manip.generate()
c_1 = c_manip.generate()
py_1.command_array = np.copy(temp)
c_1.stack = np.copy(temp)
c_manip.simplify_stack(c_1)
assert py_training_data.x.all() == pytest.approx(c_training_data.x.all())
assert py_training_data.dx_dt.all() == pytest.approx(
c_training_data.dx_dt.all())
py_fit = py_implicit_regressor.evaluate_fitness(py_1, py_training_data)
py_fit = py_implicit_regressor.evaluate_fitness_vector(
py_1, py_training_data)
c_fit = c_implicit_regressor.evaluate_fitness(c_1, c_training_data)
c_fit = c_implicit_regressor.evaluate_fitness_vector(c_1, c_training_data)
assert py_fit.all() == pytest.approx(c_fit.all())
def test_getting_subset_of_training_data(python):
data_input = np.arange(5).reshape((5, 1))
training_data = ImplicitTrainingData(data_input, data_input)\
if python \
else bingocpp.ImplicitTrainingData(data_input, data_input)
subset_training_data = training_data[[0, 2, 3]]
expected_subset = np.array([[0], [2], [3]])
np.testing.assert_array_equal(subset_training_data.x,
expected_subset)
np.testing.assert_array_equal(subset_training_data.dx_dt,
expected_subset)
def test_correct_partial_calculation_in_training_data(python):
data_input = np.arange(20, dtype=float).reshape((20, 1))
data_input = np.c_[data_input * 0,
data_input * 1,
data_input * 2]
training_data = ImplicitTrainingData(data_input) \
if python \
else bingocpp.ImplicitTrainingData(data_input)
expected_derivative = np.c_[np.ones(13) * 0,
np.ones(13) * 1,
np.ones(13) * 2]
np.testing.assert_array_almost_equal(training_data.dx_dt,
expected_derivative)
def test_implicit_get_item():
print("-----test_implicit_get_item-----")
n_lin = int(math.pow(500, 1.0 / 3)) + 1
x_1 = np.linspace(0, 5, n_lin)
x_2 = np.linspace(0, 5, n_lin)
x_3 = np.linspace(0, 5, n_lin)
x = np.array(np.meshgrid(x_1, x_2, x_3)).T.reshape(-1, 3)
x = x[np.random.choice(x.shape[0], 500, replace=False), :]
# make solution
y_t = (x[:, 0] * x[:, 0] + 3.5 * x[:, 1])
y = y_t.reshape(-1, 1)
x = np.hstack((x, y))
py_training_data = ImplicitTrainingData(x)
c_training_data = bingocpp.ImplicitTrainingData(x)
items = [5, 10, 15, 18, 64, 92, 129, 186, 201, 215, 293, 355, 389]
py_result = py_training_data.__getitem__(items)
c_result = c_training_data.__getitem__(items)
assert py_result.x.all() == c_result.x.all()
assert py_result.dx_dt.all() == c_result.dx_dt.all()
def implicit_data_nan(request, dummy_sum_equation, dummy_sum_equation_cpp):
x, y = init_x_and_dx_dt()
x[0, 0] = np.nan
if request.param == "python":
return (SampleTrainingData(x, y), dummy_sum_equation)
return (bingocpp.ImplicitTrainingData(x, y), dummy_sum_equation_cpp)
def dummy_training_data_cpp():
if bingocpp is None:
return None
x, dx_dt = init_x_and_dx_dt()
return bingocpp.ImplicitTrainingData(x, dx_dt)
def compare_cpp_agcpp_implicit(X, Y, operator, params):
"""does the comparison"""
X = np.hstack((X, Y.reshape([-1, 1])))
Y = None
# make solution manipulator
sol_manip = bingocpp.AcyclicGraphManipulator(X.shape[1], 16, nloads=2)
sol_manip.add_node_type(2)
sol_manip.add_node_type(3)
sol_manip.add_node_type(4)
sol_manip.add_node_type(5)
sol_manip.add_node_type(6)
sol_manip.add_node_type(7)
sol_manip.add_node_type(8)
sol_manip.add_node_type(9)
sol_manip.add_node_type(10)
sol_manip.add_node_type(11)
sol_manip.add_node_type(12)
# make true equation
equ = sol_manip.generate()
stack = np.copy(equ.stack)
stack[0] = (0, 0, 0)
stack[1] = (0, 1, 1)
stack[2] = (0, 2, 2)
stack[3] = (operator, params[0], params[1])
stack[-1] = (3, 3, 2)
equ.stack = np.copy(stack)
sol_manip.simplify_stack(equ)
print(stack)
print("equstack\n", equ.stack)
# make predictor manipulator
pred_manip = fpm(32, X.shape[0])
# make training data
training_data = bingocpp.ImplicitTrainingData(X)
# make fitness_metric
explicit_regressor = bingocpp.ImplicitRegression()
# make and run island manager
islmngr = SerialIslandManager(N_ISLANDS,
solution_training_data=training_data,
solution_manipulator=sol_manip,
predictor_manipulator=pred_manip,
fitness_metric=explicit_regressor)
epsilon = 1.05 * islmngr.isles[0].solution_fitness_true(equ) + 1.0e-10
print("EPSILON IS - ", epsilon, equ.latexstring())
converged = islmngr.run_islands(MAX_STEPS,
epsilon,
step_increment=N_STEPS,
make_plots=False)
if not converged:
# try to run again if it fails
islmngr = SerialIslandManager(N_ISLANDS,
solution_training_data=training_data,
solution_manipulator=sol_manip,
predictor_manipulator=pred_manip,
fitness_metric=explicit_regressor)
