consistently avoid dense matrix conversion for glmnet(x = ...) #1315
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
currently, some of the functions for the glmnet model check if the training data is a sparseMatrix, and some don't. the result is that initial operations in train() might succeed, and then later in the workflow, a step will fail (usually with "Cholmod error 'problem too large'" for a sparseMatrix with very large dimensions) when some of the training data is inadvertently converted to a dense Matrix. For instance, this currently would happen when prob() in glmnet.R is called (if trainControl(classProbs = TRUE)), or if tuneLength is used instead of tuneGrid in train(), because tuneLength = ... triggers a call to grid() in glmnet.R which does not do a check for a sparseMatrix. this is essentially a more complete version of the fix for topepo#1096
|
@powerpak thanks for making this change |
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Outdated suggestions cannot be applied.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.
(This is a more complete version of the fix submitted for #1096.)
Currently, some of the functions for the
glmnetmodel check if the training data is asparseMatrix, and some don't. The result is that initial operations intrain()might succeed, and then later in the workflow, a step will fail (usually with "Cholmod error 'problem too large'" for asparseMatrixwith very large dimensions) because some of the training data is inadvertently converted to a (impossibly large) denseMatrix.For instance, this bug currently occurs whenever
prob()inglmnet.Ris called (which happens iftrainControl(classProbs = TRUE)), or iftuneLengthis used instead oftuneGridfortrain(), becausetuneLength = ...triggers a call togrid()inglmnet.Rwhich does not check for a sparseMatrix before executingMatrix::as.matrix().