Adám Gosztolai
Fields of expertise
non-linear dynamics – geometric deep learning – animal behaviour – computational neuroscience
Mission
Both our brains and AI systems solve computational challenges in a highly distributed manner encoded in the collective activity of neural populations. My research focuses on understanding the dynamical processes underpinning neural computations to derive common algorithmic principles shared by these fundamentally different systems. I am motivated by two synergistic aims: (1) developing novel methods using machine learning, geometry and dynamical systems theory that facilitate discovering better models of how the brain works and (2) reverse-engineering the dynamical processes that underpin complex animal behaviours to develop more advanced AI systems that benefit clinical applications such as brain-machine interfacesEducation
PhD
Mathematics
Imperial College London
2014-2018
MRes
Systems and Synthetic Biology
Imperial College London
2013-2014
MASt
Mathematics
University of Cambridge
2012-2013
BEng
Mechanical Engineering
University College London
2008-2011
Publications
Selected publications
Gosztolai A, Günel S, Lobato-Ríos V, Pietro Abrate M, Morales D, Rhodin H, Fua P, Ramdya P Nature Methods |
LiftPose3D, a deep learning-based approach for transforming two-dimensional to three-dimensional poses in laboratory animals |
Gosztolai A, Arnaudon A Nature Communications |
Unfolding the multiscale structure of networks with dynamical Ollivier-Ricci curvature |
Gosztolai A, Peach L R, Arnaudon A, Barahona M, Vandergheynst P |
Interpretable statistical representations of neural population dynamics and geometry |