Examples for SWING LASSO, and PLSR plus aggregating results from those examples

bagherilab / SWING

Sliding Window Inference for Network Generation

10 stars 12 forks source link

Examples for SWING LASSO, and PLSR plus aggregating results from those examples #22

Open jiawu opened 6 years ago

lujonathanh commented 4 years ago

Hello! I just wanted to follow up on the SWING LASSO-- is there an example with result aggregation? I tried modifying the SWING RF example notebook but it still is not working.

In addition, is it known what alpha parameters were used for SWING LASSO?

Thank you so much for your help!!

justinfinkle commented 4 years ago

Hi @lujonathanh. We've only posted the example using SWING-RF, but it should run the same with LASSO as long as you change the window type and set any LASSO specific parameters.

Are you interested in the alpha parameters for LASSO that we used in the paper? I wouldn't expect them to be useful with other data. We provided two "convenience" methods for estimating alphas:

get_null_alpha - will find the smallest alpha value that produces all zeros. This can be a good starting place to for a forward-selection type alpha search
cv_select_alpha - will use cross-validation to select the optimal alpha value. At the time, I don't think scikit-learn had implemented a LassoCV, but they do now. The code could probably be updated to use that.

lujonathanh commented 4 years ago

Hi @justinfinkle ,

Thank you for your quick and helpful response! That makes sense. After playing around a bit I got it to work.

Gratefully,

Jonathan

justinfinkle commented 4 years ago

Excellent! Feel free to submit a PR for a working solution if you want, I'd be happy to review.

lujonathanh commented 4 years ago

Thank you! I don't think a Pull Request would be best since my directory is set up pretty differently. Is it okay if I email you the files I used instead?

jiawu commented 4 years ago

Hi Jonathan,

Not sure if you got a reply to this yet, but feel free to email the files with your questions.

Best, Jia

On Thu, Nov 14, 2019 at 2:12 PM Jonathan Lu notifications@github.com wrote:

Thank you! I don't think a Pull Request would be best since my directory is set up pretty differently. Is it okay if I email you the files I used instead?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bagherilab/SWING/issues/22?email_source=notifications&email_token=AAVVEGXE6LSDIJRSCHCQHL3QTWPLJA5CNFSM4ERZO3D2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEEC6U2I#issuecomment-554035817, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAVVEGVGY3RG53PJP6XUQSDQTWPLJANCNFSM4ERZO3DQ .

-- PhD Student in IBiS, Northwestern University Bagheri and Miller Lab jiawu@u.northwestern.edu | 408-386-6907 <(408)%20386-6907>