Search Papers | Poster Sessions | All Posters
Poster A142 in Poster Session A - Tuesday, August 6, 2024, 4:15 – 6:15 pm, Johnson Ice Rink
Multi-stage Cortical Recurrent Circuit Implementing Normalization
Asit Pal1 (), Shivang Rawat1, David Heeger1, Stefano Martiniani1; 1New York University
Communication between cortical areas is supported by long-range reciprocal connections. Given feedback connections’ hypothesized role in attentional modulation, it is essential to develop a multistage network model of the brain for studying attention and inter-area communication. Here, we present a dynamically stable hierarchical recurrent neural circuit model with feedback that implements divisive normalization exactly at each stage of its hierarchy. We consider a two-stage model (V1 and V4), each stage receives input from the preceding area and feedback from the subsequent area, and the responses in each area are normalized by local inhibitory signals. We note that an increase in feedback from V4 to V1, amplifies responses in both stages, with a more pronounced increase in higher cortical areas, consistent with experimental findings. Additionally, our model predicts that feedforward and feedback signals in the brain propagate via distinct frequency channels, gamma and alpha frequencies respectively, in line with empirical evidence. Furthermore, our model admits a low-dimensional communication subspace (within and across areas) and predicts that enhancing feedback improves inter-areal communication, yet decreases within-area communication. In summary, our hierarchical model provides a robust and analytically tractable framework for exploring normalization, attention, and interareal communication
Keywords: Multi-stage Attention Communication Coherence