Conversation
initial commit for output space exploration * new module for output space exploration build on platypus-opt * new example with the lake problem * modifications to optimization.py to make new module work with existing optimization functionality * updates to TODO outlining further possible cleanup of optimization.py
seems unnecessary because archive is already non dominated for e-NSGAII and is problematic for OutputSpaceExploration
basic version now works properly. HitBox as per Plos paper, but also behaves as epsilon archive. per epsilon box, the most central solution is kept. Tournament selection is used based on novelty. Novelty is 1/number of seen solutions in hitbox
added link to Cherel paper and outlined the rest of the algorithm.
|
Looks awesome! I will review it from a code-quality and user perspective tomorrow. |
|
I started on a review, in general, it looks quite good! Reviewing the docs on the other hand is quite a challenge without them begin built. Could you check if it's easy to enable doc building (see #171) before I review them? |
|
I have enabled the building of docs for pull requests. I guess we need one more commit here to trigger it. |
|
If you can find the time, cleaning up you commits would be a great way to also trigger the CI! :) Otherwise I will do some ci-trigger-magic when I review it (probably tomorrow). |
* new module for output space exploration build on platypus-opt * new example with the lake problem * modifications to optimization.py to make new module work with existing optimization functionality * updates to TODO outlining further possible cleanup of optimization.py * unit tests for all new code
|
That review app looks nice and the suggestions will indeed make the notebook a lot better. I'll try to make the suggested changes over the coming days. |
|
I have updated the notebook broadly in line with your suggestions. Note that I don't intend for these notebooks to be a tutorial on MOEAs so I have kept the additional text short but with links to relevant literature in case you want to know more. I also added a small check to ensure the number of items in |
initial commit for output space exploration * new module for output space exploration build on platypus-opt * new example with the lake problem * modifications to optimization.py to make new module work with existing optimization functionality * updates to TODO outlining further possible cleanup of optimization.py
|
I cleaned the commit history again, so if you agree with the notebook update, this is ready to be merged. |
EwoutH
left a comment
There was a problem hiding this comment.
Code quality and structure looks good, I didn't encounter strange things.
Docs are also coming along nicely, I added a few comments on things I found unclear or ambiguous.
If you take a look at those, could you let me merge afterwards? I will squash it into a nice tidy commit.
|
I think this is now ready to be merged |
|
go ahead |
|
@quaquel It's in, congratulations on an amazing new feature! |
adds output space exploration to the workbench.
Closes #154.