Nneka Okolo
Home
Research
Experiences

Nneka Okolo

Watercolor of Ariake flower by Kōkichi Tsunoi, 1921 

Norbert Wiener Postdoctoral Associate,
The Institute for Data, Systems, and Society (IDSS),
Massachusetts Institute of Technology (MIT)
E-mail: nnekaokolo3 [@] gmail [DOT] com
Google Scholar

About me

Hello, I am Nneka a Norbert Wiener Postdoctoral Associate at IDSS in MIT, where I am fortunate to be mentored by Prof. Alexander (Sasha) Rakhlin. In January 2025, I received a Ph.D. in Computer Science from Universitat Pompeu Fabra (UPF) in Spain. In the four years leading to this, I worked under the brilliant supervision of Prof. Gergely Neu, and was generously funded by the Dorcas Muthoni Fellowship.

Other highlights of my academic journey include a MSc. project in Reinforcement Learning while at the African Institute for Mathematical Sciences (AIMS) in Senegal and BSc. in Applied Mathematics from the University of Benin in my home country, Nigeria. During this time, I am very thankful to have been advised by Prof. Issmail El Hallaoui, Dr. Amira Dems, Dr. Mouhamad M. Allaya and Dr. Amos Egonmwan, and received funding from the Mastercard Foundation via AIMS.

Research

My research lies at the intersection of three main areas:

  • Reinforcement Learning Theory
    Towards creating fully autonomous systems with success in real-world environments, I am interested in provably time- and sample-efficient RL for large-scale sequential decision-making. In particular, I focus on RL with function approximation, and aim to develop implementable algorithms with both theoretical time and sample complexity guarantees.


  • Stochastic Optimization and Online learning
    My venture into these areas have been mostly as a means to an end. During the Ph.D., I primarily worked with a formulation of Markov Decision Processes (MDPs) in unknown environments, such that,

    1. Learning a near-optimal decision-making policy translates to solving a stochastic convex-concave optimization problem and,

    2. Providing performance guarantees is made feasible with techniques from online learning.

    Interestingly, this approach to RL raises further inquiry into algorithms for solving broader convex-concave problems without conventional assumptions enjoyed by the optimization and online learning communities. I find this research direction separately worth investigating.

News