Beginning Anomaly Detection Using

hyperparameters, 

parameters that are set before the training process.

Let’s create your training and testing data sets. One thing to note is that you can use 

data frames, arrays, matrices, etc. in Keras to serve as your data sets. Run the code in 

Figure 

3-16

.

You can use matplotlib to see what one of these images looks like. Run the code in 

Figure 

3-17

 and see the results in Figure 

3-18

.

EDWFKBVL]H 

QBFODVVHV 

QBHSRFKV 

LPBURZLPBFRO 

Figure 3-15.  Variables to use later

[BWUDLQ\BWUDLQ  [BWHVW\BWHVW  PQLVWORDGBGDWD

Figure 3-16.  Define the training and testing data sets

Chapter 3   IntroduCtIon to deep LearnIng

---

87

You can enter anywhere from 0 to 59,999 to visualize a sample in x_train.

Just looking at 10 examples of the digit 1, you can see there is plenty of variation in 

the data set (see Figure 

3-19

 and Figure 

3-20

).