Main Focus
My research lies at the intersection of systems and computational neuroscience. I am generally interested in understanding how animals adaptively learn to modify their behavior according to environmental dynamics. To answer this question, I use reinforcement learning models to characterize the mechanisms underlying foraging behavior in naturalistic settings.
During my PhD (working with Anna Levina and Tatiana Engel), I studied how the brain develops adaptive dynamics and behaviors across different timescales. We developed data analysis methods (Zeraati et al., Nat. Comput. Sci, 2022) and computational models (Zeraati et al., Nat. Commun., 2023; Shi, Zeraati et al., PRR, 2023) to show that timescales of neural dynamics can flexibly adapt to behavioral states (e.g., selective attention), correlate with behavioral outputs such as reaction time, and are shaped by the spatial network structure and top-down inputs. Inspired by these findings, I became interested in the functional role of neural timescales, i.e. their role in neural computations. For this purpose, we used task-optimized recurrent neural networks (RNNs) to study the link between neural timescales and working memory (Khajehabdollahi*, Zeraati* et al., ICLR, 2024).
Curriculum Vitae
- Ph.D. in Neuroscience, Graduate Training Center of Neuroscience, IMPRS MMFD, University of Tübingen, Germany (2025)
- SMARTSTART 2 Fellow in computational neuroscience funded by Bernstein Network and the Volkswagen Foundation (2019)
- M.Sc. in Neural Information Processing, Graduate Training Center of Neuroscience, University of Tübingen, Germany (2018)
- B.Sc. in Physics and Biomedical Engineering (double-major), Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran (2016)