Zeyuan Ye

This paper develops a Gaussian Process with Kernel Regression method to analyze grid cell population activity, finding that running speed increases manifold size and noise strength, with manifold dilation having a greater impact on information coding.

The paper investigates out-of-distribution generalization in mouse visual cortex and deep neural networks, finding that geometric properties of neural representations correlate with generalization ability.

The paper shows that a self-supervised neural network exhibits border ownership signals similar to those in the brain, enhancing video prediction.

This paper trains RNNs on a color delayed-response task to show that decoding phase strategies reduce memory error by attributing broader neural states to high-probability colors.