MOO with NaN outcomes
See original GitHub issueHi,
I have a multi-objective problem where one of the objectives is only defined when the other one is larger than 0. I’m using the service API and noticed that if I don’t pass the NaN objective in these cases the model does not complain and in exp_to_df
they show up as NaN as intended.
I wanted to make sure that this is causing no unexpected behaviour. How are the NaNs handled in the model? Are they ignored?
Thanks for the great work!
Issue Analytics
- State:
- Created 2 years ago
- Comments:8 (5 by maintainers)
Top Results From Across the Web
washing my hair with rice water..what will the outcome be
Missy Moo and Christmas Crafts at the zoo ! · my chicken crossaint!!! · christmas cookies !! #christmas · just a chit chat...
Read more >How to Deal With NaN Values - Datatest
NaN values represent undefined or unrepresentable results from certain mathematical operations. Mathematical operations involving a NaN will either return a NaN ...
Read more >The 26th Hahn Moo-Sook Colloquium in the Korean ...
The 26th Annual Hahn Moo-Sook Colloquium in the Korean Humanities November 9&10, 2018 “Emotion, Culture, ... Nan Kim, University of Wisconsin Milwaukee.
Read more >an observational study of a regional elimination programme
The main outcome was the change in village incidence of clinical P falciparum malaria, quantified using a multivariate, generalised, additive multilevel ...
Read more >Fe3O4/Ag/Bi2MoO6 Photoactivatable Nanozyme for Self‐ ...
Fe3O4/Ag/Bi2MoO6 Photoactivatable Nanozyme for Self-Replenishing and ... enabled and sustained in TME for outstanding therapeutic outcomes.
Read more >Top Related Medium Post
No results found
Top Related StackOverflow Question
No results found
Troubleshoot Live Code
Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start FreeTop Related Reddit Thread
No results found
Top Related Hackernoon Post
No results found
Top Related Tweet
No results found
Top Related Dev.to Post
No results found
Top Related Hashnode Post
No results found
Top GitHub Comments
Hmm having NaNs in the data should certainly cause issues if you are using Bayesian optimization because I do not believe we do any filtering of NaNs. Are you using Sobol search? Can you share an example?
Okay, thanks for the explanation!