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

Equations to define causal graph including causal mechanism #1066

Closed
wants to merge 60 commits into from
Closed

Equations to define causal graph including causal mechanism #1066

wants to merge 60 commits into from

Conversation

bhatt-priyadutt
Copy link
Contributor

  • This feature will enable user to write custom equations for each and node and get a causal model back with causal mechanisms assigned.
  • The Usage is mainly targeted for when the function/relationship model between nodes is known and allows user to specify it in a equation form as demonstrated below -
X = empirical()
Y = 12*exp(X) + halfnorm()
Z = 3*Y + empirical()

  • List of Supported functions for specifying parent-child relationships - here

  • List of Supported functions for specifying noise models here

@bhatt-priyadutt
Copy link
Contributor Author

some tasks like writing test cases, packaging new libraries used, handling for disconnected node is still remaining...

@bhatt-priyadutt bhatt-priyadutt marked this pull request as draft November 7, 2023 19:04
amit-sharma and others added 26 commits November 10, 2023 10:08
@amit-sharma
Fix frontdoor estimation bug (py-why#1060)
395d1fa 
* fixed frontdoor bug

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* fixed formatting issues

Signed-off-by: Amit Sharma <amit_sharma@live.com>

---------

Signed-off-by: Amit Sharma <amit_sharma@live.com>
@bloebp
Add new method to estimate KL divergence using classifier
eb88735 
This should work better with multivariate data and mixed data types. However, it is generally slower than the knn appraoch.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Fix issue with auto assignment with imbalanced classes
b9ae10b 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Fix issue with linear regressor with fixed parameters
57656af 
Before, when creating a linear regressor with fixed parameters, these parameters are overridden when fit to data. Now, the parameters remain fixed.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Revise display of notebook examples in documentation
9d77022 
This change aims at providing a better overview of the notebooks by displaying them as separate cards instead of a card carousel.

Other changes:
- Introductory examples and Real world-inspired examples are now more prominent with individual images and a grid layout by 2 per-row.
- All other examples are now in a grid layout with 3 examples per row.
- Clear outputs of some notebooks.
- Fix issue with rendering counterfactual example notebook.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bhatt-priyadutt
fixed fit error
3f8bf25
@bloebp
Remove 'experimental' disclaimer from GCM modules
cf14caa 
Also slightly change citation hint.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Address issue with timeouts in unit change tests
e846851 
The build sometimes randomly fails due to a timeout issue in the unit tests of the unit change methods of the GCM module. While this only happens in the github builds, this is most likely due to the prallelization of the underlying RandomForestRegressors being fitted.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bhatt-priyadutt
fixed spacing error
4d83d89
@bloebp
Add new GCM model evaluation module
1d3cd9a 
This module adds a new method for evaluating a fitted gcm. Here, we evaluate the performance of causal mechanisms, the underlying modeling assumptions (if possible), the goodness of the generated joint distribution and the graph structure. This utilizes some of the existing methods, but also introduces new ones.

This further adds a new user guide and notebook entries demonstrating the usage.

Part of introducing the module required to make some changes in other modules and implementatins, which are mostly fixes and improvements.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Remove deprecated feature.py module from GCM
601c2ae 
These methods are now available in the feature_relevance.py module.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Fix issue in KL estimation using knn
b2e75a7 
Before, the method threw an error when all samples were equal. However, in these cases, it should rather return a KL divergence of 0.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bhatt-priyadutt
fixed some errors and reformatted and added some validation logic and…
18b0f9b 
… added a test case to start
@bloebp
Fix handling of NaN values in MedianCDFQuantileScorer
ced5d72 
NaN values are now correctly counted when estimating the anomaly score.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Add GCM online shop example notebook
0b8f418 
This is an updated and slightly modified version of the blog post: https://aws.amazon.com/blogs/opensource/root-cause-analysis-with-dowhy-an-open-source-python-library-for-causal-machine-learning/

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@dependabot @bloebp
Bump actions/github-script from 6 to 7
5a6ce23 
Bumps [actions/github-script](https://github.com/actions/github-script) from 6 to 7.
- [Release notes](https://github.com/actions/github-script/releases)
- [Commits](actions/github-script@v6...v7)

---
updated-dependencies:
- dependency-name: actions/github-script
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>
@bloebp
Fix issue with MedianCDFQuantileScorer
cae656a 
Before, the scorer was not able to handle numpy object types directly. However, GCM often uses the object dtype to ensure support of mixing categorical and float values. This fixes the handling of object dtypes by explicitly converting them to floats first.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Fix issue in bar_plot with misspecified confidence intervals
c88cc83 
If the confidence intervals are misspecified, e.g., greater lower bound than upper bound, the method threw an error before. This, however, can sometimes happen due to precision errors in some algorithms and lead to random build fails. This change fixes the issue and ignores invalid intervals accordingly.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Change default parameter of SVC model in the GCM module
bd23532 
Before, the Support Vector Classifier did not produce probabilities, which are required for different algorithms in the GCM module. This changes the 'probability' parameter to True.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Add new example notebook demonstrating the use of the ICC methond in GCM
bd4f95f 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Extend GCM model evaluation by additional metrics
7c015b7 
In addition to CRPS and depending on the node data type, it now also reports the MSE, NMSE, R2 and F1 score.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bhatt-priyadutt
added sanitize and some other logics to cater to disconnected root no…
132cf95 
…de and so on.
@amit-sharma
auto identify the effect modifier columns for `effect' method for Eco…
4fd0a92 
…nML estimators (py-why#1061)

* auto identify the effect modifier columns

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* fixed formatting errors

Signed-off-by: Amit Sharma <amit_sharma@live.com>

---------

Signed-off-by: Amit Sharma <amit_sharma@live.com>
@bloebp @amit-sharma
Proposal: Finalize functional API refactor - deprecate causal graph (p…
2a8e49a 
…y-why#943)

* Deprecate CausalGraph

The effect estimation API is now based on an functional API that expects a networkx graph as input.

