scikit-learn-intelex
scikit-learn-intelex copied to clipboard
[enhancement] Add IncrementalEmpiricalCovariance to patch_map
Description
IncrementalEmpiricalCovariance update to match sklearnex estimator characteristics
Changes proposed in this pull request:
- Add mahalanobis method using sklearnex pairwise_distances to IncrementalEmpiricalCovariance
- Add error_norm
- Add onedal_cpu_support and onedal_gpu_support to IncrementalEmpiricalCovariance
- Add BaseEstimator to IncrementalEmpiricalCovariance inheritance
- Add wrap_output_data where necessary in IncrementalEmpiricalCovariance
- Fix covariance_ and location_ attributes to report as not existing before fit
- Add n_features_in_ attribute for sklearn conformity
- Add n_samples_seen_ attribute to match style of IncrementalPCA
- Add init keywords to match sklearn EmpiricalCovariance
- Add store_precision functionality
- Add assume_centered functionality
- Add relevant sklearn tests to sklearnex tests via a separate monkeypatch
Tasks
- [x] Implement changes for sklearnex/sklearn conformance
- [x] Add to patchmap to access sklearn tests
- [x] Modify patchmap to work with new estimators
- [x] Add EmpiricalCovariance tests from sklearn to test_covariance for IncrementalEmpiricalCovariance
- [x] Pass Tests
- [x] Split off changes to patchmap and tests into separate PR
/Intelci: run
/intelci: run
/intelci: run
/intelci: run
Will write tickets associated with dataframe0 and test memory leak, it is gpu and cpu side meaning its likely in the glue portion of the code. Also, it only occurs on one of the two meaning something else fishy is going on. This will have to be revisited in the future. I will now slim this PR down and remove the content from the other 3. #1677 #1678 #1676
/intelci: run
Code has now been reduced to just changes in IncrementalEmpiricalCovariance
/intelci: run
/intelci: run
Unrelated issue in private CI (logistic regression convergence on cpu)
/intelci: run
/intelci: run
/intelci: run
If its possible, please wait this pr, for avoid python computation for assume_centered
flag support https://github.com/oneapi-src/oneDAL/pull/2658
/intelci: run
/intelci: run
/intelci: run
/intelci: run
/intelci: run
/intelci: run
/intelci: run
/intelci: run
Only Issue is a errant PCA test case failure.
/intelci: run
/intelci: run
Rerun due to openblas issues: http://intel-ci.intel.com/ef02ba38-9fbd-f127-9fb6-a4bf010d0e2e
/intelci: run
/intelci: run