We are looking for a PhD student with an interest in experimental methodology and Bayesian machine learning.
Active learning has originated as a subfield of machine learning that is used when data is expensive or difficult to obtain: Rather than randomly labelling all data, an algorithm chooses informative data points about which it is uncertain and requests a label from a human. This approach has been applied to healthcare in some clinical tests. For example, in the audiogram, a test to detect the softest tones a patient can hear, a Bayesian active-learning algorithm led to a more detailed estimate of the threshold in less time. The latter is particularly important for more complex tests, where test duration was shortened from several hours to 30 minutes, making them applicable in clinical practice.
The active-learning approaches of those tests were hand-tailored and determined in extensive research. For the present project we are looking for a PhD student who will design, implement and evaluate an active-learning framework that can be applied by clinicians, scientists and other practitioners for a wide range of tests and experiments without further need of laborious fine tuning. This has the potential to transform clinical practice and scientific methods by enabling more precise results in a given time. For example, perceptual tests could be applied more easily to study or diagnose autism.
We are looking for a student with a background in psychology, engineering, computer science, physics, or similar, and an interest in psychological methods, perception, mathematical psychology, probabilistic machine learning and information theory. Prior experience in some of these areas is expected but the opportunity to learn will be given. Experience in electrophysiology is welcome but not necessary. The department provides a vibrant environment with about twenty PhD students and ten early-career researchers. Activities like seminars or journal clubs provide an opportunity to gain further knowledge in related areas. Source: https://ucl-epsrc-dtp.github.io/2023-24-project-catalogue/projects/2228bd1227.html
The project is competitively funded, closing deadline 26 January 2023. A successful candidate gets tuition fees covered, an additional stipend of at least £20,668 per year and a research support grant.
Applications must be made before the deadline via this webpage: https://www.ucl.ac.uk/epsrc-doctoral-training/prospective-students/apply-ucl-epsrc-dtp-studentship
which also provides detailed information about the programme and application.
For questions, please contact Josef Schlittenlacher under firstname.lastname@example.org