Kernel Analysis Of Multi-neuronal Spike Trains

We recorded multi-neuronal spike activities from hippocampal CA1 regions of rats performing a conditional discrimination task. Separating single unit activities from multi-neuronal spike activities, we obtained spike count data. Then, we calculated kernel matrices from the spike count data. The kernel we used, called Spikernel, measures the similarities among spike count data. We performed the kernel k-means clustering and the kernel PCA with the kernel matrices. The data can be separated into two clusters, which represent the behaviors of rats. Our results suggest that spike activities of suitable length are crucial to predict the behaviors of rats.