Robustness and quality of life improvements to Shared Sparsity contribution. #58
Conversation
This file contains hidden or 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
Updates to improve robustness and quality-of-life: -Addition of a max_norm constraint to avoid overflow/nonsparse outcomes given the nonconvex regularizer (involves a minor change to the optimization algorithm). This was a parameter that the theory had included, but we had omitted this from the original version since it only seems to matter in extreme circumstances where the hyperparameters are poorly chosen. There is also an auto option here for easy use, as well as none to ignore. -Addition of an auto option for the mcp_alpha parameter, since this may be somewhat mysterious to some. -Addition of an include_bias option defaulted to True. This subtracts off the mean of each group of covariates and outcomes (something we typically did outside the module), which is equivalent to adding an unregularized bias term to the regressions. -The replacement of the auto option for mcp_lambda with an option proportional_lambda, which, if set to True, interprets the value of mcp_lambda as a constant of proportionality to the previous auto option. This is because I found in practice, the coefficient of the auto option pretty much always needs to be tuned, but that the analytic expression it uses was extremely helpful otherwise.
edit parameter description
|
|
Fixes a bug where `proportional_lambda` was always set to `True` in `MCPSelector` regardless of the value passed to the actual user-instantiated `SharedSparsityConfounderSelection` model.
Member
|
Apologies for the late response, I was under the impression everything was ok and just waited for the right version change to merge and release. However, it sems these changes broke the existing tests. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
No description provided.