The Faculty of Engineering and Mathematical Sciences would like to congratulate computer science and software engineering PhD candidate, Christopher Bartley, whose thesis paper on Data Science and Machine Learning has been accepted to one of the top three international artificial intelligence (AI) conferences in 2019, The Thirty-Third Association for the Advancement of Artificial Intelligence (AAAI) Conference on Artificial Intelligence, held in Honolulu, Hawaii, USA.
The purpose of the AAAI Conference is to promote research in AI and provide a platform for scientific conversations among AI researchers, practitioners, scientists, and engineers in related disciplines. The conference welcomes diverse submissions ranging from student abstracts to posters sessions, invited speakers, tutorials, workshops, and exhibit and competition programs.
Mr Bartley’s thesis paper, “High Accuracy Partially Monotone Ordinal Classification”, was co-supervised by UWA Senior Lecturer Wei Liu and Professor Mark Reynolds, and was a yearlong endeavour from inspiration to acceptance to the AAAI conference.
“AI is an exciting field to be in. I think we are at a point where AI has been commoditised (think K-mart for off-the-shelf algorithms), making it both technically feasible and economically viable to apply in almost every discipline and industry,” Mr Bartley said.
“AI are algorithms, or models that predict things. The usual approach in AI is to let historical data shape (train) these models, ignoring any supposed 'domain knowledge' entirely. Instead of ignoring it, my PhD focused on whether we can find ways to leverage domain knowledge to improve some of the best predictive algorithms in terms of both comprehensibility and accuracy”.
The research project warranted intense mathematical development of the underlying techniques, rigorous proofs and heavy experimental testing on a wide range of large benchmarking datasets.
Professor Reynolds said, “to combine a simple form of general Expert Knowledge to some of the most popular state of the art data-driven Machine Learning techniques, allows the machine to learn from real-world datasets which range across Medicine, Science, and Finance, to make predictions (about new unseen data) which are both more accurate and in line with the expert advice.”
The new hybrid technique is complementary to ‘data-hungry’ deep learning methods, and well suited to smallish datasets which are ubiquitous in real-world applications. “The future application of this technology is very broad,” Professor Reynolds said.
“Getting accepted at AI conferences is brutal. There is so much competition! It took three rejections from other conferences, and subsequent revisions, to get to this point. I'm super happy, and relieved, to have this recognition just before finishing my PhD,” Mr Bartley said.
Mr Bartley will present his thesis paper at the AAAI Conference in Hawaii in late January 2019.
Caitlin White (UWA Faculty of Engineering Mathematical Sciences) (+61 8) 6488 2260