That could be applied to any meta-estimator that uses a base estimator, right?

Yes, it could be. I didn't have time when I wrote this issue to check the applicability to other ensembles.

Updated title and description

@jnothman I think that we have two options.

• Validate the input early as it is now and introduce a new parameter check_input in fit, predict, etc with default vaule True in order to preserve the current behavior. The check_input could be in the constrcutor.
• Relax the validation in the ensemble and let base estimator to handle the validation.

What do you think? I'll sent a PR.

IMO assuming the base estimator manages validation is fine.

Is this still open ? can I work on it?

@Chaitya62 I didn't have the time to work on this. So, go ahead.

@chkoar onit!

After reading code for 2 days and trying to understand what actually needs to be changed I figured out that in that a call to check_X_y is being made which is forcing X to be 2d now for the patch should I do what @chkoar suggested ?

Cool I ll submit a PR soon

@Chaitya62, Let me inform if you are not working on this anymore. I want to work on this.

@devanshdalal I am working on it have a minor issue which I hope I ll soon solve

@Chaitya62 Are you still working on this?

@dalmia go ahead work on it I am not able to test my code properly

@Chaitya62 Thanks!

I'd like to work on this, if that's ok

As a first step, I tried looking at the behavior of meta-estimators when passed a 3D tensor. Looks like almost all meta-estimators which accept a base estimator fail :

>>> pytest -sx -k 'test_meta_estimators' sklearn/tests/test_common.py
<....>
AdaBoostClassifier raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data
AdaBoostRegressor raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data
BaggingClassifier raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data
BaggingRegressor raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data
ExtraTreesClassifier raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data
ExtraTreesRegressor raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data

Skipping GradientBoostingClassifier - 'base_estimator' key not supported
Skipping GradientBoostingRegressor - 'base_estimator' key not supported
IsolationForest raised error 'default contamination parameter 0.1 will change in version 0.22 to "auto". This will change the predict method behavior.' when parsing data   

RANSACRegressor raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data                                     
RandomForestClassifier raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data
RandomForestRegressor raised error 'Found array with dim 3. Estimator expected <= 2.' when parsing data

@jnothman @amueller considering this, should this be a WONTFIX, or should all the meta-estimators be fixed?

@jnothman Adaboost tests are testing the sparsity of the X. This means that we should skip these tests in order to relax the validation, right?

@jnothman Could you please tell me the solution you landed on? I am working with CNNs as my weak classifiers for which the input features need to be 3D. I still face the same issue as referenced in #7767.

Please give code that shows the error you encounter

I am having the error in the following code snippet:
boosted_cnn = AdaBoostClassifier(base_estimator= model(), n_estimators=20)
boosted_audio.fit(X,Y)

shape of X (a,100,1) and Y (a, b) where b is the number of classes and a is the number of examples
fit gives be the following error:
*** ValueError: Found array with dim 3. Estimator expected <= 2.

@jilt-sebastian This functionality hasn't released yet. You probably need to install scikit-learn from master.

Cool. Thank you