Training a neural network with Chaotic data

i have a set of differential equations which i have solved numerically. The solution of the differential equations is chaotic in nature. I need to train my neural network with this data below 20,000 input data for the time variable 20,000 output data which will be solution of the differential equations i.e x(t) but when i am training my neural network it is not giving me results. Is it not possible to train a neural network with chaotic data, or can it be done by some other generalization of the neural network