Daniel McKenzie
Applied Math and ML.
Positions
2019-present
University of California, Los Angeles. Assistant Adjunct Professor
Research interests
Optimization and signal processing applied to machine learning and data science. More specifically: zeroth-order optimization and applications, implicit neural networks, data-driven metrics for unsupervised learning, and knowledge graphs.
Education
2013-2019
University of Georgia, USA. PhD. Advisor: Ming-Jun Lai.
03/2013-06/2013
University of Bayreuth, Germany. Short term visitor.
2011-2013
University of Cape Town, South Africa. M.Sc (with distinction).
2007-2010
University of Cape Town, South Africa. B.Sc (with distinction)
Awards
2021
Liggett Instructor Award.
2019
William Armor Wills Memorial Scholarship.
2017
University Outstanding Teaching Assistant Award.
2014-2016
NRF Doctoral Scholarship for Study Abroad.
2013
DAAD Short Term Research Exchange (to University of Bayreuth).
2012
DAAD-NRF Joint Masters Bursary.
2010
UCT Council Merit Scholarship.
2010
NRF Honours Bursary.
2009
Jakob Burlak Memorial Trust Scholarship.
2007,2008,2009
UCT Science Faculty Scholarship.
Publications
A list is also available online. *= undergraduate coauthor.
Journals
2021
Balancing geometry and density: Path distances on high-dimensional data. SIMODS Anna Little, Daniel McKenzie, and James M. Murphy (alphabetical).
2021
Zeroth-order regularized optimization (ZORO): Approximately sparse gradients and adaptive sampling. SIOPT (to appear) HanQin Cai, Daniel McKenzie, Wotao Yin and Zhenliang Zhang (alphabetical).
2020
Compressive sensing for cut improvement and local clustering. SIMODS. Ming-Jun Lai and Daniel McKenzie (alphabetical).
2019
Power weighted shortest paths for clustering Euclidean data. Foundations of Data Science. Daniel McKenzie and Steve Damelin
Conferences
2022
JFB: Jacobian-Free Backpropagation for Implicit Networks. AAAI 2022 (15% acceptance rate). Samy Wu Fung, Howard Heaton, Qiuwei Li, Daniel McKenzie, Stanley Osher and Wotao Yin.
2021
A zeroth-order, block coordinate descent algorithm for huge-scale black-box optimization. ICML 2021 (22% acceptance rate). HanQin Cai, Yuchen Lou*, Daniel McKenzie, and Wotao Yin (alphabetical).
2020
Who killed Lilly Kane? A case study in applying knowledge graphs to crime fiction. IEEE Big Data GTA3 Workshop. With Mariam Alaverdian*, William Gilroy*, Veronica Kirgios*, Xia Li, Carolina Matuk*, Daniel McKenzie, Tachin Ruangkriengsin*, P. Jeffrey Brantingham and Andrea Bertozzi (alphabetical).
Submitted
2021
From the simplex to the sphere: Faster constrained optimization using the Hadamard parametrization Qiuwei Li, Daniel McKenzie, and Wotao Yin (alphabetical).
2021
Curvature-Aware Derivative-Free Optimization Bumsu Kim, Daniel McKenzie, HanQin Cai and Wotao Yin.
2021
Learn to predict equilibria via Fixed Point Networks. Howard Heaton and Daniel McKenzie, Qiuwei Li, Samy Wu Fung, Stanley Osher and Wotao Yin.
2020
A one-bit, comparison-based gradient estimator. HanQin Cai, Daniel McKenzie, Wotao Yin and Zhenliang Zhang (alphabetical).
Talks
All talks invited unless otherwise stated.
03/2022
SIAM UQ22: Deep Learning for Optimization. Atlanta, USA
03/2022
Math Machine Learning Seminar. Max Planck Institute, Germany.
10/2021
INFORMS2021: Recent Advances in Derivative-free Optimization. Anaheim, USA.
08/2021
Mathematics of Machine Learning Conference. Bielefeld University, Germany. (contributed)
07/2021
SIAM OP21: Optimization, Data Science and their Applications. Online.
06/2021
Optimal Transport and Mean Field Games Seminar. Online.
11/2020
INFORMS2020: Session on Recent Progress in Blackbox Optimization. Online.
04/2020
Tufts Math of Data Science Lecture Series. Online.
02/2020
UGA Applied Math Seminar. Athens, USA.
09/2019
SIAM South-Eastern Sectional. Knoxville, USA.
05/2018
International Conference on Computational Harmonic Analysis (ICCHA7). Nashville, USA. (contributed)
10/2018
AMS Central Sectional. Ann Arbor, USA.
Teaching
2021-present
Math 151BH: Honors Applied Numerical Methods II, UCLA. (Taught twice. I also co-developed this course).
2021-present
Math 151AH: Honors Applied Numerical Methods I, UCLA. (Taught twice. I also co-developed this course).
2020-present
Math 118: Mathematical Methods of Data Theory, UCLA. (Taught four times. I also co-developed this course).
2020
Math 170S: Statistics, UCLA.
2019
Math 151A: Applied Numerical Methods I, UCLA.
2019
Math 32A: Calculus III, UCLA.
2015-2019
Math2250: Calculus I, UGA. (taught three times).
2014-2018
Math1113: Precalculus, UGA. (taught six times).
Service
Organization
2021
Exploiting structure in zeroth-order optimization. Workshop at INFORMS2021.
2020
ZOOM: Zeroth Order Online Meeting. (Online) mini-conference.
Undergrad. Students Mentored
2021
Yuchen Lou (Hong Kong Univ. –> Northwestern Univ.), Isha Slavin (UCLA), Allen Zou (UCSD).
2020
Mariam Alaverdian (Los Angeles Community College –> Yale), William Gilroy (Harvey Mudd), Veronica Kirgios (Notre Dame), Carolina Matuk (Univ. Iowa), Tachin Ruangkriengsin (UCLA–> Princeton), Charles Stoksik (UCLA), Chenglin Yang (UCLA –> Columbia), Allen Zou (UCSD).
2018
Lucas Connell (UGA).
Graduate Students Mentored
2019-present
Howard Heaton, Bumsu Kim.
Outreach
2020
CSST Mentor. The resulting paper was presented at ICML.
2020
UCLA REU Mentor. The resulting paper was presented at IEEE Big Data 2020
2018
UGA MathCamp. I mentored a group of five high school students on the “Monster Epidemiology” project.
2011-2012
SHAWCO. I tutored students, trained volunteers and co-led the KenSMART project.
Skills
Proficient in Python (specifically: NumPy, SciKitLearn and PyTorch), Matlab, LaTeX, Git, and Markdown. Experienced in mentoring junior researchers, and in explaining technical concepts to non-technical audiences.