PCA on spike data and ML on Brian neurons


I am currently working on a model of the primary motor cortex (built with brian) and have been producing some spontanoues(ish) activity (driven from thalamic inputs). I was wondering if anyone has used or experimented with implimentation of PCA or any other dimentionality reduction technique in order to visualise any underlying neural manifolds? Additionally, I am hoping to train this network to produce some desired dynamics given some specific input. I have been looking into Spytorch (which seems to map the SNN to a RNN), has anyone had any experience with this?

Any help/advice is greatly appreaciated, cheers.