Xiaohan Zhang

This study uses computational models to show that the mediodorsal thalamus improves decision-making by reducing sensory noise, enhancing working memory, and enabling context switching.

The paper trains a biologically-inspired recurrent neural network to simulate visual grid cell activity in rat V2 and finds grid-like patterns emerge robustly across different network configurations and tasks.

This study uses a recurrent neural network to model context-dependent computation in the prefrontal cortex, revealing geometric patterns of information coding and integration.