grzegorzmikaagh/towards-adaptivity-via-a-new-discrepancy-principle-for-poisson-inverse-problems

Morozov discrepancy principle in statistical inverse problems with Poisson noise

Towards adaptivity via a new discrepancy principle for Poisson inverse problems

Simulation code accompanying the article G.Mika, Z.Szkutnik, "Towards adaptivity via a new discrepancy principle for Poisson inverse problems".

The replication of simulation results is possible via a provided set of codes. For each of considered methods (Landweber, 1- and 2-times iterated Tikhonov and TSVD) and for each presented function (Beta, Bimodal, SMLA, SMLB) there is a separate code file with the following naming convention: FunctionName_MethodName.py. The list of requirements is provided in file requirements.txt. To execute the code a minimal required version of Python is 3.7.

Each code file produces a set of output files in a format of csv files with following naming convention: MethodName_fun_FunctionName_size_SampleSize_tau_ParamterTau.csv, where SampleSize and ParamterTau are customizable parameters and can be edited inside a respective code files. By default, they are set to the values used in the article.

For Linux-based systems, a shortcut is provided in a format of shell file (run_simulations.sh). The code expects to be executed inside a repository folder and in a proper python environment with the required packages installed.