Beginning Anomaly Detection Using

Figure A-25.  Code to define a custom metric and use that for the model

Figure A-24.  The code definition of categorical accuracy as seen in the Keras 

GitHub page

Of course, there are many more metrics that are available on the Keras 

documentation, and you can even define 

custom metrics. To do that, just simply define 

a function that takes in y_true and y_pred, and call that function name in your metrics, 

as shown in Figure 

A-25

.

In this example, you simply rewrite the binary accuracy metric in several lines and 

return the score. You can actually condense this function to just one line like in the actual 

implementation seen above, but this is just an example to showcase a custom metric.