Search Papers | Poster Sessions | All Posters
Poster A134 in Poster Session A - Tuesday, August 6, 2024, 4:15 – 6:15 pm, Johnson Ice Rink
Layer-specific cortical mechanisms underlying visual perceptual learning
Monika Jozsa1, Clara Pecci Terroba1, Ke Jia2, Mengxin Wang3, Zoe Kourtzi1, Yashar Ahmadian1 (); 1University of Cambridge, 2Zhejiang University, Hangzhou, China, 3University of Oxford
Training in perceptual tasks can enhance the brain’s internal representations of task-relevant features for improved decision making. Recent ultra-high-field neuroimaging studies have found that training in fine discrimination tasks leads to increased signal-to-noise ratio of activity patterns in the superficial layers of primary visual cortex (V1), accompanied by an increase in GABAergic inhibition. However, the causal circuit mechanisms underlying the layer-specific representational changes, and the layer-specificity of changes in cortical excitation and inhibition, are unknown. Here, we theoretically study these questions by training a biologically-constrained mechanistic model of V1 in a fine orientation discrimination task near a fixed orientation. Training led to (1) strengthening (weakening) of cortical inhibition (excitation), which was larger and more robust in the model’s superficial layer, and (2) sharpening of the superficial-layer tuning curves at the trained orientation, as previously reported in neurophysiology experiments. Further, these changes correlated with improvement in the network’s task performance. Finally, the mechanistic nature of our model allows making testable predictions about the causal pathways linking layer-specific changes in excitation and inhibition with representational or behavioral improvements.
Keywords: visual perceptual learning cortical excitation and inhibition recurrent network dynamics deep learning