Unlocking time: machine learning for understanding dynamic processes

The world is dynamic and in a constant state of flux, yet most machine learning models learn static models from a dataset that represents a single snapshot in time. My group's research is revolutionising the field of temporal analytics. We have refocused the field on methods that are both effective and feasible for non-trivial problems. We received a prestigious best paper award at the SDM data science conference [1]; one of our papers is recognised as Clarivate Web of Science HighCite (top 1% of papers for the field of research) [2]; three of our algorithms (TS-Chief, InceptionTime and Rocket) have been independently identified as three of four algorithms that define the current state of the art [3]; our papers on these three algorithms [4-6] are the three most cited papers published in 2020 in the primary journal for time series classification research, Data Mining and Knowledge Discovery [7]; and our algorithm Rocket has been independently assessed as ‘the most important recent development in the field’ [8]. As part of a large, active Australian Research Council funded group, this project will build on these foundations to develop new technologies for supervised machine learning from temporal data. See my time series research for more details of the research program on which this research will build.