Towards Maintaining and Reusing Complex Event Processing Systems

Abstract Understanding situations formed of patterns of in- terrelated events is a complex problem: often available training data are sparse and either noisy or potentially manipulated by other members of the coalition, if not by an adversary. In previous research we introduced DeepProbCEP, a hybrid neuro- symbolic architecture that leverages both a neural architecture, to interpret raw data, and logical rules, to express patterns defining complex events, while allowing for end-to-end learning. While DeepProbCEP has many advantages, including its ability of adapting to new contexts by leveraging only sparse data, it suffers from several drawbacks, notably the problem of maintaining logical rules. Inductive Logic Programming (ILP) systems are able to learn logical rules from examples. In this demonstration we use an extension of the recent FastLAS ILP system to learn the definitions of complex events from a small number of examples. This method for automatic derivation of the logical rules using ILP overcomes the previous problem of requiring a manual encoding of the complex event definitions.
Authors
  • Marc Vilamala (Cardiff)
  • Mark Law (Imperial)
  • Harrison Taylor (Cardiff)
  • Tianwei Xing (UCLA)
  • Luis Garcia (UCLA)
  • Dave Braines (IBM UK)
  • Daniel Cunnington (IBM UK)
  • Graham White (IBM UK)
  • Alessandra Russo (Imperial)
  • Mani Srivastava (UCLA)
  • Lance Kaplan (ARL)
  • Federico Cerutti (Cardiff)
Date Sep-2020
Venue 4th Annual Fall Meeting of the DAIS ITA, 2020