Model Selection Flashcards

1
Q

Occams razor

Training Error / Generalization Error

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Statistical Learning Theory

Learning Machines

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Statistical Learning Theory

MSE

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Bias Variance Tradeoff

A
  • underfitting:
    • high bias
    • low variance
  • overfitting:
    • low bias
    • high variance
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

James Stein paradox

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

No free lunch theorem

A
  • Occams razor (simples model / most intuitive method is preferable):
    • choice of model depends on scenario
  • utilitaristic approach (lowest prediction error is preferable):
    • without knowledge of the scenario, its impiossible to predict the prediction error
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Methods for model selection

A
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

k-fold cross validation

A
  • never use the test set!
How well did you know this?
1
Not at all
2
3
4
5
Perfectly