Daniel Moyer

Computer Science Ph.D. Candidate

University of Southern California

moyerd [at] aardvark.usc.edu
(no animal in actual address)

About Me

I am a 5th year CS grad student at USC. My research is focused on machine learning and its applications to medical imaging, specifically MRI and the human brain. My advisors are Greg Ver Steeg and Paul Thompson.


  • 2018-12-01: Went to NeurIPS (Neural Information Processing Systems) in Montreal to present Invariant Representations without Adversarial Training [arXiv].
  • 2018-11-28: We've renamed our paper from Evading the Adversary in Invariant Representations to Invariant Representations without Adversarial Training [arXiv], hopefully that should clear up some confusion.
  • 2018-09-18: Congrats to Fabian Corlier on his accepted paper at SIPAIM-18!
  • 2018-09-16: Went to MICCAI 2018, it was a great conference! I gave at talk at the CDMRI Workshop, and won their Best Oral Presentation.


Invariant Representations [link]:

Often we want a representation of our data that is invariant to known outside factors:

  • scans ⊥ site
  • images ⊥ rotations
  • predictions ⊥ protected classes.
This project is about the construction of general invariant representations.

Continuous Brain Connectivity [link]:

Brain connectivity is synonymous with discrete brain networks. Connectivity, however, persists without explicit subsets and can be generalized to continuous coordinate systems. This project explores the continuous connectivity paradigm.

Selected Publications and Preprints

See the full list here: [link]

  1. Invariant Representations without Adversarial Training
    D Moyer, S Gao, R Brekelmans, G Ver Steeg, A Galstyan
    Neural Information Processing Systems (NIPS) 2018
  2. Measures of Tractography Convergence
    D Moyer, PM Thompson, G Ver Steeg
    MICCAI CDMRI (Computational Diffusion MRI) Workshop, 2018
  3. Continuous Representations of Brain Connectivity using Spatial Point Processes
    D Moyer, B Gutman, N Jahanshad, J Faskowitz, PM Thompson
    Medical Image Analysis (MedIA) 2017.
  4. A Restaurant Process Mixture Model for Connectivity Based Parcellation of the Cortex
    D Moyer, B Gutman, N Jahanshad, PM Thompson
    Information Processing in Medical Imaging (IPMI) 2017.
  5. A Continuous Model of Cortical Connectivity
    D Moyer, B Gutman, N Jahanshad, J Faskowitz, PM Thompson
    Medical Image Computing and Computer Assisted Intervention (MICCAI), 2016.

Mathematics is very much like poetry. What makes a good poem -- a great poem -- is that there is a large amount of thought expressed in very few words. In this sense \[e^{\pi i} + 1 = 0 ~~~\text{ or }~~~ \int_{-\infty}^{\infty} e^{-x^2}dx = \sqrt{\pi}\] are poems
Lipman Bers

Other random stuff (unsorted).

This website design was copied lock-stock-and-barrel from Sam Way.