Skip to content

My notes about possible testing of models for overfitting

Notifications You must be signed in to change notification settings

Alexander-Minyushkin/ml_cheat_sheet

 
 

Repository files navigation

#Jitter test for overfitting

Jitter_Test_for_Overfitting.ipynb notebook demonstrate some intuition about possible testing of models for overfitting.

Sample models are taken from "Supervised learning superstitions cheat sheet"

  • Logistic regression
  • Decision trees
  • Support vector machines

Trained on various versions of the "two moons" dataset and plot empirical decision boundaries.

JTO applied to regression model. Comparison to Cross Validation added.

TL/DR : JOT gives small improvement over Cross Validation. Computational speed of JOT is slower than CV.

About

My notes about possible testing of models for overfitting

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published