During the UNCECOMP 2019 conference, the minisymposium “Surrogate models: benchmark problems and solutions” offered not only interesting talks but also lively discussions regarding benchmark models for UQ and surrogate modelling purposes and how to be shared among the members of the UQ community. In my opinion, UQWorld is the venue to do so. Therefore, I would like to share with you the following:
The “Virtual Library for Simulation Experiments” of the Simon Fraser University (SFU) (https://www.sfu.ca/~ssurjano/index.html) offers a number of test functions and datasets for UQ, optimization, prediction, or other studies. MATLAB and R implementations are available for most models.
In my Github page (https://github.com/dlouk/UQ_benchmark_models), I have collected a number of models that I have used and still use to test my own UQ codes. All models are given in Python.
Indeed, UQWorld can be a platform for us to share such resources.
Thanks for the links, especially to your repository! As far as I know, a collection of UQ benchmark problems implemented in Python are still missing. In UQWorld, we have some MATLAB implementations, while in Virtual Library for Simulation Experiments there are MATLAB and R implementations of the benchmark problems.
Are you going to maintain (and from time to time) update the repository?
Hi @damarginal, thanks for the warm welcome. I do maintain and update my repositories, according to my current research activity. In case you want to copy the models to a central repository related to UQWorld, I would not have any problem with that as well.
That’s cool! But, I was thinking about, maybe, making a pull request to your repo, say for new benchmark problems implementation in Python. This way, we open a door for contributions from the community (if any; well, I for one would do it ) while centralizing the location of the benchmark problems. What do you think about that?
) while centralizing the location of the benchmark problems. What do you think about that?