Deprecate grid_scores_ everywhere, replace by cv_results_
See original GitHub issueIssue Analytics
- State:
- Created 5 years ago
- Comments:15 (12 by maintainers)
Top Results From Across the Web
How to Grid Search Hyperparameters for Deep Learning ...
This is a map of the model parameter name and an array of values to try. By default, accuracy is the score that...
Read more >Intro to Model Tuning: Grid and Random Search - Kaggle
The grid cross validation score increases over time. This indicates that whatever hyperparameters are changing in grid search are gradually increasing the score...
Read more >RandomizedSearchCV best_params_ changes everytime ...
My question is: Why do best_params_ change every time I run this program? I assume that best_params_ should always stay the same. Code...
Read more >Linear Models for Regression - Andreas Mueller
FIXME show how to get feature names out of complex pipelines FIXME remove isocontour slides ... The scores are R squared or coefficient...
Read more >Complete Guide to Parameter Tuning in XGBoost with codes ...
XGBoost on the other hand make splits upto the max_depth specified and then start pruning the tree backwards and remove splits beyond which ......
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 Free
Top 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
If you look at the documentation for GridSearchCV’s
cv_results_
and compare to the existinggrid_scores_
, you’ll see they’re very different. The former is a dict of arrays; the latter is a 2d array. The former includes information about how long each fit takes, but I’m not sure that’s relevant here. The former can also report scores under multiple metrics.@amueller, I’m not immediately sure what benefit we provide users by ditching the existing 2d array representation, unless we also support other features like multiple scorers. The use of warm start in specialised CV means we can’t really report timings either…
@hemant659, it’s much easier to look at your changes in a pull request than in a zip. Please open one when you are ready, but perhaps wait for us to clarify exactly what form
cv_results_
should take. Yes, the existing attributes would need deprecation.Well I think it’s sensible to provide a
cv_results_
-style interface, for consistency, but straight-out scores is also simpler for some use cases.