Abstract | ||
---|---|---|
We present a practicable procedure which allows us to decide if a given time series is pure noise, chaotic but distorted by noise, purely chaotic, or a Markov process. This classification is important since the task of modelling and predicting a time series with neural networks is highly related to the knowledge of the memory and the prediction horizon of the process. Our method is based on a measure of the sensitive dependence on the initial conditions which generalizes the information-theoretical concept of Kolmogorov-Sinai entropy. |
Year | DOI | Venue |
---|---|---|
1996 | 10.1007/3-540-61510-5_130 | ICANN |
Keywords | Field | DocType |
information theoretic measure,time series,markov process,neural network,initial condition | Information flow (information theory),Markov process,Pattern recognition,Computer science,Horizon,Artificial intelligence,Conditional entropy,Artificial neural network,Chaotic,Machine learning | Conference |
ISBN | Citations | PageRank |
3-540-61510-5 | 0 | 0.34 |
References | Authors | |
1 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Christian Schittenkopf | 1 | 55 | 6.95 |
Gustavo Deco | 2 | 1004 | 156.20 |