Beginning Anomaly Detection Using

Figure 7-30.  Outputs the shapes to provide an understanding of how the data sets 

are structured

Figure 7-31.  The shapes of both data sets

[BWUDLQ QSDUUD\ [BWUDLQ UHVKDSH [BWUDLQVKDSH>@

[BWUDLQVKDSH>@

[BWHVW QSDUUD\ [BWHVW UHVKDSH [BWHVWVKDSH>@

[BWHVWVKDSH>@

LQSXWBVKDSH  [BWUDLQVKDSH>@

\BWUDLQ NHUDVXWLOVWRBFDWHJRULFDO \BWUDLQ

\BWHVW NHUDVXWLOVWRBFDWHJRULFDO \BWHVW

Figure 7-32.  Makes the x sets three-dimensional and the y sets two-dimensional 

by reshaping the x sets and changing the y sets to be categorical. The reshaping of 

the x sets is done to fit the input shape of the model

Chapter 7   temporal Convolutional networks

---

276

Let’s take a look at how the operations changed the data sets. Run the code in 

Figure 

7-33

.

The output should look like Figure 

7-34

.

Alright, now both of the data sets have been reshaped successfully. The input shape 

tells the model how many columns and rows to accept per entry. In this case, the input 

shape indicates that there will be 1 row and 31 columns.

Now let’s move on to defining your model. The code chunk in Figure 

7-35

 defines the 

one-dimensional convolutional layers and the dropout layers.

SULQW

6KDSHV?Q[BWUDLQV?Q\BWUDLQV?Q [BWUDLQVKDSH

\BWUDLQVKDSH

SULQW

[BWHVWV?Q\BWHVWV?Q [BWHVWVKDSH\BWHVWVKDSH

SULQW

LQSXWBVKDSH^`?QIRUPDW LQSXWBVKDSH