The .zip file contains the code required to reproduce the plots in the paper "A study of uncertainty quantification in overparametrized high-dimensional models" ( https://arxiv.org/abs/2210.12760 ).
The code is organized as follows
* The "gcmpyo3" folder contains the Rust library that solves the state evolution equation to compute the performance for the different estimators. Run install.sh (at the root) to compile the library and generate the Python bindings.
* The "core" folder contains code to run the different estimators : the subfolder "erm" is used to compute logistic regression and the "amp" subfolder is used to run the Bayesian estimators (empirical Bayes and Bayes optimal) as described in the paper.
The "core" and "gcmpyo3" folders are independent are can be used independently. They are both used in the following files / folders : 
* The "experiments" folder contains notebooks to compute the empirical values for each estimator (bayes optimal, erm, pseudo-Bayes), and checks that the overlaps (m, q, etc.) are in agreement with the theoretical predictions of the state-evolution equations.
* The state_evolution.ipynb notebook produces the main Figures of the paper, using the state evolution equations implemented in the "gcmpyo3".