search

notsebastiano / GP_algorithm

Implementation of the Grassberger-Procaccia algorithm to estimate the Correlation Dimension of a set of points
14 stars 10 forks source link

data & data_std not defined #2

Closed yeahjack closed 2 years ago

yeahjack commented 4 years ago

Hi, I was using your script to implement the GP algorithm, but I found variable "data" not defined in function td_embedding and variable "data_std" not defined in function grassberg_procaccia, and I do not know what these mean, please help!

ricardofrantz commented 2 years ago

hey, any update here?

yeahjack commented 2 years ago

nope... haven't got any feedbacks

notsebastiano commented 2 years ago

Hi, sorry I did not maintain this, I'll be updating this by tomorrow

ricardofrantz commented 2 years ago

Hi Sebastiano,

I am also new to this field and have used the routine according to this reference https://hess.copernicus.org/articles/22/5069/2018/. It would be great if you could prove a functional example and if you could outline some comments on the choice of parameters (and maybe a reference). As far as I understand, when finding the Hausdorff dimension, we have to loop over a range of, say, m=1..10 until a reasonable convergence of d is reached.

Thank you for sharing your work with us, R

On 18 Jan 2022, at 10:34, Sebastiano @.***> wrote:

Hi, sorry I did not maintain this, I'll be updating this by tomorrow

— Reply to this email directly, view it on GitHub https://github.com/notsebastiano/GP_algorithm/issues/2#issuecomment-1015228914, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABAGNTZXATL6M4M3LHOQNFTUWUX2XANCNFSM4TAPX2HQ. You are receiving this because you commented.

notsebastiano commented 2 years ago

Hi @ricardofrantz , Hi @yeahjack I have updated the code and solved the issue with data_std.

I have also added a Notebook with a functional example on the Lorenz attractor. I use that dynamical system to first obtain a timeseries of scalars (as you would have from the real world in absence of the knowledge of the underlying deterministic dynamical model). Then in the example the GP algorithm is used to calculate the correlation dimension of the reconstructed phase space for different embedding dimensions (refer to takens delay embedding) and you will se that after emb_dim = 3 the dimension is constant. This as an indicator that the dynamical manifold is unfolded as well reconstructed in 3 dims

Feel free to let me know if further problems arise, or doubts about the algorithm! Sebastiano