no code implementations • 14 May 2020 • Oliver Ammann, Gabriel Michau, Olga Fink
We demonstrate that mixing kervolutional with convolutional layers in the encoder is more sensitive to variations in the input data and is able to detect anomalous time series in a better way.