Use best parameters as found by experimentation

train.py

@@ -4,16 +4,19 @@ import torch
 from matplotlib import pyplot
 from mpl_toolkits.axisartist.axislines import SubplotZero
 
-torch.manual_seed(42)
+torch.manual_seed(57)
 
 # Number of symbols to learn
 order = 16
 
+# Initial value for the learning rate
+initial_learning_rate = 0.1
+
 # Number of batches to skip between every loss report
 loss_report_batch_skip = 50
 
 # Size of batches
-batch_size = 32
+batch_size = 2048
 
 # File in which the trained model is saved
 output_file = 'output/constellation-order-{}.pth'.format(order)
@@ -30,8 +33,8 @@ pyplot.show(block=False)
 # Train the model with random data
 model = constellation.ConstellationNet(
     order=order,
-    encoder_layers_sizes=(8, 4),
-    decoder_layers_sizes=(4, 8),
+    encoder_layers_sizes=(8, 4,),
+    decoder_layers_sizes=(4, 8,),
     channel_model=constellation.GaussianChannel()
 )
 
@@ -52,7 +55,7 @@ total_change = float('inf')
 
 # Optimizer settings
 criterion = torch.nn.CrossEntropyLoss()
-optimizer = torch.optim.Adam(model.parameters(), lr=0.1)
+optimizer = torch.optim.Adam(model.parameters(), lr=initial_learning_rate)
 scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
     optimizer, verbose=True,
     factor=0.25,