Self-organising neural systems

The brain has access to a rich set of mechanisms governing neural connectivity and plasticity, including gap junctional coupling, homeostasis, and spike-timing dependent plasticity. An outstanding challenge in relating artificial neural network studies to neuroscience is to reconcile standard network training protocols with these mechanisms. Here, we will ask a complementary question: what network motifs can be established using these basic biological building blocks? Using meta-learning approaches, we show examples of several neural computations that are able to self-organize using intrinsic, unsupervised biological learning rules, including ensemble formation, sequence generation and integration. We explore the character of the emerging rules and compare with analytical network models.