alexander iannantuono ['ja.nan.two.no]

Myself, circa 2016

http303 -> 127.0.0.1

Kelowna, BC

Canada, V1V 000

I am a MSc student in mathematics at the University of British Columbia Okanagan Campus. I am working under the supervision of Yves Lucet. I graduated virtually from McGill University with a BSc in mathematics and computer science.


Research

Current: nothing for now! just taking neat classes.

In the near future: optimization for road design applications

Timeline

2020
  • Worked on a neat class (MATH562 at UBCO with Warren Hare) project which involved investigating surrogate problems in (black box) optimization.
    • I wrote a small document to gather my thoughts (as well as data and figures) in preparing for a presentation.
    • Coding in MATLAB + Python was involved, code is available on GitHub
  • Worked on a project that aims to provide a measure of confidence that a deep neural network has given some classification task
    • Pre-print on arXiv
    • Recently chosen for a spotlight presentation at the ICML 2020 Uncertainty & Robustness in Deep Learning (UDL) workshop
2019
  • Worked with Aram-Alexandre Pooladian on a project involving Deep Learning (DL) perturbation attacks using a Langevin dynamics based approach (preprint)
  • Worked on an algorithm that became the No-collision transportation maps paper
    • It fares well empirically for approximating the optimal transport solution in \(\mathbb{R}^n\), \(n > 1\) using \(\ell_1^2, \ell_2^2, \ell_\infty^2, \ell_2\) as measures
    • A poster (pdf) was presented the NeurIPS 2019 OTML workshop in December 2019)
    • Pre-print on arXiv, code on GitHub
Interests
  • Optimization
  • Optimal transport
  • Numerical methods
  • Machine learning
  • Distributed and parallel computing

I hope to specify further as stronger interests develop.


Small Talk Topics

Some hobbies of mine that do not pertain to the above
  • Bicycles: cycling, maintenance, etc.
  • Collecting (vinyl) records, high-fidelity sound
  • Almost anything digital: repair, building, design, etc.
  • Photography (sometimes with artistic distortion), graphic design and typography
  • Tinkering in the vain of N-O-D-E’s projects
Nationality: Canadian
Languages (spoken + written):
  • English
  • French
  • ~Italian (looking to improve)

Colleagues & Friends


news

Oct 23, 2020 I will be giving a talk on the No-collision transportation maps work at the COCANA on January 14 March 4, 2021 via Zoom.
Jul 5, 2020 Work on using Out-of-Distribution (OOD) data to calibrate deep neural networks has been accepted to the UDL workshop!

selected publications

  1. JSC
    No-Collision Transportation Maps
    Nurbekyan, L., Iannantuono, A., and Oberman, A. M.
    2020