Beginning Anomaly Detection Using

 Model Compilation and Training

In most cases, the code to compile your model will look something like Figure 

A-3

.

Figure A-3.  Code to compile a model in Keras

However, there are many more parameters to consider:

• 

optimizer: Passes in the name of the optimizer in the string or an 

instance of the optimizer (you call the optimizer with whatever 

parameters you like. We will elaborate on this further below in the 

Optimizers section.)

• 

loss: Passes in the name of the loss function or the function itself.  

We elaborate on what we mean by this below in the 

Losses section.

• 

metrics: Passes in the list of metrics that you want the model to 

evaluate during the training and testing processes. Check out the 

Metrics section for more details on what metrics you can use.

• 

loss_weights: If you have multiple outputs and multiple losses, the 

model evaluates based on the total loss. The loss_weights are a list 

or dictionary that determines how much each loss factors into the 

overall, combined loss. With the new weights, the overall loss is now 

the weighted sum of all losses.

•