Tom Ragonneau

Hi @Katherine-Gilber, sorry for the delay. May I know which operating system you are using? Also, could you please run ```bash pip uninstall pdfo pip install pdfo --no-cache-dir ``` and...

Hi @MJDRobbins, The conda-forge stating of PDFO is not maintained by us. Could you please try installing PDFO using ``pip`` (also available in Anaconda)? Thank you, Tom.

It now compiles, but MATLAB crashes every time I run `nomadOpt`. More precisely, I tried ```matlab nomadOpt(@(x)x^2,1) ``` Did I misunderstood the syntax and broke the MEX-file? Thanks :+1:

I understand. The tests runs on my machine, and produces ``` ------------------------------------------------------------------ This is NOMAD v4.2.0 Authors: C. Audet, S. Le Digabel, V. Rochon Montplaisir and C. Tribes MEX Interface...

> Agreed, it would be great to see the result of those. @ragonneau you can find those in `benchmarks/benchmarks/optimize.py`. It would be very useful to see the results of those...

> It looks like something went awry with the named problems (`BenchSmoothUnbounded`)? COBQA's objective value is always _worse_. Or are the other algorithms really always better on smooth problems? I...

> @ragonneau I suggest you edit your results. They are difficult to read. In addition, there seems no reason to use a table different from the other one. Thanks. I...

> Based on the benchmarks, I would consider COBYQA a great improvement over COBYLA and for our derivative-free solvers in general. Also it should be noted that currently COBYLA is...

> It depends a bit on plans I'd say. E.g., is the standalone package expected to continue to be developed for years into the future? Or once it's stable in...

> +1 to what Robert said on using relative imports and avoiding `importlib`. That's not even much of a compromise I'd say; relative imports within a package are idiomatic. I...