H2O Generalized Linear Model Learner (Regression) - H2O Model Port to variable or table

KNIME Analytics Platform
1

image

image
image959×358 7.7 KB

I am trying to automatically access the selected lambda (found via lambda search in the H2O Generalized Linear Regression Learner Regression node). It outputs in the H2O Model Port, however I am having trouble automatically accessing it through another node. I was hoping Model to Cell would work, but unfortunately that only outputs the port object: Generalized Linear Modeling.

Does anyone know how to automatically access the selected lambda (5.07E-4 pictured above in the regularization row)?

Thanks!

2

Hello @elliegreib ,

I’ve spoken with our engineers here at KNIME and unfortunately that is not possible. Sorry for the inconvenience.

Please tag my name and ask a follow up question if there is anything else I can do to help.

Best,
Victor

3

Hi @victor_palacios,

Could you elaborate why it is not possible to get numerical outputs in a table instead of an image?

Thanks,
Robin

4

@victor_palacios - this seems like this should be available in a future enhancement.

The current node calculates a number that would be useful to have available for the automation within the workflow. What needs to be done to create a ticket to look into creating this enhancement?

Kevin

5

@jichaohe @jichaohe @elliegreib @kevweath well actually there is a way to do this. One could employ a Python node and H2O.ai in the background to extract a buch of parameters and also the summary from the MOJO model file. Though admittedly not the most elegant way you could have the information in a file :slight_smile:

image
image1920×1113 155 KB

kn_example_ml_regression_h2o_glm_parameter.knwf (222.2 KB)

If you are interested you might also check out these article and workflows (I plan on having a walkthrough on some H2O.ai AutoML code also …):

6

Hi @kevweath! Hope things are going well.

Victor is out for a few days, but I’ll see what I can do about getting a ticket created for this. Thanks for bringing it to our attention!

(And thanks @mlauber71 for the workaround!)

EDIT: Feature request ticket AP-20504 created (thanks @victor_palacios !)

7

Thank you, @mlauber71 ! I learned a few things from your solution here.

We are trying to avoid messing with Python for this project due to its users. For h2o Python API, I know the following 1-line code can export the lambda easily:

H2OGeneralizedLinearEstimator.getLambdaBest(model_obj)

We are hoping to see it being exported natively in KNIME.