common tests: ``test_get_params_invariance`` invariance should be part of _yield_all_tests

[amueller]

I think the test_get_params_invariance test should be removed and the check_get_params_invariance should be added to _yield_all_tests in estimator_checks.py.
This is less code and will also make this test run as part of check_estimator.

We might want to do something similar with the n_iter tests above.

[JungeAlexander]

Looks like a feasible first contribution, I'd like to look into this.

[jnothman]

go for it

[JungeAlexander]

Before looking into moving the *_n_iter tests, could someone please have a look at my commit linked above and tell me if I am roughly on the right track? Especially regarding the seemingly redundant test for the get_params attribute that I removed as mentioned in the commit message.

[jnothman]

I suppose that elsewhere we practically require estimators to have get_params, so I suppose that's fine. Your change LGTM.

[JungeAlexander]

Great, thank you for the feedback! I'll continue here.

[JungeAlexander]

Commit above moves check_transformer_n_iter and I will soon continue with moving check_non_transformer_estimators_n_iter.

[amueller]

thanks @JungeAlexander. You can already create a pull request and keep iterating on that branch.

[JungeAlexander]

Currently some tests fail. Some hints:

AttributeError: 'Isomap' object has no attribute 'n_iter_'
AttributeError: 'KernelPCA' object has no attribute 'n_iter_'
AttributeError: 'LocallyLinearEmbedding' object has no attribute 'n_iter_'
ValueError: For mono-task outputs, use ElasticNet (MultiTaskElasticNet)
ValueError: For mono-task outputs, use ElasticNetCV (MultiTaskElasticNetCV)
ValueError: For mono-task outputs, use ElasticNet (MultiTaskLasso)
ValueError: For mono-task outputs, use LassoCV (MultiTaskLassoCV)
AttributeError: 'OneClassSVM' object has no attribute 'n_iter_'

Will look into this later. Please let me know if you have a better idea than skipping the test for specifically those estimators where it fails right now