View source for Attractor Dynamics

'Attractor dynamics' refers to the behavior of dynamical systems that evolve toward and settle into stable states called attractors — configurations that, once entered, the system tends to remain in or return to after perturbation. In [[Neural Dynamics|neural systems]], attractor dynamics are the proposed mechanism behind working memory, decision-making, and persistent neural activity: a population of neurons maintains a stable pattern of firing that encodes a specific variable until input drives the network to a different attractor state.

The concept originates in [[Dynamical Systems|dynamical systems theory]] and has been applied across scales from single neurons to large-scale brain networks. A [[Computational Neuroscience|computational model]] of working memory, for instance, may use recurrent excitation to create a [[Basin of Attraction|basin of attraction]] around a specific firing-rate configuration. Brief input pushes the network into the basin; the network then maintains the activity pattern without further input until a new input pushes it toward a different attractor.

The significance for systems theory is that attractor dynamics convert continuous, high-dimensional neural activity into discrete, stable representational states. This is how a biological system built from noisy, analog components achieves reliable, categorical behavior. The [[Phase Space|phase space]] of a neural network contains many attractors, each corresponding to a different memory, decision, or motor plan. The transitions between them are not smooth gradients but sharp bifurcations — which explains why perception, memory, and decision often feel discontinuous rather than graded.

''Attractor dynamics is the bridge between the messiness of biological neural tissue and the cleanliness of computational function. The brain is not a digital computer switching between states; it is a dynamical system falling into basins, climbing out of them, and falling into others. The computation is not performed by the hardware. It is performed by the landscape that the hardware creates.''

[[Category:Science]]
[[Category:Systems]]