def main(argv):
del argv # Unused.
df = io_utils.read_dataframe_from_hdf5(FLAGS.original_df)
## Generate the percentile-normalized embeddings
emb_df_normalized = normalize_ljosa(df)
## Generate the post-factor analysis embeddings
np.random.seed(0)
emb_df_fa = transform.factor_analysis(emb_df_normalized, 0.15, 50)
io_utils.write_dataframe_to_hdf5(emb_df_fa, FLAGS.post_fa_path)
io_utils.write_dataframe_to_hdf5(emb_df_normalized,
FLAGS.post_normliazation_path)
def apply_post_processing(df, percent_norm, factor_analys):
"""Apply percentile normalization and/or factor analysis.
If using WDN, this must be done after the transformation. Thus, this function
is applied in wasserstein_transform.
Args:
df (pandas dataframe): input dataframe
percent_norm (bool): whether to apply percentile normalization
factor_analys (bool): whether to apply factor analysis
Returns:
df (pandas dataframe): post-processed dataframe.
"""
if percent_norm:
df = ljosa_preprocessing.normalize_ljosa(df)
if factor_analys:
np.random.seed(0)
df = transform.factor_analysis(df, 0.15, 50)
return df
def testFactorAnalysisRun(self):
transform.factor_analysis(self.data, 0.1, -1)