epsilon = 1.05 * islmngr.isles[0].solution_fitness_true(equ) + 1.0e-10
print("EPSILON IS - ", epsilon, equ.latexstring())
converged = islmngr.run_islands(MAX_STEPS,
epsilon,
step_increment=N_STEPS,
make_plots=False)
def implicit_regression_cpp():
training_data = cppBackend.ImplicitTrainingData(TEST_X_PARTIALS, TEST_DX_DT)
return cppBackend.ImplicitRegression(training_data)
def main(max_steps, epsilon, data_size):
"""main function which runs regression"""
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
# load data on rank 0
if rank == 0:
# make data
# n_lin = int(math.pow(data_size, 1.0/3)) + 1
# x_1 = np.linspace(0, 5, n_lin)
# x_2 = np.linspace(0, 5, n_lin)
# x_3 = np.linspace(0, 5, n_lin)
# x = np.array(np.meshgrid(x_1, x_2, x_3)).T.reshape(-1, 3)
# x = x[np.random.choice(x.shape[0], data_size, replace=False), :]
# make solution
# y = (x[:,0]*x[:,0]+3.5*x[:,1])
# x_true = x
# y_true = y
x = snake_walk()
y = (x[:, 0] + x[:, 1])
x_true = np.hstack((x, y.reshape([-1, 1])))
y_true = None
else:
x_true = None
y_true = None
# then broadcast to all ranks
x_true = MPI.COMM_WORLD.bcast(x_true, root=0)
y_true = MPI.COMM_WORLD.bcast(y_true, root=0)
# make solution manipulator
# sol_manip = agm(x_true.shape[1], 64, nloads=2)
# sol_manip.add_node_type(AGNodes.Add)
# sol_manip.add_node_type(AGNodes.Subtract)
# sol_manip.add_node_type(AGNodes.Multiply)
# sol_manip.add_node_type(AGNodes.Divide)
# sol_manip.add_node_type(AGNodes.Exp)
# sol_manip.add_node_type(AGNodes.Log)
# sol_manip.add_node_type(AGNodes.Sin)
# sol_manip.add_node_type(AGNodes.Cos)
# sol_manip.add_node_type(AGNodes.Abs)
# sol_manip.add_node_type(AGNodes.Sqrt)
# make solution manipulator
# y_true = y_true.reshape(-1, 1)
# sol_manip2 = AGraphCpp.AGraphCppManipulator(x_true.shape[1], 64, nloads=2)
sol_manip2 = bingocpp.AcyclicGraphManipulator(x_true.shape[1], 64, nloads=2)
# sol_manip2 = bingocpp.AcyclicGraphManipulator(x_true.shape[1], 64, nloads=2, opt_rate=0)
# sol_manip.add_node_type(2) # +
# sol_manip.add_node_type(3) # -
# sol_manip.add_node_type(4) # *
# sol_manip.add_node_type(5) # /
# sol_manip.add_node_type(6) # sin
# sol_manip.add_node_type(7) # cos
# sol_manip.add_node_type(8) # exp
# sol_manip.add_node_type(9) # log
# # sol_manip.add_node_type(10) # pow
# sol_manip.add_node_type(11) # abs
# sol_manip.add_node_type(12) # sqrt
sol_manip2.add_node_type(2) # +
sol_manip2.add_node_type(3) # -
sol_manip2.add_node_type(4) # *
sol_manip2.add_node_type(5) # /
sol_manip2.add_node_type(6) # sin
sol_manip2.add_node_type(7) # cos
sol_manip2.add_node_type(8) # exp
sol_manip2.add_node_type(9) # log
# sol_manip2.add_node_type(10) # pow
sol_manip2.add_node_type(11) # abs
sol_manip2.add_node_type(12) # sqrt
# make predictor manipulator
pred_manip = fpm(128, data_size)
# make training data
# training_data = ImplicitTrainingData(x_true)
training_data = bingocpp.ImplicitTrainingData(x_true)
# training_data = ExplicitTrainingData(x_true, y_true)
# training_data2 = bingocpp.ExplicitTrainingData(x_true, y_true)
# make fitness metric
# implicit_regressor = ImplicitRegression()
implicit_regressor = bingocpp.ImplicitRegression()
# explicit_regressor = StandardRegression(const_deriv=True)
# explicit_regressor2 = bingocpp.StandardRegression()
# make and run island manager
islmngr = ParallelIslandManager(#restart_file='test.p',
solution_training_data=training_data,
solution_manipulator=sol_manip2,
predictor_manipulator=pred_manip,
solution_pop_size=64,
fitness_metric=implicit_regressor)
# fitness_metric=explicit_regressor)
# islmngr2 = ParallelIslandManager(#restart_file='test.p',
# solution_training_data=training_data,
# solution_manipulator=sol_manip,
# predictor_manipulator=pred_manip,
# solution_pop_size=64,
# fitness_metric=explicit_regressor)
# islmngr = ParallelIslandManager(#restart_file='test.p',
# data_x=x_true, data_y=y_true,
# solution_manipulator=sol_manip,
# predictor_manipulator=pred_manip,
# solution_pop_size=64,
# fitness_metric=StandardRegression)
# islmngr2 = ParallelIslandManager(#restart_file='test.p',
# data_x=x_true, data_y=y_true,
# solution_manipulator=sol_manip,
# predictor_manipulator=pred_manip,
# solution_pop_size=64,
# fitness_metric=StandardRegression)
non_one = time.time()
islmngr.run_islands(max_steps, epsilon, min_steps=500,
step_increment=500, when_update=50)
non_two = time.time()
non_time = non_two - non_one
timesN.append(non_time)
agesN.append(islmngr.age)
