Skip to content

SelectionΒΆ

Last Updated: 3 months ago2024-12-26 ; Contributors: AhmedThahir, web-flow

Note: Make sure to correct for multiple hypothesis testing

Model SelectionΒΆ

  1. Fit multiple models gig_i on the training data and eyeball dev data
  2. Use dev data for hyper parameter tuning of each model gig_i
  3. Use external validation data for model selection and obtain gβˆ—g^*
  4. Combine the training and validation data. Refit gβˆ—g^* on this set to obtain gβˆ—βˆ—g^{**}
  5. Assess the performance of gβˆ—βˆ—g^{**} on the test data

Finally, train gβˆ—βˆ—g^{**} on the entire data to obtain f^\hat f

Check the results on the self-hosted competition

Comments