This repository contains a source files for all figures, simulated data, parameter fits and all source code used in the research project
Signatures of criticality arise from random subsampling in simple population models
You can find statistical methods and algorithms used in this projects as a toolbox that we hope can be of use for other projects that analyse multivariate binary datasets.
For reasons of transparency, this repository is a direct copy of the code used internally within the 'criticality' project, complete with the full and raw submission history of all commits. Given that, the code (found in the "code" folder) uses absolute pathnames and is included as it was run on our workstation and on our compute-cluster. It is likely that it will not work ‘out of the box’.
We have additionally uploaded the simulated data, source files with any data that went into any figure as well as scripts for generating raw versions of all figures. The figure scripts were updated to work with the file structure and should run out of the box when exectued from within the "figures" folder.
The data and results files were uploaded using Git Large File Storage (https://git-lfs.github.com/). Downloading the .mat and .gc files from this repository requires to install Git LFS.
During the project, we generated long Markov Chain Monte Carlo runs that were used to gauge the gain in sampling effeciency obtained from Rao-Blackwellizing the Monte Carlo sampler. This repository only contains data on the quantified gain (i.e. the root mean square errors (RMSEs) referred to in the manuscript) computed from the sampled Monte Carlo chains, but not the raw Monte Carlo chains themselves (~4GB).
This repository also contains the code used to generate the activity of our simple simulated population of retinal ganglion cells, found under "data/data_generation/". This code can be readily used to generate additional simulated activity.