牧野 南山 Pronounced like “moo-yeh nahn-shahn” IPA: [mu˥˩ jɛ˨˩˦ nan˧˥ ʂan˥]
PhD Candidate in Statistics
Simon Fraser University · Burnaby, Canada
I am a PhD candidate in Statistics at Simon Fraser University. My research focuses on statistical methodology for longitudinal and high-dimensional data, particularly when observations are sparse, irregular, and structurally complex.
I am especially interested in functional data analysis, neuroimaging, survival analysis, and statistical machine learning. I aim to develop methods that are both theoretically principled and useful for scientific applications.
Functional data · Online learning
Online functional principal component analysis for streaming multidimensional data. We combine spline representations with Riemannian stochastic gradient descent to learn smooth, orthonormal components, and use beam search with rolling validation to tune smoothing as new data arrive.
Scientific computing · Medical imaging
A research codebase for simulating and reconstructing glioma progression on patient-specific 3D brain domains. A reaction–diffusion model describes tumor infiltration; sensitivity equations support calibration of growth, diffusion, and tumor origin from single or multiple scans.
Survival analysis · Neuroimaging
Our SMMR work uses a regularized mixture cure-rate model to distinguish susceptibility to Alzheimer’s conversion from time to conversion, with localized MRI effects estimated over irregular brain domains.
In ongoing work, we jointly model longitudinal images and interval-censored event times through functional principal components linked to a Cox model, enabling efficient estimation and inference for the overall image effect.
Dynamical systems · Sparse estimation
Two papers on learning sparse dynamical systems from noisy, non-Gaussian observations. We develop estimation and structure-selection methods for high-dimensional linear ODEs and a joint estimation approach for sparse additive ODEs, with applications including yeast gene regulation.
Mini tool · MRI preprocessing
A practical toolkit for downloading and preparing ADNI MRI data for imaging and survival analysis. It brings together DICOM-to-NIfTI conversion, N4 bias correction, SynthStrip skull stripping, affine and nonlinear registration, and deformation-derived log-Jacobian maps that quantify local volume change over time.
The longitudinal workflow measures change within each subject before mapping it to a common brain template.
Cover: illustrative mean of three processed annualized maps, shown with the MNI template and analysis masks.
PhD in Statistics, Simon Fraser University — in progress
MSc in Statistics, Fudan University — 2023
BSc in Applied Mathematics, Fudan University — 2020