Connectome reveals cerebellum-like circuit that rapidly cancels predictable sensory input
A Nature study maps cell types and synaptic connections in a cerebellum-like electrosensory circuit of electric fish, showing inhibitory and disinhibitory pathways, structured feedforward and recurrent connectivity, and multi-site plasticity that together solve a continual-learning problem by cancelling predictable sensory input while preserving unpredictable signals; a constrained computational model demonstrates how GC, MG, and SG plasticity and MG recurrence cooperate for fast, accurate, noise-robust predictions.