Skip to content

Conversation

@CIGbalance
Copy link
Contributor

@CIGbalance CIGbalance commented Feb 10, 2025

This is a first draft for my task in #4, relating to the research questions that can be answered by comparing across different functions. We agreed that I would first add something to the webpage (which is this PR, and then we could potentially include similar groupings in cocopp).

Before converting this to a full PR, I would like to discuss the following:

  • Do the visualisations make sense? I figured for some points (like separability), it is nice to see. But I am currently only showing 2D, not sure if that is too simplistic
  • Do the headers make sense? They are mostly just what is already in the documentation, but not entirely. I think we should also add more details to a lot of them, but that is secondary for now.

row-1-column-1
row-2-column-1
row-3-column-1
row-4-column-1
row-5-column-1

@CIGbalance CIGbalance changed the title Research questions Research questions to function mapping Feb 10, 2025
@nikohansen
Copy link
Contributor

nikohansen commented Feb 10, 2025

I think it's currently difficult to understand that there are two semantic approaches and (only) the second compares two cases (or investigates a specific change).

The Rosenbrock function should not be characterized as separable or an example of exploitation of separability. One could say that we test whether a dependency on the given coordinate system is exploited. The 2-D case isn't likely to reflect the question in this case too.

@CIGbalance
Copy link
Contributor Author

I think it's currently difficult to understand that there are two semantic approaches and (only) the second compares two cases (or investigates a specific change).

That is a very good point. Just to clarify, you mean the difference between (1) observing optimal convergence by looking at the sphere function only, vs (2) comparing the behaviour across 2 similar functions, right? I'll think about how to make this clearer. I am in generally not entirely happy with the presentation, but haven't thought of a good alternative yet.

The Rosenbrock function should not be characterized as separable or an example of exploitation of separability. One could say that we test whether a dependency on the given coordinate system is exploited. The 2-D case isn't likely to reflect the question in this case too.

Good point, I'll clarify 👍 (at some point, when I return to this 😇 )

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants