Anomaly detection and system regulation
When is a system behaving normally?
What does normal mean? Is it something we decide according to
policy, or is it a matter of empiricism?
At Oslo University College, we are examining these questions by building policy
based control systems along side machine-learning agents. The agents
are able to integrate what they learn about their environments and their
own previous behaviour into a consistent picture of the world. This is
used to motivate decision making during automation.
Our test implementation of the ideas is contained in
cfengine's cfenvd component. |
