ADVANCED CHAOS FORECASTING

The exponential separation of initially adjacent trajectories restrict s the predictability of deterministic chaotic motions. The predictabil ity depends on the initial state from where the trajectory starts that shall be forecasted. By calculating the predictability simultaneously with the forecast, we are able to reject forecasts with low reliabili ty immediately, thereby decreasing drastically the average forecast er ror. We test this scheme experimentally on Chua's circuit [Komuro, Tok unaga, Matsumoto, Chua, and Hotta, Int. J. Bifurc. Chaos 1, 139 (1991) ], basing all calculations only on a time series of a single scalar va riable.