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Outdated suggestions cannot be applied.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.
Motivation and context:
Currently
Sparse
transforms are only usable when they are manually specified. This PR allows certain solvers such asLstsqL1
andLstsqDrop
to signal to the backend that the returned weights should be implemented using a sparse representation. This currently only works forweights=True
. Whenweights=False
this continues to useDense
decoders.Interactions with other PRs:
Might conflict with some documentation changes in #1540.
How has this been tested?
Made sure an existing test is now using sparse weight matrices when
scipy
is installed. Added tests to each new warning.How long should this take to review?
Where should a reviewer start?
Start in
solvers.py
with the changes to existing solvers. Then see how this is handled in the connection builder.Types of changes:
Checklist:
Still to do:
I've done a bit of profiling to see when it makes sense to do this. On my Ubuntu machine with a
conda
+scipy
install and Python 3.6 I'm seeing that it helps when using aLasso
solver with< 20%
sparsity (number of non-zero coefficients) and1000 x 1000
weight matrices.Accuracy seems to be consistent with
LstsqL2
when within 10-20% sparsity: