fix bugs

Author: chenyangkang

stemflow/model/AdaSTEM.py

@@ -485,18 +485,28 @@ def SAC_ensemble_training(self, index_df: pd.core.frame.DataFrame, data: pd.core
             bootstrap_random_state = index_df['bootstrap_random_state'].iloc[0]
             rng = np.random.default_rng(bootstrap_random_state)  # NumPy's random generator
             bootstrap_indices = rng.choice(data.index, size=len(data), replace=True)  # Full bootstrap sample
+        else:
+            bootstrap_indices = None # Place holder
 
         res_list = []
         for start in unique_start_indices:
-            valid_index_window_data_df = data.index[
-                (data[self.Temporal1] >= start) & (data[self.Temporal1] < start + self.temporal_bin_interval)
-            ]
-            window_data_df_index = bootstrap_indices[np.isin(bootstrap_indices, valid_index_window_data_df)]
-            window_data_df = data.loc[window_data_df_index] # So that we don't need to make a whole copy of the data
+            
+            if self.ensemble_bootstrap:
+                valid_index_window_data_df = data.index[
+                    (data[self.Temporal1] >= start) & (data[self.Temporal1] < start + self.temporal_bin_interval)
+                ]
+                window_data_df_index = bootstrap_indices[np.isin(bootstrap_indices, valid_index_window_data_df)]
+                window_data_df = data.loc[window_data_df_index] # So that we don't need to make a whole copy of the data
+                del window_data_df_index, valid_index_window_data_df
 
+            else:
+                window_data_df = data[
+                    (data[self.Temporal1] >= start) & (data[self.Temporal1] < start + self.temporal_bin_interval)
+                ]
+            
             window_data_df = transform_pred_set_to_STEM_quad(self.Spatio1, self.Spatio2, window_data_df, index_df)
             window_index_df = index_df[index_df[f"{self.Temporal1}_start"] == start]
-
+            
             # Merge
             def find_belonged_points(df, df_a):
                 return df_a[

stemflow/utils/quadtree.py

@@ -191,15 +191,21 @@ def get_one_ensemble_quadtree(
         bootstrap_random_state = rng.integers(1e9)
         rng = np.random.default_rng(bootstrap_random_state)  # NumPy's random generator
         bootstrap_indices = rng.choice(data.index, size=len(data), replace=True)  # Full bootstrap sample
+    else:
+        bootstrap_indices = None # Place holder
 
     ensemble_all_df_list = []
 
     for time_block_index, bin_ in enumerate(temporal_bins):
         time_start = bin_[0]
         time_end = bin_[1]
-        valid_index_sub_data = data.index[(data[Temporal1] >= time_start) & (data[Temporal1] < time_end)]
-        sub_data_index = bootstrap_indices[np.isin(bootstrap_indices, valid_index_sub_data)]
-        sub_data = data.loc[sub_data_index] # So that we don't need to make a whole copy of the data
+        
+        if ensemble_bootstrap:
+            valid_index_sub_data = data.index[(data[Temporal1] >= time_start) & (data[Temporal1] < time_end)]
+            sub_data_index = bootstrap_indices[np.isin(bootstrap_indices, valid_index_sub_data)]
+            sub_data = data.loc[sub_data_index] # So that we don't need to make a whole copy of the data
+        else:
+            sub_data = data[(data[Temporal1] >= time_start) & (data[Temporal1] < time_end)]
 
         if len(sub_data) == 0:
             continue