Beginning Anomaly Detection Using

 Back End (TensorFlow Operations)

You can also perform operations with TensorFlow (if it is the back end) through Keras by 

importing the back end. Below, we will demonstrate some basic functions, but keep in 

mind that TensorFlow has a vast variety of operations and functions.

You can use the back end to create custom layers, metrics, loss functions, etc., 

allowing for a much deeper level of customization. However, you must basically be 

knowledgeable in TensorFlow to accomplish all of this, since this is practically just using 

TensorFlow.

If you want the most customization possible, then using tf.keras along with 

TensorFlow is better, since tf.keras is wholly compatible with all of TensorFlow, and 

you’ll have access to many more TensorFlow commands that you can’t get with just the 

Keras back end.

Here are some of the commands you can execute using the back end (Figure 

A-40

, 

Figure 

A-41

, Figure 

A-42

, Figure 

A-43

).

Appendix A   intro to KerAs

---

359

Figure A-40.  Some TensorFlow operations such as defining placeholders and 

variables done through the Keras back end