I'm a data scientist and software developer based in Dunedin, New Zealand. Prior to working as a developer with ADInstruments, I was a neuroscience researcher. This is my personal GitHub profile but you can find links to my external projects below too.
Brain Imaging | Data Science | Machine Learning | Software Engineering
- LinkedIn: https://www.linkedin.com/in/samharrison5
- ORCID: https://orcid.org/0000-0002-5886-2389
- Google Scholar: https://scholar.google.com/citations?user=SpoKPkgAAAAJ
PROFUMO: Inference of subject-specific brain netorks (PRObabilistic FUnctional MOdes) from resting-state functional MRI data.
- Role: developed the model and implemented an efficient framework for variational Bayesian inference of matrix-factorisation models.
- Based at the Wellcome Centre for Integrative Neuroimaging, University of Oxford.
- Code: https://git.fmrib.ox.ac.uk/profumo/profumo
- Papers:
PhysIO: a novel Hilbert-based method for improved processing of respiratory signals as part of the PhysIO functional MRI denoising toolbox.
- Role: Devised and implemented a new signal processing approach for physiological data, based on a novel application of ideas from EEG/MEG research.
- Based at the Translational Neuromodeling Unit, ETH Zurich.
- Code: https://github.com/translationalneuromodeling/tapas/tree/master/PhysIO
- Paper: https://doi.org/10.1016/j.neuroimage.2021.117787
pd-apathy: Analysis of the predictors of apathy for patients with Parkinson's disease.
- Role: Formulated the longitudinal modelling approach and performed the analysis.
- Based at the New Zealand Brain Research Institute, University of Otago.
- Code & results: https://github.com/nzbri/pd-apathy
- Paper: https://doi.org/10.1093/brain/awad113
solar_analyses: Analysis of the output from a home solar system.
- Role: Developed the probabilistic model and implemented the sampling scheme to do Bayesian inference using Stan.
- Personal project.
- Code & results: https://github.com/sharrison5/solar_analyses
genbed: toolbox for classification and visualisation, designed for computational psychiatry workflows.
- Role: Put together the functionality needed to provide a sensible default workflow for classification, with a focus on visualisation and interpretation of the outputs.
- Based at the Translational Neuromodeling Unit, ETH Zurich.
- Code: https://github.com/translationalneuromodeling/tapas/tree/master/genbed
- Paper: https://doi.org/10.3389/fpsyt.2021.680811
Pyro: Tutorial for performing boosting black-box variational inference in Pyro.
- Role: co-supervised a student project based on adapting a novel variational inference technique to new deep learning frameworks.
- Based at the Translational Neuromodeling Unit, ETH Zurich.
- Code: pyro-ppl/pyro#2308
- Tutorial: http://pyro.ai/examples/boosting_bbvi.html