Bubbles Sampling 16 Parameters dataset

Published: 28-05-2020| Version 1 | DOI: 10.17632/37zzz2kfvm.1
Contributor:
Christos Lemesios

Description

In our paper (to be submitted) we investigate the capacity of a new sampling method, "Bubbles Sampling". This dataset is complementary to the submission in order to reproduce some of the results presented in the paper. Under the investigation of the capacity of the algorithm to identify and extract the parameter values correctly in an optimization problem, we created a toy model, a Tilted Mexican hat. We compare our results with Brute Force Uniform Sampling to assess the algorithm's robustness and its efficiency as opposed to an adaptive Metropolis-Hastings. We provide the data for the most complex Mexican-Hat case we present in the paper for the reproducibility of the results. The data are compressed in a python ".npz" format and need to be decompressed to be used in the attached software. The Bubbles dataset are a 2-D array with the row being the solutions in the AMIAS ensemble of solutions and columns 0-15 the parameters of the model. Column 16 is the chi-squared value of each solution, column 17 is the Bubble ID, column 18 being the MC step, and column 19 the assigned weights as described in the paper. The format is for the Adaptive Metropolis-Hastings and Uniform Sampling, without the last "weights" column. With the data we provide the steps to extract and reproduce our results in the attached software.

Files

Steps to reproduce

Read and run the file CLASS_AMIAS_MENDELEYDATA.py