Dynamical archetypes identify attractor structure in neural circuits

Neuroscientists call circuits integrators or compasses, yet no physical system matches such an idealisation exactly and there has been no general way to test the claim against recordings. Ábel Ságodi presents a method that scores each mechanism in a catalogue of dynamical archetypes by its forecast and by the deformation needed to match a circuit’s dynamics, identifying ring attractors in mouse and fly head-direction circuits and task-trained networks and a line attractor in the hypothalamus, enabling hypothesis-driven identification of the computation a circuit implements across circuits and species.