Feedback and Questions from JOSS Paper Review

[giovannifereoli]

Hi everyone - thanks for the great work. I had a few comments and questions while going through the JOSS manuscript:

• Line 45: maybe clarify the phrasing - something like “without additional memory allocation” might read more cleanly.
• Line 47: when you say “several solvers,” could you specify which ones you are referring to?
• Line 96: it would be helpful to see the same plot/analysis also for the constrained case.
• Line 103: could you expand on what is meant by “very encouraging”? The performance is close to Ipopt, so I’m curious about the specific reasons a user would choose this package and what drives the comparable results.
• Figure 1: the axes are a bit unclear - an explanation of what each axis represents would help. Also, the plot quality could be improved (SVG, EPS, or PDF would be ideal).

Thanks again — happy to clarify anything if needed!

[giovannifereoli]

JOSS Review

[tmigot]

Thanks for the feedback @giovannifereoli . (I took the liberty to modify your comment to add a tick-list)

For the bound-constrained benchmark, in order not to charge the paper too much, we decided to add a variant in the documentation: https://jso.dev/JSOSolvers.jl/latest/benchmark/ . Let me know, if you think that's sufficient.

The results are encouraging because TRUNK and LBFGS were clearly faster on the performance profile for solving approximately the same amount of problems. It's also encouraging considering that the Ipopt has been carefully optimized over the years. One of our short term goal is to refine algorithmic hyper-parameter, which we expect to improve further the quality.

I am opening a pull request for the other comments #327 .

[giovannifereoli]

Of course - thanks for adding the checklist!

1. Sounds good to me. I think it’s clear, though it might be worth adding a short sentence in the paper mentioning that these additional benchmarks are available.
2. Got it! I think it would help to expand that part of the paper a bit - maybe just a slightly longer explanation of why the results are described as “encouraging.”

[giovannifereoli]

Updated the checklist!

[tmigot]

Hi @giovannifereoli I think #332 close the last item on the list. Can we close this issue ?
Thanks again for the feedbacks :).

[giovannifereoli]

Yes, of course! Thank you guys!