Pm. Alsing et al., USING NEURAL NETWORKS FOR CONTROLLING CHAOS, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics, 49(2), 1994, pp. 1225-1231
A feed-forward backpropagating neural network is trained to achieve an
d maintain control of the unstable periodic orbits embedded in a chaot
ic attractor. The controlling algorithms used for training the network
are based on the now standard scheme developed by Ott, Grebogi, and Y
orke [Phys. Rev. Lett. 64, 1196 (1990)], including variants that utili
ze previous perturbations and/or delayed time-series data.