#32 · Systems biology

Correlated noise and plasticity shape how neural networks represent and replay activity

The dominant thread is how activity at successive levels of neural organization, from population noise and firing to synaptic release, determines the representations a network carries and how they are replayed. A mean-field reduction of correlated-noise quadratic integrate-and-fire networks reports that correlation suppresses firing and induces metastability. A dynamic mean-field theory for nonlinear noise in recurrent networks uses a Gaussian-equivalent OU noise replacement with lognormal moment closure to derive a closed dynamical description. An information-theoretic treatment of associative presynaptic short-term plasticity proposes release-probability learning rules that enhance onset coding and recurrent replay. A published simulation study finds that excitatory-only connectivity can amplify sequential assembly replay. What is new is joint attention to how each layer shapes the representation, from population noise to synaptic learning. The links are thematic rather than methodological, since the works rely on different model classes and analytical approaches.