Victor Geadah

(I pronounce my last name as "Jada")

I am a Flatiron Research Fellow in the Center for Computational Neuroscience at the Flatiron Institute, NYC.

My research draws on dynamical systems, control theory, and probabilistic machine learning to understand how intelligent systems learn and adapt to new environments. I approach this by building statistical methods grounded in normative theory, which aim to explain both what we observe in data, and why. You can learn more about my research in this video.

Bio: I received my PhD in Applied and Computational Mathematics from Princeton University, where I was advised by Jonathan Pillow in the Princeton Neuroscience Institute and was awarded the Porter Ogden Jacobus Fellowship. Before Princeton, I studied applied mathematics at the University of Cambridge (master's student, Part III), and mathematics at the Université de Montréal (BSc), where I worked with Guillaume Lajoie on dynamic coding in recurrent neural networks.

Google Scholar | GitHub | CV

News

Sept. 2026: Excited to start as a postdoctoral research fellow at the Flatiron Institute, Center for Computational Neuroscience (CCN) in NYC!
Apr. 2026: I had the pleasure of defending my PhD thesis on Statistical Inference of Normative Processes in Neural and Behavioral Dynamics. You can find it in open access here.
Sep. 2025: Our NeurIPS submission on Conditionally Linear Dynamical Systems has been accepted!
May 2025: Excited to announce I'm starting as a research scientist intern at Meta CTRL-Labs (NYC) this summer.
Apr. 2025: I'm honored to have been awarded Princeton's Porter Ogden Jacobus Fellowship for 2025-26!
Feb. 2025: Our ICLR submission on comparing noisy population dynamics has been selected for an oral presentation!

Victor Geadah
Flatiron Institute, Simons Foundation
New York, NY 10010, U.S.A.
vgeadah [at] flatironinsitute.org

Publications

Journal Articles

Sparse-coding variational autoencoders
Victor Geadah*, Gabriel Barello*, Daniel Greenidge, Adam S. Charles, and Jonathan W. Pillow (2024)
Neural Computation 36(12), 2571-2601. [link | pdf | code]
Neural networks with optimized single-neuron adaptation uncover biologically plausible regularization
Victor Geadah, Stefan Horoi, Giancarlo Kerg, Guy Wolf, and Guillaume Lajoie (2024)
PLOS Computational Biology 20 (12), e1012567. [link | bioRxiv].

*: Equal contribution.

Conference Proceedings

Modeling neural activity with conditionally linear dynamical systems
Victor Geadah, Amin Nejatbakhsh, David Lipshutz, Jonathan W. Pillow, and Alex H. Williams (2025)
The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS). [link | code | bibtex]
Flexible inference of learning rules from de novo learning data using neural networks
Y. Helena Liu, Victor Geadah, and Jonathan W. Pillow (2025)
The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS). [link | bibtex]
Comparing noisy neural population dynamics using optimal transport distances
Amin Nejatbakhsh, Victor Geadah, Alex H. Williams, and David Lipshutz (2025)
The Thirteenth International Conference on Learning Representations (ICLR). [oral | link | bibtex]
Inferring system and optimal control parameters of closed-loop systems from partial observations
Victor Geadah, Juncal Arbelaiz, Harrison Ritz, Nathaniel D. Daw, Jonathan D. Cohen, and Jonathan W. Pillow (2024)
The 63rd IEEE Conference on Decision and Control (CDC), 8006-8013. [link | pdf | bibtex]
Parsing neural dynamics with infinite recurrent switching linear dynamical systems
Victor Geadah, International Brain Laboratory, and Jonathan W. Pillow (2024)
The Twelfth International Conference on Learning Representations (ICLR). [link | pdf | bibtex]
Low-dimensional dynamics of encoding and learning in recurrent neural networks
Stefan Horoi, Victor Geadah, Guy Wolf, and Guillaume Lajoie (2020)
Canadian Conference on Artificial Intelligence.

Preprints

Inferring learning rules during de novo task learning
Victor Geadah and Jonathan W. Pillow.
[bioRxiv]
Modeling complex animal behavior with latent state inverse reinforcement learning
Aditi Jha, Victor Geadah, and Jonathan W. Pillow.
[bioRxiv]