add new explanations

AirShower_WGAN/ShowerGAN.py

@@ -14,6 +14,7 @@ import tensorflow as tf
 KTF.set_session(utils.get_session())  # Allows 2 jobs per GPU, Please do not change this during the tutorial
 log_dir="."
 
+# basic trainings parameter
 EPOCHS = 10
 GRADIENT_PENALTY_WEIGHT = 10
 BATCH_SIZE = 256
@@ -45,6 +46,7 @@ def build_generator(latent_size):
     generator.add(Conv2D(1, (3, 3), padding='same', kernel_initializer='he_normal', activation='relu'))
     return generator
 
+
 # build critic
 # Feel free to modify the critic model
 def build_critic():
@@ -70,20 +72,19 @@ critic = build_critic()
 print(critic.summary())
 
 # make trainings model for generator
-utils.make_trainable(critic, False)
-utils.make_trainable(generator, True)
+utils.make_trainable(critic, False) # freeze the critic during the generator training
+utils.make_trainable(generator, True) # unfreeze the generator during the generator training
 
 generator_training = utils.build_generator_graph(generator, critic, latent_size)
 generator_training.compile(optimizer=Adam(0.0001, beta_1=0.5, beta_2=0.9, decay=0.0), loss=[utils.wasserstein_loss])
 plot_model(generator_training, to_file=log_dir + '/generator_training.png', show_shapes=True)
 
 # make trainings model for critic
-utils.make_trainable(critic, True)
-utils.make_trainable(generator, False)
-
+utils.make_trainable(critic, True) # unfreeze the critic during the critic training
+utils.make_trainable(generator, False) # freeze the generator during the critic training
 
 critic_training, averaged_batch = utils.build_critic_graph(generator, critic, latent_size, batch_size=BATCH_SIZE)
-gradient_penalty = partial(utils.gradient_penalty_loss, averaged_batch=averaged_batch, penalty_weight=GRADIENT_PENALTY_WEIGHT)
+gradient_penalty = partial(utils.gradient_penalty_loss, averaged_batch=averaged_batch, penalty_weight=GRADIENT_PENALTY_WEIGHT) # construct the gradient penalty
 gradient_penalty.__name__ = 'gradient_penalty'
 critic_training.compile(optimizer=Adam(0.0001, beta_1=0.5, beta_2=0.9, decay=0.0), loss=[utils.wasserstein_loss, utils.wasserstein_loss, gradient_penalty])
 plot_model(critic_training, to_file=log_dir + '/critic_training.png', show_shapes=True)
@@ -104,15 +105,19 @@ for epoch in range(EPOCHS):
     generated_map = generator.predict_on_batch(np.random.randn(BATCH_SIZE, latent_size))
     utils.plot_multiple_signalmaps(generated_map[:,:,:,0], log_dir=log_dir, title='Generated Footprints Epoch: ', epoch=str(epoch))
     for iteration in range(iterations_per_epoch):
+
         for j in range(NCR):
+
             noise_batch = np.random.randn(BATCH_SIZE, latent_size) # generate noise batch for generator
-            shower_batch = shower_maps[BATCH_SIZE*(j+iteration):BATCH_SIZE*(j++iteration+1)]
-            critic_loss.append(critic_training.train_on_batch([noise_batch, shower_batch], [negative_y, positive_y, dummy]))
+            shower_batch = shower_maps[BATCH_SIZE*(j+iteration):BATCH_SIZE*(j++iteration+1)] # take batch of shower maps
+            critic_loss.append(critic_training.train_on_batch([noise_batch, shower_batch], [negative_y, positive_y, dummy])) # train the critic
             print "critic loss:", critic_loss[-1]
+
         noise_batch = np.random.randn(BATCH_SIZE, latent_size) # generate noise batch for generator
-        generator_loss.append(generator_training.train_on_batch([noise_batch], [positive_y]))
+        generator_loss.append(generator_training.train_on_batch([noise_batch], [positive_y])) # train the generator
         print "generator loss:", generator_loss[-1]
 
+# plot critic and generator loss
 utils.plot_loss(critic_loss, name="critic", log_dir=log_dir)
 utils.plot_loss(generator_loss, name="generator",log_dir=log_dir)