MF2: Multi-Fidelity Functions

[sjvrijn]

Submitting Author: Sander van Rijn(@sjvrijn)
Package Name: mf2 One-Line Description of Package: A collection of multi-fidelity analytical benchmark functions Repository Link (if existing): https://github.com/sjvrijn/mf2

---

Description

• Include a brief paragraph describing what your package does:

This package provides a number of multi-fidelity function implementations commonly used to benchmark the performance of multi-fidelity (surrogate-model-assisted) optimization algorithms. It provides a common and simple interface to all included functions, with corresponding references available in the documentation.

Scope

• Please indicate which category or categories this package falls under:

  • [ ] Data retrieval
  • [ ] Data extraction
  • [ ] Data munging
  • [ ] Data deposition
  • [ ] Data visualization
  • [x] Reproducibility
  • [ ] Geospatial
  • [ ] Education
  • [x] Unsure/Other (explain below)

• Explain how the and why the package falls under these categories (briefly, 1-2 sentences). Please note any areas you are unsure of:

Analytical benchmark functions are a like well-established datasets in the field of machine-learning: by comparing your algorithm on the same ones, you can accurately compare your results. They are essential for reproducibility, as typos in the formulas will obviously give very different results, but are not an 'external' utility like the examples given under Package Categories in the guidebook.

• Who is the target audience and what are scientific applications of this package?

The target audience is any scientist working in the field of multi-fidelity optimization algorithms, as they will likely want to use functions from this collection to benchmark the performance of their algorithms.

• Are there other Python packages that accomplish the same thing? If so, how does yours differ?

I was not aware of any when writing the package. Since recently, there seem to be two packages emukit and smt that also contain a partially overlapping set of the same functions. Those packages however are mostly focussed on providing algorithms to apply on these kinds of problems, while my mf2 is a separate package consisting of just functions.

In my opinion, that makes mf2:

1. easier to find when searching for function implementations,
2. easier to include when working on new or different methods, and
3. better tested as mf2 has dedicated and comprehensive tests to verify the functions (property and regression tests)

• Any other questions or issues we should be aware of?:

None

[lwasser]

thank you for this submission @sjvrijn our next pyopensci checkin in is tomorrow at 11am mountain time! we will discuss this package then. you are welcome to join us as well as it's an open meeting!!!

[sjvrijn]

As discussed in the December 5 meeting, this submission is withdrawn/on hold

See the meeting notes for details.

[lwasser]

hey @sjvrijn thank you for closing this and for understanding!! I hope that you still are interested in being involved with us. We just don't have the capacity yet to review this particular submission BUT i do suspect that in the future we may have that capacity!!

[sjvrijn]

@lwasser No problem, and yes I'm still interested. How do I sign up as a reviewer/supporter?

[lwasser]

awesome! and funny you should ask!! i finally created a good form for this very purpose. You can fill it out here -- https://forms.gle/5MuYMVs8xpCxk9w29 and please let me know if you have any issues with filling it out or if there are other "categories" of domain expertise that i am missing!! i am sure i've missed some important ones.