Added extra branch to main autoencoder for rule_based prediction

deepof/models.py

@@ -341,11 +341,11 @@ class SEQ_2_SEQ_GMVAE:
             "bidirectional_merge": "concat",
             "clipvalue": 1.0,
             "dense_activation": "relu",
-            "dense_layers_per_branch": 1,
+            "dense_layers_per_branch": 3,
             "dropout_rate": 0.05,
             "learning_rate": 1e-3,
             "units_conv": 64,
-            "units_dense2": 64,
+            "units_dense2": 32,
             "units_lstm": 128,
         }
 
@@ -399,23 +399,15 @@ class SEQ_2_SEQ_GMVAE:
             use_bias=True,
         )
 
-        # Nested list comprehension adding a series of dense and batch norm layers
         Model_E4 = [
-            i
-            for s in [
-                [
-                    Dense(
-                        self.DENSE_2,
-                        activation=self.dense_activation,
-                        # kernel_constraint=UnitNorm(axis=0),
-                        kernel_initializer=he_uniform(),
-                        use_bias=True,
-                    ),
-                    BatchNormalization(),
-                ]
-                for _ in range(self.dense_layers_per_branch)
-            ]
-            for i in s
+            Dense(
+                self.DENSE_2,
+                activation=self.dense_activation,
+                # kernel_constraint=UnitNorm(axis=0),
+                kernel_initializer=he_uniform(),
+                use_bias=True,
+            )
+            for _ in range(self.dense_layers_per_branch)
         ]
 
         # Decoder layers

deepof/train_utils.py

@@ -103,6 +103,7 @@ def get_callbacks(
     run_ID = "{}{}{}{}{}{}{}{}{}{}{}{}".format(
         ("GMVAE" if variational else "AE"),
         ("_input_type={}".format(input_type) if input_type else "coords"),
+        ("_window_size={}", format(X_train.shape[1])),
         ("_NextSeqPred={}".format(next_sequence_prediction) if variational else ""),
         ("_PhenoPred={}".format(phenotype_prediction) if variational else ""),
         ("_RuleBasedPred={}".format(rule_based_prediction) if variational else ""),

deepof_experiments.smk

@@ -23,6 +23,7 @@ entropy_knn = [100]
 next_sequence_pred_weights = [0.0, 0.15]
 phenotype_pred_weights = [0.0]
 rule_based_pred_weights = [0.0, 0.15]
+window_lengths = range(11,56,11)
 input_types = ["coords"]
 run = list(range(1, 11))
 
@@ -47,6 +48,7 @@ rule deepof_experiments:
         expand(
             outpath + "train_models/trained_weights/"
             "GMVAE_input_type={input_type}_"
+            "window_size={window_size}_"
             "NextSeqPred={nspredweight}_"
             "PhenoPred={phenpredweight}_"
             "RuleBasedPred={rulesweight}_"
@@ -58,6 +60,7 @@ rule deepof_experiments:
             "run={run}_"
             "final_weights.h5",
             input_type=input_types,
+            window_size=window_lengths,
             loss=losses,
             encs=encodings,
             k=cluster_numbers,
@@ -121,6 +124,7 @@ rule train_models:
     output:
         trained_models=outpath + "train_models/trained_weights/"
         "GMVAE_input_type={input_type}_"
+        "window_size={window_size}_"
         "NextSeqPred={nspredweight}_"
         "PhenoPred={phenpredweight}_"
         "RuleBasedPred={rulesweight}_"
@@ -149,7 +153,7 @@ rule train_models:
         "--encoding-size {wildcards.encs} "
         "--entropy-knn {wildcards.entknn} "
         "--batch-size 256 "
-        "--window-size 24 "
-        "--window-step 12 "
+        "--window-size {window_size} "
+        "--window-step 11 "
         "--run {wildcards.run} "
         "--output-path {outpath}train_models"