Robert D. Nowak

Also known as Rob Nowak, Robert Nowak

Robert Nowak is a University of Wisconsin–Madison engineering professor whose work in signal processing and machine learning ranges from sparse signal recovery to a mathematical theory of neural networks. He earned his degrees in electrical engineering at Wisconsin by 1995 and held posts at Rice University and Michigan State University before taking the McFarland-Bascom professorship in engineering. With Stephen Wright and Mário Figueiredo, he wrote the SpaRSA paper that won the 2014 IEEE Baker Award, and his crowdsourcing algorithms began sorting captions for The New Yorker's cartoon contest in 2016. In 2020 and 2021, he and Rahul Parhi posted papers on which functions neural networks learn, and by 2023 he directed a Data Efficient Machine Learning center. He is the Grace Wahba Professor of Data Science at Wisconsin and Chief Scientist at Ludo Robotics.

Now at: University of Wisconsin–Madison, Toyota Technological Institute at Chicago and Ludo Robotics
Badges: Bridge
Boomerang
Loyal

Timeline

1990–1995

Degrees at Wisconsin–Madison

He earns B.S., M.S. and Ph.D. degrees in electrical engineering from the University of Wisconsin–Madison in 1990, 1992 and 1995.
1995–2003

Posts at Rice and Michigan State

He is a postdoctoral fellow at Rice University in 1995–1996 and an assistant professor at Michigan State University from 1996 to 1999. From 1999 to 2003, he holds assistant and associate professor positions at Rice, and he then becomes McFarland-Bascom Professor of Engineering at Wisconsin–Madison.
2009–2014

SpaRSA and a Baker Award

With Stephen J. Wright and Mário A.T. Figueiredo, he publishes "Sparse Reconstruction by Separable Approximation" in the IEEE Transactions on Signal Processing in July 2009. The SpaRSA method combines convex optimization, linear algebra and dimensionality reduction to find near-optimal sparse approximations, and the paper receives the IEEE W.R.G. Baker Award in 2014. He also wins the IEEE Signal Processing Society Best Paper Award in 2011.
2016

Algorithms for The New Yorker

In March 2016, The New Yorker reveals that it is using crowdsourcing algorithms from a University of Wisconsin team led by Nowak to find the best captions in its cartoon caption contest. The adaptive algorithms weed out the weakest captions quickly so that more voters see the likely winners. The U.S. Air Force and American Family Insurance also use the algorithm.
2020–2021

What neural networks learn

In June 2020, he and Rahul Parhi post "Banach Space Representer Theorems for Neural Networks and Ridge Splines," which shows that finite-width, single-hidden layer neural networks solve a family of inverse problems posed in the Radon domain. In May 2021, they post "What Kinds of Functions do Deep Neural Networks Learn? Insights from Variational Spline Theory," which does the same for deep ReLU networks over a new function space.
By 2023

Directing a data-efficient learning center

By 2023, he holds the Keith and Jane Nosbusch Professorship in Electrical and Computer Engineering at Wisconsin–Madison and directs the AFOSR/AFRL University Center of Excellence on Data Efficient Machine Learning. He serves on the editorial boards of the SIAM Journal on the Mathematics of Data Science and the IEEE Journal on Selected Areas in Information Theory.
As of October 2026

Wahba Professor and Ludo Robotics

He is the Grace Wahba Professor of Data Science and the Keith and Jane Nosbusch Professor in Engineering at the University of Wisconsin–Madison, an adjoint professor at the Toyota Technological Institute at Chicago, and Chief Scientist at Ludo Robotics. His current work investigates the mathematical foundations of AI, human-AI interaction, and robotics.

Career

Since 2003University of Wisconsin–MadisonMcFarland-Bascom Professor of Engineering
NowToyota Technological Institute at ChicagoAdjoint Professor
NowLudo RoboticsChief Scientist
1996–1999Michigan State UniversityAssistant Professor
1995–2003Rice UniversityPostdoctoral Fellow, Assistant and Associate Professor