George Dimitriadis

This paper shows that RNNs learn abstract relational representations through low-rank recurrent dynamics, which support temporal abstraction and generalization.

This study challenges the traditional view of motor cortex function by showing that rats with motor cortical lesions show impairments in unexpected obstacle course scenarios but not in repeated trials.

This paper introduces a high-density CMOS scanning probe for neural recording, demonstrating its ability to capture neural activity across multiple brain regions with improved resolution compared to lower density devices.

The paper describes a method to record from the same neuron using patch-clamp and CMOS probes, presents a dataset, and discusses collaborative analysis through an open repository.

The paper uses t-SNE to visualize and cluster extracellular spike data from neural recordings, showing it can identify single units and improve sorting efficiency.