- The graph should now be defined via a networkx graph. Most identification methods now expect an additional "observed_nodes" parameter accordingly.
- CausalModel and CausalGraph still exist and should be compatible with the old API.

---------

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
Signed-off-by: Amit Sharma <amit_sharma@live.com>
Co-authored-by: Amit Sharma <amit_sharma@live.com>
@bhatt-priyadutt
added mode checks and validaion and removed compile func
ae9ccd9
@bhatt-priyadutt
added handling for undefined nodes and node redudancy
af61e42
bloebp and others added 29 commits December 1, 2023 06:58
@bloebp
Update
bddbb3a 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Update
031d49e 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Fix documentation box
3475b19 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Add pywhy refrence
591e9d9 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Make example sections more prominent
f455c56 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Revised documentation
9ebf9e3 
- Slightly update and revise existing GCM notebooks
- Moving mediation analysis, direct arrow strength and ICC to their own "Quantify Causal Influence" section
- Adding brief overview to describe differences between the quantification methods
- Change navigation image to reflect newest changes
- Adding related notebooks links to some of the causal task entries
- Adding a direct arrow strength example to the ICC notebook
- Adding a brief overview of the available root cause analysis and explanation methods
- Smaller revision of other GCM entries, such as the basic example
- Smaller typos and missing refernce fixes

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bloebp
Change readme notebook links to compiled HTML versions
918efc6 
Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bhatt-priyadutt
updated extract parent nodes logic
357c092
@amit-sharma
Bug fix: Linear regression CATE estimates were not shown even when ne…
1d050f0 
…ed_conditional_estimates is True (py-why#1092)

* fixed bug where CATE is not returned by lr

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* added test

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* formatted file

Signed-off-by: Amit Sharma <amit_sharma@live.com>

---------

Signed-off-by: Amit Sharma <amit_sharma@live.com>
@amit-sharma
Bug fix for frontdoor identification and a new set of tests (py-why#1093
048117c 
)

* fixed frontdoor bug and added tests

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* updated docstring

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* reformatted file

Signed-off-by: Amit Sharma <amit_sharma@live.com>

---------

Signed-off-by: Amit Sharma <amit_sharma@live.com>
@amit-sharma
Link list of estimators in estimate_method docstring (py-why#1094)
2900fd7 
* linked to up-to-date list of estimators

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* updated docs

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* using absolute paths

Signed-off-by: Amit Sharma <amit_sharma@live.com>

---------

Signed-off-by: Amit Sharma <amit_sharma@live.com>
@amit-sharma
Remove DeepIV econml estimator from notebook and fix flaky test (py-w…
299b1d7 
…hy#1091)

* removed deepiv and updated flaky test

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* black reformattingb

Signed-off-by: Amit Sharma <amit_sharma@live.com>

* removed all outputs from nb

Signed-off-by: Amit Sharma <amit_sharma@live.com>

---------

Signed-off-by: Amit Sharma <amit_sharma@live.com>
@bloebp
Fix issue in falsify method when no tests were performed
1df8561 
It now does not raise a division by zero error anymore. Other changes:
- Add new parameter indicating whether the method requires data for all nodes in the graph or also allows a subset of data.
- If no tests were performed, the summary now returns "Cannot be evaluated".

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>
@bhatt-priyadutt
added unknown mech logic support and did some refactoring
5ccc808
@bhatt-priyadutt
added more test cases
da8141e
@bhatt-priyadutt
added initial graph to equation logic
ebc8dca
@bhatt-priyadutt
corrected some logic and refactored, modularized
3375122
@bhatt-priyadutt
fixed fit error
9e63ed3
@bhatt-priyadutt
fixed spacing error
8aa873d
@bhatt-priyadutt
fixed some errors and reformatted and added some validation logic and…
7b543d7 
… added a test case to start
@bhatt-priyadutt
added sanitize and some other logics to cater to disconnected root no…
4edb3a4 
…de and so on.
@bhatt-priyadutt
added mode checks and validaion and removed compile func
01bc536
@bhatt-priyadutt
added handling for undefined nodes and node redudancy
e38bdd5
@bhatt-priyadutt
updated extract parent nodes logic
1458a05
@bhatt-priyadutt
added unknown mech logic support and did some refactoring
516c14c
@bhatt-priyadutt
added more test cases
4213703
@bhatt-priyadutt
Merge remote-tracking branch 'origin/equations-to-define-causal-graph…
dee6b9a 
…' into equations-to-define-causal-graph

# Conflicts:
#	dowhy/gcm/__init__.py
#	dowhy/gcm/causal_models.py
@bhatt-priyadutt
removed comment
619d894
@bhatt-priyadutt
cmit
7190e30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@bhatt-priyadutt @amit-sharma @bloebp