Liqin Zhou

This paper trains recurrent neural networks to mimic human spatial attention dynamics across different cue types, revealing shared low-dimensional manifolds and sparse coding mechanisms.

This study investigates how precision of contextual predictions modulates social attention using a hierarchical Bayesian model and fMRI, showing that attention allocation is adjusted based on prediction error precision.

The paper shows that deep convolutional neural networks trained for facial identity recognition develop expression-selective units with human-like perception characteristics when exposed to facial expressions.

This paper identifies a whole-brain functional connectivity network linked to non-verbal number acuity and arithmetic skills using connectome-based predictive modeling on resting-state fMRI data.