Industrial plants, aircraft, cybersecurity, but also medicine, finance, and
more: anomaly detection systems serve a vital role in numerous contexts,
promptly identifying malfunctions and irregularities without the need for (and
often outperforming) constant expert oversight. There are multiple types of
anomaly detection. In this article, I will demonstrate how to use neural networks
to build a simple anomaly detection system specifically for time series data, i.e., the kind of
data that is mostly generated by sensors.