[add] bias-var trade-off in resampling: make clearer / more explicit - Githubissues

slds-lmu / lecture_i2ml

I2ML lecture repository

https://slds-lmu.github.io/i2ml/

Creative Commons Attribution 4.0 International

147 stars 81 forks source link

[add] bias-var trade-off in resampling: make clearer / more explicit #1202

Open lisa-wm opened 8 months ago

lisa-wm commented 8 months ago

:heavy_plus_sign: :heavy_plus_sign: :heavy_plus_sign:

:warning: Please provide specific & actionable issue titles!

Chapter name / context

Performance evaluation, resampling

Description

Conflicting goals: large train sets (for low bias) and large test sets (for low variance) ➡️ make crystal-clear why we need resampling and how we balance conflicting goals; and what we average over when aggregating over folds / repeating CV
Explain: what's the problem with estimating variance in CV?
Explain: what's the problem with LOO-CV?

Additional sources

...