Code

GH-6723 - AdaBoost#15639

Code

Merged

valenad1 merged 34 commits intomasterh2oai/h2o-3:masterfrom valenad-6723-adaboosth2oai/h2o-3:valenad-6723-adaboostCopy head branch name to clipboardSep 26, 2023

+1,577-14

Commits

Commits on Sep 26, 2023

• GH-6723 - First version of Adaboost with hardcoded weaklearner to DRF - works similarly to scikit on prostate, airlines, and higgs

  valenad1

  committed
  60382e6

• WIP add glm learner

  valenad1

  committed
  11a0e26

• Add probabilities by Obtaining Calibrated Probabilities from Boosting paper

  valenad1

  committed
  2cc9785

• try to fix a test

  valenad1

  committed
  87c13d9

• try to fix test

  valenad1

  committed
  6fa3b5d

• Try to fix the tests

  valenad1

  committed
  89f2d44

• try to fix the tests - this works - pass weights externally each time

  valenad1

  committed
  d23e46c

• Add weights inside of the algorithm - this is working

  valenad1

  committed
  4c768fb

• fix for GLM

  valenad1

  committed
  40e6285

• GH-6723 - add unit test to inner tasks

  valenad1

  committed
  0fd0a43

• use test files again in the large tests

  valenad1

  committed
  109b348

• Improve basic training test to look into structure of weak learners

  valenad1

  committed
  f994b1e

• clenup - remove isAdaBoost

  valenad1

  committed
  3c07b25

• add simple model summary

  valenad1

  committed
  98191ba

• fix java api

  valenad1

  committed
  589033d

• Implement possibility to have a custom weights column and ensure that the created weights collumn will always be the one we use

  valenad1

  committed
  b413535

• Add categorical test

  valenad1

  committed
  6077424

• Cleanup that didn't change tests results

  valenad1

  committed
  7a65eb7

• Remove toCSV since there is API already and commented code

  valenad1

  committed
  e0d4845

• Refactor learning rate to learn rate

  valenad1

  committed
  f84d9cb

• Add documentation and validation to parameters

  valenad1

  committed
  e3a0309

• Add documentation to AdaBoost class

  valenad1

  committed
  5345b9b

• add log

  valenad1

  committed
  1840cab

• Refactor AdaBoost - simple refactor

  valenad1

  committed
  0485715

• Fix GLM as a weak learner

  valenad1

  committed
  4810d4f

• Add GBM as a weak learner

  valenad1

  committed
  40bca78

• test cleanup

  valenad1

  committed
  32d96b2

• Refactor n_estimators to nlearners

  valenad1

  committed
  8574bd2

• fixup! Implement possibility to have a custom weights column and ensure that the created weights collumn will always be the one we use

  valenad1

  committed
  bf7a435

• Fix for different model as a weak learner - use upperclass instead of DRF

  valenad1

  committed
  445dc39

• Ensure that adaboost create exactly nlearners models

  valenad1

  committed
  4ec9229

• Refactoring according to suggestions

  valenad1

  committed
  e52c287

• fixup! Fix for different model as a weak learner - use upperclass instead of DRF

  valenad1

  committed
  aeaa11b

• GH-6723 AdaBoost API (#15732)

  Show description for 5b2b780
  valenad1

  authored
  5b2b780