[SYSTEMDS-3928] Second-order Builtin Function for Independent Subnet Training#2427
Closed
arnobock wants to merge 6 commits intoapache:mainfrom
Closed
[SYSTEMDS-3928] Second-order Builtin Function for Independent Subnet Training#2427arnobock wants to merge 6 commits intoapache:mainfrom
arnobock wants to merge 6 commits intoapache:mainfrom
Conversation
Contributor
|
Thanks for the great contribution @arnobock. During the merge, I only replaced the hard-coded reference to the non-existing mnist-train and disabled the long-running test. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR implements a first version of a builtin realising independent subnet training [1] as DML script.
[1] https://www.vldb.org/pvldb/vol15/p1581-wolfe.pdf
The implementation provides:
As hinted this version needs further runtime optimisation to enhance performance.
A current determined bottleneck is the gradient computation, which (even though executed through a parfor-loop) remains rather slow. One potential reason for this could be the sparse masking of disjoint subnets since dimensionalities of the model matrices are effectively not reduced.
The code contains some TODO's which hint at potential performance boosts.
Note: A dense masking which reduces subnet matrices size is currently in progress and could solve the remaining issues.