Computational Neuroscience Seminar

---

Building models with biologically-realistic neural dynamics

Speaker: Grace Lindsay, NYU

Location: Warren Weaver Hall 1314

Date: Tuesday, October 6, 2026, 4 p.m.

Synopsis:

Artificial neural networks can be useful models of information processing in the brain. However these models frequently lack dynamics or have only very coarse dynamics that cannot be directly related to neural data easily. In this talk, I will cover two projects in my lab aimed at building models that can perform biologically-relevant tasks with more realistic neural dynamics. The first is DynVision, a toolbox for building ODE-based recurrent convolutional neural networks that serve as models of the visual system. The second is NCAP, a biologically-inspired model of recurrent circuits in the spinal cord that controls a 3-D simulated body. Both of these projects offer the opportunity to connect neural dynamics to behavior on challenging tasks.