Dr Roman Marchant
PhD Sydney (2016)
Lecturer
Data Scientist
Centre for Translational Data Science
Faculty of Engineering and Information Technologies
J12 - Computer Science Building
The University of Sydney
| Telephone | +61 2 8627 4344 |
|
|
|
| Website |
Centre for Translational Data Science Sydney Institute of Criminology Google Scholar |
| Curriculum vitae | Curriculum vitae |
Biographical details
Dr Marchant completed a PhD at the School of Information Technologies, University of Sydney in 2015. His current research at the Centre for Translational Data Science explores applying data science to the social sciences, currently focusing on predicting crime and understanding criminal behaviour. His area of expertise is Sequential Bayesian Optimisation (SBO), which is a novel probabilistic method for finding the optimal sequence of decisions that maximise a long-term reward. Although SBO has been readily applied to robotics and environmental monitoring, it can be applied to any optimisation problem.
Dr Marchant has considerable experience in machine learning and data science. Since 2011 he has participated as an active researcher at National ICT Australia (NICTA). Here, he actively contributed in the development of several projects that engage with industry and government agencies:
- Environmental Monitoring for EPA: A platform that continuously learns patterns over space and time for predicting air pollution in the hunter valley region.
- Network Optimisation for the Department of Environment, Water and Natural Resources of South Australia: Optimise a network of sensors that monitor groundwater reservoirs.
- Anomaly Detection for Ecotech: Autonomous system that detects anomalies in environmental sensors.
His PhD thesis, entitled 'Bayesian Optimisation for Planning in Dynamic Environments', proposes the use of a probabilistic framework for finding the optimal sequence of decisions to monitor a time changing phenomena in an efficient manner. During his PhD, he published in top machine learning, robotics and artificial intelligence conferences, including:
- Roman Marchant, Fabio Ramos, and Scott Sanner. Sequential Bayesian Optimisation for Spatial-Temporal Monitoring. In Conference on Uncertainty in Artificial Intelligence (UAI), 2014.
- Roman Marchant and Fabio Ramos. Bayesian Optimisation for Informative Continuous Path Planning. In IEEE International Conference on Robotics and Automation (ICRA), 2014.
- Jefferson Souza, Roman Marchant, Lionel Ott, Denis F. Wolf, and Fabio Ramos. Bayesian Optimisation for Active Perception and Smoot Navigation. In IEEE International Conference on Robotics and Automation (ICRA), 2014.
- Roman Marchant and Fabio Ramos. Bayesian Optimisation for Intelligent Environmental Monitoring. In IEEE Conference on Intelligent Robots and Systems (IROS), 2012.
Throughout his career, Dr Marchant has received several awards, which include Best Student Presentation at the University of Sydney Student Conference 2013 and 2014 and the Google Publication Prize 2013. He was selected by the Chilean Government for a full PhD scholarship in 2011 and then received a Top-up scholarship from NICTA in 2012.
Research interests
There are hidden patterns in large datasets in every domain. Using machine learning algorithms, Dr Roman Marchant is uncovering previously unknown patterns that will help us to understand the key drivers and dynamics of crime, with the ultimate goal of effectively allocating resources for the prevention of criminal behaviour.
"The goal of my research is to extract useful information from large quantities of data. My current focus is on criminology, where statistical models can help us to understand criminal behaviour and determine the key drivers and dynamics of crime. In particular, I am interested in achieving a future reduction in crime arising from data-driven and informed policy decisions.
"Understanding crime as a spatial-temporal phenomenon can help us to determine what are the more dangerous locations at specific times of day. More importantly, it can help us to identify the key drivers of crime. Policy makers will then be able to make more informed decisions, and thus achieve an overall reduction in crime.
"For example, by examining datasets relating to domestic violence assaults, authorities can target specific problems in society that are proven launch pads to domestic violence. Through such direct treatment of specific problems, the number of people affected by domestic violence will diminish.
"Some of the research questions I hope to answer include: What are the key drivers of crime? What types of crime will be predominant in the future? What are the drivers that affect young children and their early involvement with crime? How do life events and other important factors influence the criminality levels of people over time?
"Ultimately I expect to provide evidence that supports the efficient expenditure of resources towards crime reduction."
Teaching and supervision
Current research students
| Project title | Research student |
|---|---|
| Bayesian methods for analysing non-stationary, multivariate time series: applications for machine learning | Nick JAMES |
In the media
https://www.cmo.com.au/article/658978/black-box-algorithms-should-applied-human-outcomes/
Selected grants
2018
- Modelling and Uncertainty Aware Decision Making in the Social Sciences; King G, Marchant Matus R; Office of Global Engagement/Travel Grants.
2017
- Review of the Juvenile Justice NSW Objective Detainee Classification System; Clancey G, Marchant Matus R; NSW Dept of Juvenile Justice/Prequalification Scheme: Performance and Management Services.
Selected publications
|
|
|
|
|
|
|
|
|
|
|
|