Research Publications
Kelvin Hsu
Bayesian Perspectives on Conditional Kernel Mean Embeddings: Hyperparameter Learning and Probabilistic Inference
PhD Thesis
Kelvin Hsu and Fabio Ramos
Bayesian Deconditional Kernel Mean Embeddings
In International Conference on Machine Learning (ICML) 2019
BibTeX
Kelvin Hsu and Fabio Ramos
Bayesian Learning of Conditional Kernel Mean Embeddings for Automatic Likelihood-Free Inference
In International Conference on Artificial Intelligence and Statistics (AISTATS) 2019
BibTeX
Kelvin Hsu, Richard Nock, and Fabio Ramos
Hyperparameter Learning for Conditional Kernel Mean Embeddings with Rademacher Complexity Bounds
In European Conference on Machine Learning (ECML) 2018
Best Student Paper Award Winner
BibTeX
Kelvin Hsu, Simon O’Callaghan, Alistar Reid, and Stefan Williams
Informative seafloor exploration using the linearised differential entropy of Gaussian process classifiers
In Australasian Conference on Robotics and Automation (ACRA) 2015
Teaching Experience
Postgraduate Level Courses
COMP5318 (Machine Learning and Data Mining), Postgraduate, 2017 - 2018
QBUS6810 (Statistical Learning and Data Mining), Postgraduate, 2018 - 2019
BUSS6002 (Data Science in Business), Postgraduate, 2017 - 2019
QBUS6820 (Business Risk Management), Postgraduate, 2017 - 2018
Undergraduate Level Courses
AERO4701 (Space Engineering 3), Capstone Undergraduate, 2016
QBUS3820 (Machine Learning and Data Mining), Final Year Undergraduate, 2017
AMME3500 (System Dynamics and Control), Third Year Undergraduate, 2016
STAT1005 (Statistics), First Year Undergraduate, 2017
ENGG1801 (Engineering Computing), First Year Undergraduate, 2013 - 2018
Others
Academic Record: Machine Learning PhD | University Medal | Advanced Mathematics | Transcript
References: Dr. Simon O’Callaghan | Dr. Matthew Whiting | Dr. Jason Chan | Dr. Alistar McEwan
"During his time at NICTA, he demonstrated an exemplary work ethic coupled with an innate ability to think critically about complex theoretical issues. Kelvin has always expressed a strong desire to develop a deep understanding of the machine learning techniques that we use daily and, as a result, has been able to effectively leverage their capabilities to solve real world scientific problems.
Kelvin designed and coded an active sampling algorithm for guiding autonomous underwater vehicles during habitat mapping expeditions. I was struck by Kelvin’s ability to operate under minimal guidance; suggestions that I would make to him would quickly become an implemented reality and iterated upon within days. His work has resulted in a paper getting accepted to the Australasian Conference on Robotics and Automation 2015.
Kelvin’s communication skills are a rarity amongst undergraduates. On several occasions during his internship and honours work, he has been required to present his progress to others both in the form of written reports and oral presentations. My colleagues and I are frequently impressed with his talent to distil complicated methodologies into concise points and make them accessible to a variety of audiences."
"Kelvin has built a very impressive graphical interface that allows a user to examine the results of a Duchamp search. This will help astronomers make better sense of their search result. This was a complex task, requiring understanding of the various data structures and the ability to put together different elements of python programming.
I was also very impressed with Kelvin's presentation skills. Kelvin's presentation was, in my opinion, one of the best. He was very confident, and did an excellent job at presenting something that can be difficult."
"Many students cited that he was very generous and patient, and that they felt comfortable due to his friendly nature. He would get them to think and guide them towards the solution without actually giving them the answers, showing Kelvin's commitment to quality work. Students were able to understand complicated concepts easily because of Kelvin’s very clear explanations and presentation style. Kelvin’s overwhelmingly positive attitude and enthusiasm for learning is clearly felt by students.
In the UoS surveys (run by the School of IT and ITL) since 2013 when Kelvin began tutoring, Kelvin has always had a very high quantity of strong and enthusiastically positive comments, without a single negative comment at all. Even amongst the excellent standard of tutors, this suggests to me that Kelvin is one of the best tutors of all time, even better than previous tutoring award winners."