Speaker
Steffen Seitz
(TU Dresden)
Description
Industrial machines tend to have a variety of errors and in general there is no dataset on every error available, since you can‘t run a „good to failure“ experiment at one‘s convenience. In the talk we will show how to use machine learning techniques like Autoencoders to infer information of the hidden condition into latent variables to monitor these industrial processes. Therefore we cover Variational Inference as a Neural Networks training method and the application onto time series data.
