def run(data_filenames, fold_filename, alpha):
fold_filepath = '{!s}/folds/{!s}'.format(dir_path, fold_filename)
layers_config = []
df_train_input, df_train_output, df_test_input, df_test_output = data_rw.loadData(
data_filenames, fold_filepath)
model = getModel(nDInput=df_train_input.shape[1],
nDOutput=df_train_output.shape[1],
alpha=alpha)
keras_util.fitModel(model, df_train_input.as_matrix(),
df_train_output.as_matrix(), layers_config, alpha,
"lasso", dir_path, fold_filename)
# Save test predictions:
test_pred = model.predict(df_test_input.as_matrix())
df_test_pred = pd.DataFrame(data=test_pred,
columns=df_test_output.columns,
index=None)
data_rw.savePreds(df_test_pred, dir_path, fold_filename, layers_config,
alpha, "lasso")
# Save training predictions:
train_pred = model.predict(df_train_input.as_matrix())
df_train_pred = pd.DataFrame(data=train_pred,
columns=df_train_output.columns,
index=None)
data_rw.savePreds(df_train_pred, dir_path, fold_filename, layers_config,
alpha, "train_lasso")
def run(data_filenames, fold_filename, alpha, nDepth, ge_range_all):
fold_filepath = '{!s}/folds/{!s}'.format(dir_path, fold_filename)
df_train_input, df_train_output, df_test_input, df_test_output = data_rw.loadData(data_filenames, fold_filepath)
nM = df_train_input.shape[0]
model = getModel(nDInput = df_train_input.shape[1],
nDOutput = df_train_output.shape[1],
alpha = alpha,
nDepth = nDepth)
if ge_range_all:
model = keras_util.apply_range(model, ge_range_all, df_train_output.columns)
train_input_ext = np.repeat(df_train_input.as_matrix()[:, np.newaxis, :],
nDepth, axis=1)
keras_util.fitModel(model, train_input_ext, df_train_output.as_matrix(), [nDepth], alpha, "biRnn", dir_path, fold_filename)
test_input_ext = np.repeat(df_test_input.as_matrix()[:, np.newaxis, :],
nDepth, axis=1)
# Save test predictions:
test_pred = model.predict(test_input_ext)
df_test_pred = pd.DataFrame(data = test_pred,
columns=df_test_output.columns,
index=None)
data_rw.savePreds(df_test_pred, dir_path, fold_filename, [nDepth], alpha, "biRnn")
# Save training predictions:
train_pred = model.predict(train_input_ext)
df_train_pred = pd.DataFrame(data = train_pred,
columns=df_train_output.columns,
index=None)
data_rw.savePreds(df_train_pred, dir_path, fold_filename, [nDepth], alpha, "train_biRnn")