ACEMS Research Briefs

These are short summary descriptions of our latest, just-published research.

ACEMS Research Briefs Projects

Anomaly Detection in Streaming Nonstationary Temporal Data

Anomalies can be the main carriers of significant and often critical information, and the identification of anomalies is a key task in many fields such as cyber security, intrusion detection, water quality monitoring, system health monitoring, environmental monitoring. ACEMS researchers, led by Priyanga Dilini Talagala, have proposed a framework that provides early detection of anomalous series within a large collection of streaming time-series data.

Rapid spatial risk modelling for management of early weed invasions: balancing ecological complexity and operational needs

Invasive weeds can threaten native biodiversity and negatively impact on agriculture. A group of researchers at QUT and CSIRO, including an ACEMS member, designed and implemented a new framework for rapidly modelling the invasion risk of weeds using Bayesian belief networks. 

Variance reduction properties of the reparameterization trick

ACEMS researchers have now provided a mathematical treatment that sheds light on the variance reduction properties of the reparameterization trick. The work was presented at the 22nd International Conference on Artificial Intelligence and Statistics (AISTATS) in Okinawa, Japan in April 2019, and was subsequently published in the conference proceedings. AISTATS is considered a top conference in Machine Learning.