Beginning Anomaly Detection Using

semi-supervised 

anomaly detection.

 Encoder-Decoder Temporal Convolutional Network

The version of the encoder-decoder TCN you will be exploring involves a combination 

of one-dimensional causal convolutional and pooling layers to encompass the encoding 

stage and a series of upsampling and one-dimensional causal convolutional layers to 

comprise the decoding stage. The convolutional layers in this model aren’t dilated, but 

they still count as layers of a temporal convolutional network. To better understand the 

structure of this model, take a look at Figure 

7-46

.

Chapter 7   temporal Convolutional networks

---

284

The diagram might seem pretty complicated, so let’s break it down layer by layer.

First, look at the encoding stage and start with the input layer at the very bottom. 

From this layer, you perform a