Beginning Anomaly Detection Using

 Model  Creation

In Keras, you can build a 

sequential model, or a functional model.

The 

sequential model is built as shown in Figure 

A-1

.

Figure A-1.  Code defining a sequential model in Keras

Figure A-2.  Code defining a functional model in Keras

Once you’ve defined a sequential model, you can simply add layers to it by calling 

model_name.add(), where the layer itself is the parameter. Once you’ve finished adding 

all of the layers that you want, you are ready to compile and train the model on whatever 

data you have.

Now, let’s look at the 

functional model, the format of which is what you’ve used in 

the book thus far (see Figure 

A-2

).

The functional model allows you to have more flexibility in how you define your 

neural network. With it, you can connect layers to any other layer that you want, instead 

of being limited to just the previous layer like in the sequential model. This allows you 

to share a layer with multiple other layers or even reuse the same layer, allowing you to 

create more complicated models.

Appendix A   intro to KerAs

---

322

Once you’re done defining all of you layers, you simply need to call Model() with 

your input and output parameters respectively to finish your whole model. Now, you can 

continue onwards to compiling and training your model.