U.S. Department of Energy

Pacific Northwest National Laboratory

Autoencoder based anomaly detection for streaming analytics

Publish Date: 
Monday, March 2, 2015
Some of the most pressing machine learning applications such as cyber security and object recognition lack enough ground-truth training data to build a classifier. Rather than build a classifier, our approach is to determine when data is anomalous or deviates from the norm. We demonstrate the use of an autoencoder to both learn a feature space and then identify anomalous portions without the aid of labeled data or domain knowledge. To perform the second step, anomaly detection, we the trained autoencoder to reconstruct new incoming data. A similarity metric between input and reconstruction is used to determine the data subsets that deviate most from the expected from past observations.
Robinson SM, NM Nichols, and RJ Jasper. 2016. "Autoencoder based anomaly detection for streaming analytics." In Autoencoder based anomaly detection for streaming analytics. PNNL-SA-118742, Pacific Northwest National Laboratory, Richland, WA.
| Pacific Northwest National Laboratory