Convergent cross mapping algorithm from:
Sugihara, George, et al., Detecting Causality in Complex Ecosystems, Science 26 October 2012, Vol. 338, no. 6106, pp. 496-500.
for detecting causality.
+ L and M causality.
SugiLM - predict values based on historical data
SugiLM1 - predict values based on all available data
Sugi - same as SugiLM1, but without L and M (faster than SugiLM1)
Jozef Jakubik (2020). convergent cross mapping (https://www.mathworks.com/matlabcentral/fileexchange/52964-convergent-cross-mapping), MATLAB Central File Exchange. Retrieved .
thanks for ur sharing ,but why my convegent map shows downward and volatility,my data form is daily,please reply me email@example.com,thank u
Sugiharas CMM correlation for estimate of X and original X in coupled case
Sugiharas CMM correlation for estimate of Y and original Y in coupled case
Sugiharas CMM correlation for estimate of X and original X in non-coupled case
Sugiharas CMM correlation for estimate of Y and original Y in non-coupled case
why do you say couple and non-couple, I observed that the methods invoked are the same, please mail to me (chenxofhit[at]gmail.com)Jakubik, waiting for your reply! Thanks very much.
SugiLM uses for the manifold just historical data (older than the actual point), SugiLM1 reconstructs the manifold from all available data.
E depends on the system. A good approximation of E could be the correlation dimension or lyapunov exponents. Also, you could consider E as a hyperparameter, try different values for it and choose the best one.
tau is data dependent. To get tau, it is best to explore different ones and choose the best one for you.
Thanks for you sharing this codes! But how to get the optimal parameters tau and E? And what's the difference between historical data and all available data?
Thanks for sharing!
What is the difference between the codes
and b) SugiLM1
thanks for your share.
corrected typo in example function
add Sugi (see description)
+ algorithm modification
+ function CCM