Welcome to my home page!
My name is Nina, I am an Assistant Professor of Electrical and Computer Engineering at UC Santa Barbara. My research group aims to create computational representations of the human body at different scales - thus contributing to the development of Computational Medicine. Before joining UC Santa Barbara, I performed my postdoctoral work in Statistics at Stanford under the supervision of Susan Holmes, and my doctoral work in Computer Science at Inria in the Asclepios-Epione team under the supervision of Xavier Pennec.
To Students: M.Sc. and PhD positions are available! If interested, contact me at ninamiolane at ucsb.edu.
To Postdocs: Open position in geometry and deep learning to reveal the structure of membrane proteins, see details here.
At the macroscopic scale, I am interested in learning quantitative descriptions of organ shapes and functions, together with their normal and pathological variations in the population; for example, learning the manifold of the brain anatomy or the manifold of the brain activity during certain functional tasks. At the microscopic scale, I am interested in studying cells shapes and functions, and at the nanoscopic scale, molecular shapes and functions. My goal is to leverage these heterogeneous multiscale representations to implement new computer-assisted diagnosis methods and develop innovative treatments for diseases.
My methodological interests lie at the intersection of Geometry, Statistics and Computer Science.
Geometric Statistics: Statistical theory for data belonging to non-linear spaces, like metric spaces and Riemannian manifolds. Shape data, or weighted graph data naturally belong to such spaces, while high-dimensional data can be naturally projected to such spaces.
Dimensionality reduction in non-linear spaces: Fréchet mean, Submanifold learning in metric spaces and Riemannian manifolds, with a special interest for manifolds with additional properties, like Lie groups and Quotient spaces.
Fast implementation of the above techniques: Variational inference and variational autoencoders for submanifold learning.
Statistical properties of the above techniques: Asymptotic studies and bias-correction methods.
My application domains are medical imaging and biological imaging, with a special interest in brain data:
Brain shapes, as observed in brain MRIs (macroscopic scale),
Brain BOLD (blood oxygen level dependent) activation, as observed in brain functional MRIs (macroscopic scale),
Brain structural and functional connectomes (macroscopic scale),
Neuronal electric signals, either from EEG or from deep brain implants (macroscopic or microscopic scale),
Molecular imaging using cryo electron microscopy (nanoscopic scale).
Beside my research interests, I am a lecturer for the classes “Introduction to Statistical Methods: Precalculus“ (2019) and “Statistical Methods for Engineering and the Physical Sciences“ (2018, 2019) at Stanford University. I am a reviewer for the scientific journals Journal of Mathematical Imaging and Vision JMLR (2015, 2017), Biometrika (2018) and the Journal of Mathematical Neuroscience (2019), as well as a reviewer for the conferences NeurIPS (2016, 2018, 2019), International Conference of Machine Learning ICML (2019), Geometric Science of Information GSI (2017, 2019). I am a member of the scientific committee of the Conference of Geometric Science of Information (GSI) (2017, 2019) and I was one of the two lead organizers of the workshop on Brain Computing at the Berkeley-Inria-Stanford annual meeting (2017).
You can follow me on: Github, LinkedIn, Twitter: @ninamiolane.