def make_experiments(system, dataset):
x = dataset[["x1", "x2"]].values
funcs = {}
for col, depvar in zip(("log_conductivity", "solution"), ("log_a", "u")):
y = dataset[col].values
gp = GP(x, y[:, np.newaxis])
gp.train(num_epochs=1000, learning_rate=0.1)
gp.predict_mode()
# Extract function
# Sorry for the a-kludge :(
f = Function(gp.model.get_mean_function())
if depvar == "log_a":
# Convert to normal conductivity
# TODO make this remain a random function
f = f.exp()
depvar = "a"
funcs[depvar] = f
experiments = [
Experiment({
"u": funcs["u"],
"a": funcs["a"]
}, dataset[["x1", "x2"]])
]
fitness_threshold = ground_truth_fitness(experiments, ground_truth_model)
return experiments, fitness_threshold
def make_experiments(system, dataset):
x = dataset["x"].values
funcs = {}
for col, depvar in zip(("conductivity", "solution", "force"),
("a", "u", "f")):
y = dataset[col].values
gp = GP(x[:, np.newaxis], y[:, np.newaxis])
gp.train(num_epochs=1000, learning_rate=0.1) # , show_loss=True)
gp.predict_mode()
funcs[depvar] = Function(gp.model.get_mean_function())
experiments = [
Experiment(
{
"u": funcs["u"],
"a": funcs["a"]
},
pd.DataFrame({"x": x}),
inhomogeneous=funcs["f"],
)
]
fitness_threshold = ground_truth_fitness(experiments, ground_truth_model)
return experiments, fitness_threshold
def make_experiments(system, dataset):
t = dataset["t"].values
lhs = {}
x = dataset["x"].values
gp = GP(t[:, np.newaxis], x[:, np.newaxis])
gp.train(num_epochs=1000, learning_rate=0.1) # , show_loss=True)
gp.predict_mode()
# Extract function for modeled dataset
fx = Function(gp.model.get_mean_function())
experiments = [Experiment({"x": fx}, pd.DataFrame({"t": t}))]
# Extras
fitness_threshold = ground_truth_fitness(experiments, ground_truth_model)
return experiments, fitness_threshold