Skip to content



Our enhanced lower bound for Dynamic Time Warping is tighter than the widely used LB Keogh without any additional computational burden. This picture shows Bart Goethals as he realises the impact it is bound to have. For more details see our SDM-19 paper here.


I'm delighted to see Wilhelmiina Hamelainen and my tutorial published. It explains how robust statistical techniques can overcome many of the numerous shortcomings of standard approaches to association and pattern mining. Read it here or see my work in the area here.


Thank you @AusDMConf and @CharlesStuartUniversity for the opportunity to present my group’s work on Concept Drift Thanks also for the lovely CSU wines. @MonashUni @@MonashInfoTech


We introduce a novel incremental decision tree learning algorithm, Hoeffding Anytime Tree, that is statistically more efficient than the current state-of-the-art, Hoeffding Tree. We demonstrate that an implementation of Hoeffding Anytime Tree–-"Extremely Fast Decision Tree", a minor modification to the MOA implementation of Hoeffding Tree–-obtains significantly superior prequential accuracy on most of the largest classification datasets from the UCI repository. Hoeffding Anytime Tree produces the asymptotic batch tree in the limit, is naturally resilient to concept drift, and can be used as a higher accuracy replacement for Hoeffding Tree in most scenarios, at a small additional computational cost.

Download the paper here.

Download the software here.



The Living with AI discussion series will feature conversations about social and ethical implications of artificial intelligence. Join us for the first of these conversations, featuring AI expert Terry Cailli addressing the question "Can Computers Behave Ethically?"

Date and Time

Wed. 8 August 2018
5:30 pm – 7:00 pm AEST


Deakin Downtown
Collins Square, Tower 2
Level 12, 727 Collins Street
Docklands, VIC 3008


Click here.


We are very honoured to have received the 2018 SIAM International Conference on Data Mining (SDM-18) Best Research Paper Award for

Efficient search of the best warping window for Dynamic Time Warping.
Tan, C. W., Herrmann, M., Forestier, G., Webb, G. I., & Petitjean, F.
Proceedings of the 2018 SIAM International Conference on Data Mining, 2018.

This paper presents algorithms that allow a critical hyper-parameter for time series classification, the warping window, to be efficiently tuned.

The paper can be downladed here:

The code can be downloaded here:

Fantastic planning session at the EODC. Amazing problems for time series analysis, starting with how to scale the state-of-the-art from being challenged by datasets containing 10 thousand series to routinely handling 10 trillion. Improved analytics for earth observation data have massive expected benefits for environmental monitoring, agriculture and economic planning, to name but a few. As always, it is both scary and exciting to be at the start of a massive undertaking.

With EODC Head of Science Wolfgang Wagner and my Monash earth observation colleague Christoph Rüdiger.

With EODC Managing Director Christian Briese and Software Developer Thomas Mistelbauer.