def get_y_normalization_scales(dataset, gap_data=False):
_x, _y, train_indices, _, test_indices = load_uci_data(f"{dataset}",
gap_data=gap_data)
y_normalization_scales = []
for i_split in range(len(train_indices)):
_y_train = _y[train_indices[i_split]].reshape(-1, 1)
y_preprocessor = StandardPreprocessor()
y_preprocessor.fit(_y_train)
y_normalization_scales.append(y_preprocessor.scaler.scale_)
return np.array(y_normalization_scales)
from data.toy_regression import cos_linear_edge_data, ground_truth_cos_function
tfd = tfp.distributions
figure_dir = "figures/linear_outer_edges_uncertainty"
figure_dir = Path(figure_dir)
figure_dir.mkdir(parents=True, exist_ok=True)
experiment_name = "linear-outer-edge"
# %%
seed = 1
n_train = 50
# train and test variables beginning with an underscore are unprocessed.
_x_train, y_train = cos_linear_edge_data(n_data=n_train, sigma=0.1, seed=seed)
preprocessor = StandardPreprocessor()
x_train, _x_plot, x_plot = preprocessor.preprocess_create_x_train_x_plot(
_x_train, test_ds=0.8)
y_ground_truth = ground_truth_cos_function(_x_plot)
layer_units = [50, 20] + [2]
layer_activations = ["relu"] * (len(layer_units) - 1) + ["linear"]
# %% codecell
y_lim = [-5, 5]
figsize = (10, 6)
fig, ax = plt.subplots(figsize=figsize)
plot_training_data(_x_train, y_train, fig=fig, ax=ax, y_lim=y_lim)
plot_ground_truth(_x_plot, y_ground_truth, fig=fig, ax=ax, alpha=0.2)
ax.set_xlabel("x")
ax.set_ylabel("y")
hidden_layers_string = (
"two-hidden-layers" if n_hidden_layers == 2 else "one-hidden-layer"
)
save_dir = f".save_uci_models/hmc-map-ensemble-comparison/{hidden_layers_string}/{dataset_name}"
save_dir = Path(save_dir)
save_dir.mkdir(parents=True, exist_ok=True)
# train and test variables beginning with an underscore are unprocessed.
_x, _y, train_indices, _, test_indices = load_uci_data(f"{dataset}")
_x_train = _x[train_indices[data_seed]]
_y_train = _y[train_indices[data_seed]].reshape(-1, 1)
_x_test = _x[test_indices[data_seed]]
_y_test = _y[test_indices[data_seed]].reshape(-1, 1)
x_preprocessor = StandardPreprocessor()
x_train = x_preprocessor.fit_transform(_x_train)
x_test = x_preprocessor.transform(_x_test)
y_preprocessor = StandardPreprocessor()
y_train = y_preprocessor.fit_transform(_y_train)
y_test = y_preprocessor.transform(_y_test)
unnormalized_ll_constant = np.log(y_preprocessor.scaler.scale_)
# %%
with open("config/uci-hyperparameters-config.yaml") as f:
experiment_config = yaml.full_load(f)
train_seed = experiment_config["train_seed"]
layer_units = experiment_config["layer_units"]
layer_activations = experiment_config["layer_activations"]
# %%
data_seed = 0
n_train = 20
# train and test variables beginning with an underscore are unprocessed.
_x_train, _y_train = create_split_periodic_data_heteroscedastic(
n_data=n_train,
lower1=-1,
upper1=0,
lower2=1,
upper2=2,
sigma1=0.2,
sigma2=0.2,
seed=42,
)
preprocessor = StandardPreprocessor()
x_train, _x_plot, x_plot = preprocessor.preprocess_create_x_train_x_plot(
_x_train, test_ds=0.5, n_test=200
)
_y_ground_truth = ground_truth_periodic_function(_x_plot)
y_preprocessor = StandardPreprocessor()
y_train = y_preprocessor.fit_transform(_y_train)
y_ground_truth = y_preprocessor.transform(_y_ground_truth)
# %% markdown
# If we simply used two neurons in the output layer we can model heteroscedastic noise
# %%
input_shape = [1]
layer_units = [20, 10, 2]
layer_activations = ["relu"] * (len(layer_units) - 1) + ["linear"]