This repository contains notebooks, code and documentation related to my final project for the 2019/20 edition of the MHPC. The thesis will be hosted here, as well.
In my project I tried to get some insights into high-performance derivative-free optimization and exploit HPC for benchmarking various algorithms for the calibration of parameters of the GEOtop model. The topic is interesting both from a scientific and technical point of view; you can find more information here.
The code developed for the project consist of a GEOtop wrapper (GEOtoPy) and various notebooks and scripts, which leverage third party libraries such as Nevergrad for optimization and SALib .
At present, the content of this repository is somewhat provisional and exploratory in nature. It contains preliminary analysis and experiments with the interface and the optimizer(s). At the end of the project, it will host an application to a case study and reports.
Nonetheless, since the very beginning, a special effort has been made to provide the code and documentation to setup an
easily reproducible environment. Once you have conda on your system, you can install the environment with
just conda env create -f environment.yaml from the root of this repository.
Distributed under the GPL3 License. See LICENSE for more information.