This test performs the test for chaotic dynamics of a noisy time series based on the Lyapunov exponent. The input is a vector of observed time series which can be stochastic or chaotic, usually time series has noise, so this code tests the positivity of the Lyapunov exponent based on a neural net approximation of the hidden chaotic map. This test computes the Lyapunov exponent using the Jacobian method without any need to specify an ODE or a suspected map giving only a vector of observation.
For a detailed information see my papers: http://www.sciencedirect.com/science/article/pii/S0960077913001185 and
http://www.sciencedirect.com/science/article/pii/S0096300313011211

Added some detailed information in file description.

18 Mar 2013

Change the default activation function to hyperbolic tangent, since the logistic function is not suited for Noisy time series. Moreover, make the code compatible with the latest version of MATLAB.