Beginning Anomaly Detection Using

Figure 26.  (continued)

Figure 6-27.  Graph of loss in TensorBoard

Figure 

6-28

 shows the plotting of the mean absolute error during the training process 

through the epochs of training.

Chapter 6   Long Short-term memory modeLS 

---

236

Figure 

6-29

 shows the plotting of the loss of validation during the training process 

through the epochs of training.

Figure 6-28.  Graph of mean absolute error in TensorBoard

Figure 6-29.  Graph of loss of validation in TensorBoard

Chapter 6   Long Short-term memory modeLS 

---

237

Figure 

6-30

 shows the plotting of the mean absolute error of validation during the 

training process through the epochs of training.

Figure 6-30.  Graph of mean absolute error of validation in TensorBoard

Figure 

6-31

 shows the graph of the model as visualized by TensorBoard.

Chapter 6   Long Short-term memory modeLS 

---

238

Once the model is trained, you can predict a test dataset that is split into 

subsequences of the same length (time_steps) as the training datasets. Once this is done, 

you can then compute the root mean square error (RMSE).

Figure 6-31.  Graph of the model as visualized by TensorBoard

Chapter 6   Long Short-term memory modeLS 

---

239

Figure 

6-32

 shows the code to predict on the testing dataset.

RMSE is 0.040, which is quite low, and this is also evident from the low loss from 

the training phase after 20 epochs: loss: 0.0251 - mean_absolute_error: 0.0251 -