Reviews: GP CaKe: Effective brain connectivity with causal kernels

This paper addresses the problem of understanding brain connectivity (i.e. More generally, perhaps, the paper attempts to uncover causal structure and uses neuroscience insights to specifically apply the model to brain connectivity. The model can be seen as an extension of (linear) dynamic causal models (DCMs) and assumes that the observations are a linear combination of latent activities, which have a GP prior, plus Gaussian noise (Eq 11). Overall the paper is readable but more clarity in the details of how the posterior over the influence from i- j is actually computed (paragraph just below Eq 12). I write this review as a machine learning researcher (i.e.