def reconstruct_data(self, Z_df):
"""
Input: n x (k+1) data frame with ID column and k latent components
Output: n x (d+1) data frame with ID column and data projected into the original (post-processed) space
"""
Z = remove_id_and_get_mat(Z_df)
X = Z.dot(self.U.T)
df = add_id(Z=X, df_with_id=Z_df)
df.columns = ['individual_id'] + self.feature_names
return df
def get_projections(self, df, **projection_kwargs):
"""
use the fitted model to get projections for df.
"""
print("Getting projections using method %s." % self.__class__.__name__)
X = self.data_preprocessing_function(df)
Z = self._get_projections_from_processed_data(X, **projection_kwargs)
Z_df = add_id(
Z, df) # Z_df will have the same index and individual id as df.
Z_df.columns = ['individual_id'
] + ['z%s' % i for i in range(Z.shape[1])]
return Z_df
def get_projections(self, df, project_onto_mean, **projection_kwargs):
"""
use the fitted model to get projections for df.
if project_onto_mean=True, projects onto the mean value of Z (Z_mu). Otherwise, samples Z.
"""
#print("Getting projections using method %s." % self.__class__.__name__)
X = self.data_preprocessing_function(df)
ages = self.get_ages(df)
Z = self._get_projections_from_processed_data(X, ages, project_onto_mean, **projection_kwargs)
Z_df = add_id(Z, df) # Z_df and df will have the same id and individual_id.
Z_df.columns = ['individual_id'] + ['z%s' % i for i in range(Z.shape[1])]
return Z_df
