Computational Neuroscience Seminar
Multiplicative feedback gating facilitates rapid learning and flexible computation in recurrent neural circuits
Speaker: Sage Chen, NYU Grossman School of Medicine
Location: Warren Weaver Hall 1314
Videoconference link: https://nyu.zoom.us/j/9726507138
Date: Tuesday, December 9, 2025, 4 p.m.
Synopsis:
The talk will start with a mathematical theory about multiplicative gating in recurrent neural network (RNN) dynamics and then transition to biological recurrent neural circuits in neuroscience. The mammalian forebrain is the seat of higher cognition with architectural parallels to modern machine learning systems. Specifically, the cortex resembles RNNs while the thalamus resembles feedforward neural networks (FNNs). How such architectural features endow the forebrain with its learning capacity, is unknown. Here, we take inspiration from empirical thalamocortical discovery and develop a multiplicative coupling mechanism between RNN-FNN architectures that collectively enhance their computational strengths and learning. The multiplicative interaction imposes a Hebbian-weight amplification onto synaptic-neuronal coupling, enabling context-dependent gating and rapid switching. Through a wide range of benchmark experiments on working memory, decision making, control, and pattern classification, we demonstrate that multiplicative gating-driven synaptic plasticity achieves 2-100 folds of speed improvement in supervised, reinforcement and unsupervised learning settings, boosting memory capacity, model robustness and generalization of RNNs. We further demonstrate the efficacy and biological plausibility of multiplicative gating in modeling four multiregional RNN-FNN circuits, including (1) a prefrontal cortex-mediodorsal thalamus network for context-dependent probabilistic decision making and context switching (Mukerjee et al. 2021 Nature; Wang et al. 2023 Nat. Commun.), (2) a cortico-thalamic-cortical network for working memory and attention (Panichello and Buschman, 2021 Nature), (3) a cerebellar-thalamic-cortical network for motor task switching (Pemberton et al., 2024 Nat. Commun.), and (4) an entorhinal cortex-hippocampus network for visuospatial navigation and sequence replay. Our model predictions not only validate various experimental findings reported independently from multiple species (rodent, monkey, human), multiple brain structures (MD thalamus, pulvinar, motor thalamus, hippocampal formation), and diverse tasks (cognitive, motor, navigation), but also provide experimentally testable hypotheses in neural perturbation studies.