More work on autoencoder.

Also, use groupyr 0.3.3

afqinsight/nn/tf_models.py

@@ -336,21 +336,25 @@ def fc_autoencoder(input_shape, encoding_dim=None, verbose=False):
     return model
 
 
-def cnn_autoencoder(input_shape, verbose=False):
+def cnn_autoencoder(input_shape, encoding_dim=8, verbose=False):
     """
     Convolutional autoencoder
     """
     ip = Input(shape=input_shape)
     # Encoder
-    x = Conv1D(32, (3), activation="relu", padding="same")(ip)
-    x = MaxPooling1D((2), padding="same")(x)
-    x = Conv1D(32, (3), activation="relu", padding="same")(x)
-    x = MaxPooling1D((2), padding="same")(x)
-
+    x = Conv1D(32, 3, activation="relu", padding="same")(ip)
+    x = MaxPooling1D(2, padding="same")(x)
+    x = Conv1D(16, 3, activation="relu", padding="same")(x)
+    x = MaxPooling1D(2, padding="same")(x)
+    shape = x.shape
+    # Latent
+    x = Flatten()(x)
+    x = Dense(encoding_dim, activation="relu")(x)
     # Decoder
-    x = Conv1DTranspose(32, (3), strides=2, activation="relu", padding="same")(x)
-    x = Conv1DTranspose(32, (3), strides=2, activation="relu", padding="same")(x)
-    x = Conv1D(1, (3), activation="sigmoid", padding="same")(x)
+    x = Reshape(shape)(x)
+    x = Conv1DTranspose(32, 3, strides=2, activation="relu", padding="same")(x)
+    x = Conv1DTranspose(16, 3, strides=2, activation="relu", padding="same")(x)
+    x = Conv1DTranspose(1, 3, activation="sigmoid", padding="same")(x)
 
     model = Model([ip], [x])
     if verbose:
@@ -401,7 +405,8 @@ def _fc_vae_decoder(input_shape, encoding_dim=None, verbose=False):
     fc = Dense((input_shape[0] * input_shape[1]) // 4, activation="relu")(fc)
     fc = Dense((input_shape[0] * input_shape[1]) // 2, activation="relu")(fc)
     pre_out = Dense((input_shape[0] * input_shape[1]))(fc)
-    return Reshape(input_shape)(pre_out)
+    out = Reshape(input_shape)(pre_out)
+    return Model([ip], [out], name="decoder")
 
 
 class _VAE(Model):
@@ -427,6 +432,11 @@ def metrics(self):
             self.kl_loss_tracker,
         ]
 
+    def call(self, inputs):
+        z_mean, z_log_var, z = self.encoder(inputs)
+        reconstructed = self.decoder(z)
+        return reconstructed
+
     def train_step(self, data):
         with tf.GradientTape() as tape:
             z_mean, z_log_var, z = self.encoder(data)