Merge pull request #190 from david-thrower/180-question-leakage-potentially-causing-the-extraordinarily-low-rmse-for-ames-dataset

180 question leakage potentially causing the extraordinarily low rmse for ames dataset

Author: Aidyn-Lopez

.github/workflows/automerge.yml

@@ -5,7 +5,7 @@ name: Python application
 
 on:
   push:
-    branches: [ "main", "187-upgrade-transformers" ]
+    branches: [ "main", "180-question-leakage-potentially-causing-the-extraordinarily-low-rmse-for-ames-dataset" ]
 
 permissions:
   contents: read

realnn-regression-example-ames-no-preproc.py

@@ -16,6 +16,8 @@
 
 ###
 
+LABEL_COLUMN = 'price'
+
 ## your data:
 
 
@@ -30,9 +32,11 @@
 
 raw_data = pd.read_csv('ames.csv')
 needed_cols = [
-    col for col in raw_data.columns if raw_data[col].dtype != 'object']
+    col for col in raw_data.columns 
+    if raw_data[col].dtype != 'object' 
+    and col != LABEL_COLUMN]
 data_numeric = raw_data[needed_cols].fillna(0).astype(float)
-label = raw_data.pop('price')
+label = raw_data.pop(LABEL_COLUMN)
 
 data_np = data_numeric.values
 

regression-example-ames-no-preproc-val-set.py

@@ -16,7 +16,7 @@
 
 ###
 
-## your data:
+LABEL_COLUMN = 'price'
 
 
 TIME = pendulum.now().__str__()[:16]\
@@ -103,11 +103,14 @@ def hash_based_split(df,  # Pandas dataframe
 
 # white = pd.read_csv('wine_data.csv')
 
+
 raw_data = pd.read_csv('ames.csv')
 needed_cols = [
-    col for col in raw_data.columns if raw_data[col].dtype != 'object']
+    col for col in raw_data.columns 
+    if raw_data[col].dtype != 'object' 
+    and col != LABEL_COLUMN]
 data_numeric = raw_data[needed_cols].fillna(0).astype(float)
-label = raw_data.pop('price')
+label = raw_data.pop(LABEL_COLUMN)
 
 
 train_df, train_labels_pd, val_df, val_labels_pd =\
@@ -147,19 +150,20 @@ def hash_based_split(df,  # Pandas dataframe
 # Params for a training function (Approximately the oprma
 # discovered in a bayesian tuning study done on Katib)
 
+
 meta_trial_number = 0  # In distributed training set this to a random number
-activation = "gelu"
-predecessor_level_connection_affinity_factor_first = 19.613
-predecessor_level_connection_affinity_factor_main = 0.5518
-max_consecutive_lateral_connections = 34
-p_lateral_connection = 0.36014
-num_lateral_connection_tries_per_unit = 11
-learning_rate = 0.095
-epochs = 145
-batch_size = 634
-maximum_levels = 5
-maximum_units_per_level = 5
-maximum_neurons_per_unit = 25
+activation = 'swish'
+predecessor_level_connection_affinity_factor_first = 0.506486683067576
+predecessor_level_connection_affinity_factor_main = 1.6466748663373876
+max_consecutive_lateral_connections = 35
+p_lateral_connection = 3.703218275217572
+num_lateral_connection_tries_per_unit = 12
+learning_rate = 0.02804912925494706
+epochs = 130
+batch_size = 78
+maximum_levels = 4
+maximum_units_per_level = 3
+maximum_neurons_per_unit = 3
 
 
 cerebros =\
@@ -172,11 +176,11 @@ def hash_based_split(df,  # Pandas dataframe
         validation_split=0.0,
         direction='minimize',
         metric_to_rank_by='val_root_mean_squared_error',
-        minimum_levels=1,
+        minimum_levels=4,
         maximum_levels=maximum_levels,
-        minimum_units_per_level=1,
+        minimum_units_per_level=2,
         maximum_units_per_level=maximum_units_per_level,
-        minimum_neurons_per_unit=1,
+        minimum_neurons_per_unit=3,
         maximum_neurons_per_unit=maximum_neurons_per_unit,
         validation_data=(val_x, val_labels),
         activation=activation,

regression-example-ames-no-preproc.py

@@ -16,6 +16,8 @@
 
 ###
 
+LABEL_COLUMN = 'price'
+
 ## your data:
 
 
@@ -30,9 +32,11 @@
 
 raw_data = pd.read_csv('ames.csv')
 needed_cols = [
-    col for col in raw_data.columns if raw_data[col].dtype != 'object']
+    col for col in raw_data.columns 
+    if raw_data[col].dtype != 'object' 
+    and col != LABEL_COLUMN]
 data_numeric = raw_data[needed_cols].fillna(0).astype(float)
-label = raw_data.pop('price')
+label = raw_data.pop(LABEL_COLUMN)
 
 data_np = data_numeric.values
 
@@ -52,18 +56,18 @@
 # discovered in a bayesian tuning study done on Katib)
 
 meta_trial_number = 0  # In distributed training set this to a random number
-activation = "gelu"
-predecessor_level_connection_affinity_factor_first = 19.613
-predecessor_level_connection_affinity_factor_main = 0.5518
-max_consecutive_lateral_connections = 34
-p_lateral_connection = 0.36014
-num_lateral_connection_tries_per_unit = 11
-learning_rate = 0.095
-epochs = 145
-batch_size = 634
-maximum_levels = 5
-maximum_units_per_level = 5
-maximum_neurons_per_unit = 25
+activation = 'swish'
+predecessor_level_connection_affinity_factor_first = 0.506486683067576
+predecessor_level_connection_affinity_factor_main = 1.6466748663373876
+max_consecutive_lateral_connections = 35
+p_lateral_connection = 3.703218275217572
+num_lateral_connection_tries_per_unit = 12
+learning_rate = 0.02804912925494706
+epochs = 130
+batch_size = 78
+maximum_levels = 4
+maximum_units_per_level = 3
+maximum_neurons_per_unit = 3
 
 
 cerebros =\
@@ -76,11 +80,11 @@
         validation_split=0.35,
         direction='minimize',
         metric_to_rank_by='val_root_mean_squared_error',
-        minimum_levels=1,
+        minimum_levels=4,
         maximum_levels=maximum_levels,
-        minimum_units_per_level=1,
+        minimum_units_per_level=2,
         maximum_units_per_level=maximum_units_per_level,
-        minimum_neurons_per_unit=1,
+        minimum_neurons_per_unit=3,
         maximum_neurons_per_unit=maximum_neurons_per_unit,
         activation=activation,
         final_activation=None,