Elex 4549 add model flexibility #109
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…CV branch for testing purposes
| @@ -5,7 +5,7 @@ | |||
|
|
|||
| INSTALL_REQUIRES = ( | |||
| "click>=8.1", | |||
| "elex-solver>=2.1.1", | |||
| "elex-solver @ git+https://github.com/washingtonpost/elex-solver.git@updating-residuals", | |||
There was a problem hiding this comment.
I changed this temporarily so we can test / run the tests here in github 🎉
| model.fit(x, y, weights=weights, taus=0.5, **model_kwargs) | ||
| else: | ||
| raise BootstrapElectionModelException( | ||
| f"Please use defined model type 'QR' or 'OLS', input is: {self.model_type}" |
There was a problem hiding this comment.
This is such a polite and helpful error message! 😄 🎉
| model_kwargs = {"fit_intercept": True, "regularize_intercept": False, "n_feat_ignore_reg": 1} | ||
| if self.model_type == "OLS": | ||
| # if we run OLS we use cross validation to find the optimal lambda | ||
| # TODO: why do we not do this for other models also |
There was a problem hiding this comment.
lol wait, why do we not do this for other models also? 🤔
Is this a "how can I write code to make this happen" question or a theoretical question (or both)? 🤔
| y_train, | ||
| weights_train, | ||
| aggregate_indicator_train, | ||
| K=5, |
There was a problem hiding this comment.
Just curious, how did you pick the default of 5 here (and elsewhere)? I mean I've done (and seen) 3-fold, 5-fold, 10-fold CV but usually my choice there depends on the number of examples in the data set and how much computing power I have 🤔
|
|
||
| # we cannot just apply ols_y_B/old_z_B to the test units because that would be missing | ||
| # our contest level random effect | ||
| # TODO: DO WE NEVER DO ANYTHING WITH THE OUTPUT |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Instead of being stuck with OLS as our base model for the bootstrap model, this PR gives us flexibility to use different models. We were wed to OLS because we had a quick way of computing the leave-one-out-residual, which we needed for the bootstrap, since the training residual would be biased towards zero. We now use k-fold cross validation to get an estimate for the leave-one-out residual (k-fold residual will be greater than or equal to loo residual, so this is if anything more conservative). We also add the ability to play with OLS vs. QR models. In the future, we may expand the models that are being used here.
This needs this branch of elex-solver, which allows us to calculate the k-fold residual. Tests will fail until this branch is merged/released.
Jira Ticket
https://arcpublishing.atlassian.net/browse/ELEX-4549
Test Steps
vs