maximally_predictive_states

This repository contains the main scripts for the maximally predictive state space reconstruction and ensemble dynamics modelling presented in

Costa AC, Ahamed T, Jordan D, Stephens GJ "Maximally predictive states: from partial observations to long timescales" Chaos 2023

The data for reproducing the figures can be found in . We provide a run through all the steps on the analysis in two model systems in the folder ./ExamplePipeline

For a follow-up application in C. elegans postural time series, check this repository and the corresponding publication.

Any comments or questions, contact antonioccosta.phys(at)gmail(dot)com. Also, suggestions to speed up the code are more than welcome!

---

---

The code is fully written in python3, and we use of the following packages:

• h5py '3.0.0'
• sklearn '0.23.1'
• matplotlib '3.3.4'
• msmtools '1.2.5'
• scipy '1.3.1'
• numpy '1.17.2'
• joblib '0.13.2'
• cython '0.29.23'
• findiff '0.9.2'