This folder contains the code for the aMR-PC toolbox by Ilja Kröker in the version used for the code in GALMAP_code. This toolbox was also used for Kröker et al. 2022 Link to current version of the toolbox here Data This folder contains inputs and simulated outputs of the CO_2 benchmark from here and referenced in Köppel et al. 2019 There are a set of 10000 random samples of inputs and corresponding simulated outputs, as well as sets of input and output samples that correspond to the roots needed for aMR-PC with different (Nr, No). Additionally the sub-folder 'DNN' contains the results of three DNN trained on the data from the benchmark. The specifics of the DNN are described in the paper. GALMaP Code aMRPCclass.py This file contains a class aMRPC, that makes use of the aMR-PC toolbox to implement an aMR-PC. The class can be used for creating and fitting an aMR-PC of specified polynomial degree on data, as well as refitting an existing aMR-PC on new data. data_load.py This file contains functions that load the data from the CO_2 benchmark contained in the folder "data". The functions are used in the file "run_iterative.py". GPE.py This file contains a variation on the Scikit GPE-implementation, which creates and fits a GPE with a standard RBF-kernel to specified data. It additionally returns the results on further, to the GPE unknown, data. PLOTS.py This file contains functions which create the figures used in the paper. This includes representations of the CO_2 dataset, as well as the results of the iterative and active learning variations of aMR-PC and GPE. Some of the functions rely on an import of the aMRPC class. run_iterative.py This file can be seen as the 'main'. It runs one of the three active learning based variations on given data and outputs the results in the form of files and figures, as can be found in the folder "results". The desired type of variation, polynomial degree of the aMR-PC and outputs can be easily changed by setting parameters at the start of the file. This code imports aMRPCclass.py, data_load.py, GPE.py and twostep.py twostep.py This file contains the calculations done in the main loop from run_iterative.py. Its function run_GPE runs a GPE on given samples and returns its approximation, as well as various error measures thereof. The approximation is depicted in figures created from the function plot_twostep. Results Results from the implementation in GALMAP_code can be found here. The results were run for the three variations GAL, GALMaP and GALMaPAT each for (Nr, No) = (1,1) and 100 iterations.

Python, 3.8