Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Shapley values support for Ensemble Models #6698

Closed
wendycwong opened this issue Apr 15, 2023 · 0 comments
Closed

Shapley values support for Ensemble Models #6698

wendycwong opened this issue Apr 15, 2023 · 0 comments
Assignees
@tomasfryda
Labels
enhancement
Milestone
3.44.0.1

Comments

@wendycwong
Copy link
Contributor

This is a customer request: https://support.h2o.ai/a/tickets/104466

Tasks probably need to be broken into several github issues:

a) Compute and provide SHAP contribs for non- tree based models.
b) Compute and provide SHAP contribs for ensemble models.
@tomasfryda tomasfryda added this to the 3.44.0.1 milestone Jun 26, 2023
@tomasfryda tomasfryda mentioned this issue Oct 2, 2023
Merged
tomasfryda added a commit that referenced this issue Oct 10, 2023
@tomasfryda @hannah-tillman
GH-6698: Shapley values support for ensemble models (#15734)
75c20c3 
* Shap for GLM

* Add bSHAP for GLM

* Interventional treeSHAP support

* GLM related edits

* Initial DeepShap

* DeepSHAP classification

* DeepSHAP for tanh and relu with dropout

* DeepSHAP with MaxOut

* Add output_space option

* Add Stacked Ensemble support

* Fix stacked ensemble with metalearner transform

* Add automl tests

* Add aggregation to get the SHAP not just per reference

* Fix tests

* Fix python tests

* Fix python tests

* Enable use in R client

* Add R and basic R support

* Add python explain support

* Add explain support and tests

* R formatting around equal sign

* Add java tests

* Add minimal documentation

* Fix different distributions with output_space=True

* Fix tweedie GLM link in python test

* Check for memory

* Fix multi-node issue and improve memory usage

* Incorporate Veronika's suggestions

* Improve tests

* Make java tests less strict (eps = 1e-8 -> 1e-6)

* Fix R tests/cran checks

* Improve parallelization for SE when there is just a small bg set + documentation

* Skip NOPASS tests in the Py3.7 Changed Only stage

* Unify error messages when no background frame is provided

* Add tests with pregenerated DeepSHAP contributions using pytorch+shap

* Add caveat about original output format to doc

* Fix formatting in documentation

* Remove duplicate 'between' in the doc

* ht/added references & readability updates

* Make R tests less strict to make tests more stable

* Fix hex/genmodel/algos/tree/ContributionsPredictorTest.java test by "implementing" a missing method

---------

Co-authored-by: Hannah Tillman <hannah.tillman@h2o.ai>
mn-mikke added a commit that referenced this issue Oct 12, 2023
@mn-mikke
[GH-6698][FOLLOWUP] Fix build of DeepLearningSHAPTest on Java 13+
b01f42a
@mn-mikke mn-mikke mentioned this issue Oct 12, 2023
Merged
mn-mikke added a commit that referenced this issue Oct 12, 2023
@mn-mikke
[GH-6698][FOLLOWUP] Fix build of DeepLearningSHAPTest on Java 13+ (#1…
096f69d 
…5818)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@tomasfryda tomasfryda

Labels
enhancement
Projects
None yet
Milestone
3.44.0.1
Development

No branches or pull requests
3 participants
@mn-mikke @wendycwong @tomasfryda