encoder data

run.sh

@@ -1,9 +1,2 @@
 rm -rf results/ARCADE/*
-python src/config.py cs=topological_0_metric_8
-python src/config.py cs=topological_0_metric_15
-
-python src/config.py cs=hemodynamic_0_metric_8
-python src/config.py cs=hemodynamic_0_metric_15
-
-python src/config.py cs=spatial_0_metric_8
-python src/config.py cs=spatial_0_metric_15
+python src/config.py cs=repr_learning_all_metric_all

src/conf/config.yaml

@@ -2,8 +2,8 @@ defaults:
   - cs: MISSING
   - _self_
 
-sobol_power: 4
-stratify: LAYOUT
+sobol_power: 2
+stratify: 
 quantity_experiment: false
-clean_data: true
+clean_data: false
 debug: true

src/conf/cs/models/mlp.yaml

@@ -6,9 +6,7 @@ mlp:
       search: log
   discrete:
     activation:
-      - identity
       - logistic
-      - tanh
       - relu
     hidden_layer_sizes:
       - (5,)
@@ -17,9 +15,6 @@ mlp:
       - (25,)
       - (25, 25)
       - (25, 50)
-      - (50,)
-      - (50, 25)
-      - (50, 50)
   static:
-    solver: lbfgs
-    max_iter: 1000
+    solver: adam
+    max_iter: 2000

src/conf/cs/models/svr.yaml

@@ -10,7 +10,6 @@ svr:
       search: log
   discrete:
     kernel:
-      - linear
       - poly
       - rbf
       - sigmoid

src/conf/cs/repr_learning_all_metric_all.yaml

@@ -0,0 +1,82 @@
+defaults:
+  - models/mlr
+  - models/rf
+  - models/svr
+  - models/mlp
+  - _self_
+experiments:
+  REPR-LEARNING_32d_0:
+    files:
+      data: full_data.csv
+    paths:
+      log: ${hydra:runtime.cwd}/logs/repr_learning_128/32_0/
+      data: ${hydra:runtime.cwd}/data/repr_learning_128/32_0/
+      results: ${hydra:runtime.cwd}/results/repr_learning_128/32_0/
+  REPR-LEARNING_32d_1:
+    files:
+      data: full_data.csv
+    paths:
+      log: ${hydra:runtime.cwd}/logs/repr_learning_128/32_1/
+      data: ${hydra:runtime.cwd}/data/repr_learning_128/32_1/
+      results: ${hydra:runtime.cwd}/results/repr_learning_128/32_1/
+  REPR-LEARNING_32d_2:
+    files:
+      data: full_data.csv
+    paths:
+      log: ${hydra:runtime.cwd}/logs/repr_learning_128/32_2/
+      data: ${hydra:runtime.cwd}/data/repr_learning_128/32_2/
+      results: ${hydra:runtime.cwd}/results/repr_learning_128/32_2/
+  REPR-LEARNING_32d_3:
+    files:
+      data: full_data.csv
+    paths:
+      log: ${hydra:runtime.cwd}/logs/repr_learning_128/32_3/
+      data: ${hydra:runtime.cwd}/data/repr_learning_128/32_3/
+      results: ${hydra:runtime.cwd}/results/repr_learning_128/32_3/
+  REPR-LEARNING_32d_4:
+    files:
+      data: full_data.csv
+    paths:
+      log: ${hydra:runtime.cwd}/logs/repr_learning_128/32_4/
+      data: ${hydra:runtime.cwd}/data/repr_learning_128/32_4/
+      results: ${hydra:runtime.cwd}/results/repr_learning_128/32_4/
+
+data:
+  features:
+    - dim_0
+    - dim_1
+    - dim_2
+    - dim_3
+    - dim_4
+    - dim_5
+    - dim_6
+    - dim_7
+    - dim_8
+    - dim_9
+    - dim_10
+    - dim_11
+    - dim_12
+    - dim_13
+    - dim_14
+    - dim_15
+    - dim_16
+    - dim_17
+    - dim_18
+    - dim_19
+    - dim_20
+    - dim_21
+    - dim_22
+    - dim_23
+    - dim_24
+    - dim_25
+    - dim_26
+    - dim_27
+    - dim_28
+    - dim_29
+    - dim_30
+    - dim_31
+
+  response:
+    - activity
+    - growth
+    - symmetry

src/permutation/experiments/experiment.py

@@ -183,6 +183,7 @@ def _get_best_cv_model(self) -> None:
 
         for algorithm, runner_list in self._models.items():
             for runner in runner_list:
+                print(runner.model.hparams)
                 runner.cross_validation()
                 progress_counter += 1
                 self.logger.log(f"Progress: {progress_counter} of {self._n_models} - {runner.id}")
@@ -330,7 +331,7 @@ def _run_experiment(self) -> None:
             pass
         self.train_models()
         self.test_models()
-        self.permutation_testing()
+        # self.permutation_testing()
 
     def _log_initialization(self) -> None:
         self.logger.log(f"(Standard Experiment) {self.name} initialized.")

src/permutation/loader.py

@@ -236,6 +236,7 @@ def clean_data(self) -> Tuple[pd.Index, pd.DataFrame, pd.DataFrame]:
             feature for feature in self.features if feature not in removed_feature_columns
         ]
         self._X, self._y = features_response_split(full_data, self.features, self.response)
+
         self._set_working()
         self._split_data(self.stratify)
 

src/permutation/models/sklearnmodel.py

@@ -51,7 +51,6 @@ def _set_model(
         """
         model = cls()
         model.hparams = hparams
-
         pipeline_list = []
 
         if preprocessing_dependencies: