Alum Jelena Diakonikolas Receives NSF CAREER Award for Pioneering Optimization Research

Award will support research on optimization and machine learning under shifting data distributions.

By
Xintian Tina Wang
February 03, 2025

Professor Jelena Diakonikolas (MSEE '12, PhD EE '16), an expert in large-scale optimization, has been awarded a prestigious National Science Foundation (NSF) CAREER Award for her project Optimization and Learning with Changing Distributions. The award provides $700,000 in funding to support her research and educational initiatives through 2029.

"Because I primarily study optimization, my research profile is atypical of the projects funded by the NSF’s Algorithmic Foundations program. Getting this feedback is both rewarding and validating,” said Diakonikolas in a news story published by the University of Wisconsin–Madison’s Department of Computer Sciences, where she is currently a faculty member.

The CAREER Award is one of the NSF’s most prestigious honors, recognizing early-career faculty with the potential to serve as academic role models in research and education. For Diakonikolas, who joined UW–Madison’s Department of Computer Sciences in 2020 after graduating from Columbia EE department with Master and PhD degrees (advised by Gil Zussman and Cliff Stein in IEOR), this achievement follows her distinction as a recipient of the Air Force Office of Scientific Research (AFOSR) Young Investigators Program award.

Advancing Optimization for Dynamic Learning Environments

Diakonikolas’ work focuses on the mathematical foundations of optimization algorithms, with growing emphasis on their applications in machine learning. Stochastic optimization—a key area of her research—plays a fundamental role in training machine learning models, where data randomness is inherent. However, conventional optimization frameworks often assume that training data distributions remain consistent with real-world applications, which is rarely the case.

Her research addresses this critical gap, developing robust optimization algorithms that adapt to dynamic and uncertain data environments. The project will explore two primary challenges:

  1. Distributionally Robust Optimization – Ensuring that machine learning models perform well even when training and testing data distributions differ significantly, such as in facial recognition models across different demographics or in e-commerce systems responding to shifting consumer behavior.
     
  2. Performative Prediction Stability – Developing algorithms that account for data distribution shifts influenced by the trained model itself, tackling challenges like nonconvex optimization problems and stochastic fixed-point equations.

These advancements are expected to enhance the efficiency and reliability of learning algorithms across various real-world applications, from AI fairness to predictive analytics.

Impact Beyond Research: Education and Collaboration

Beyond its technical contributions, Diakonikolas' CAREER Award will support broader academic initiatives, including:

  • Cross-disciplinary collaborations between optimization and machine learning communities
  • Mentorship opportunities for students, particularly from underrepresented backgrounds
  • A dedicated workshop engaging early-career researchers in advancing optimization and learning theory

The NSF’s recognition of Diakonikolas underscores the growing importance of optimization in modern machine learning and highlights the far-reaching impact of her research.