An autoencoder is a neural network that predicts its own input. The diagram in Figure 3 shows the architecture of the 65-32-8-32-65 autoencoder used in the demo program. An input image x, with 65 ...

The demo program presented in this article uses image data, but the autoencoder anomaly detection technique can work with any type of data. The demo begins by creating a Dataset object that stores the ...