Mo Shakiba · NeuroAI researcher

Mo Shakiba

I’m Mo, a NeuroAI researcher.

On: 2025-04-18 Post

Impact Scholars Program: Neuro-constrained Bio-inspired Recurrent Neural Network

Inspired by the release of the MICrONS dataset, we investigated how anatomical and functional constraints from biological neural circuits could be incorporated into recurrent neural networks and how such constraints affect both performance and network structure. Following the micropublication1, the project was presented as a poster2 at CNS*2025 and the Kempner Institute Frontiers in NeuroAI, and the resulting article has since been posted to arXiv3. This work originated in the Deep Learning stream of the Neuromatch Academy Impact Scholars, a six-month research program that provided exclusive mentorship and computational resources to pursue independent projects4. The video below is the original project presentation.

  1. Neuro-constrained Bio-inspired Recurrent Neural Network ↩︎

  2. Anatomically and Functionally Constrained Bio-Inspired Recurrent Neural Networks Outperform Traditional RNN Models ↩︎

  3. Harnessing cortical geometry, wiring, and function as inductive biases for recurrent neural networks ↩︎

  4. Impact Scholars Spotlight ↩︎