Cross-Validation Flashcards
1
Q
Cross-Validation
A
How will a strategy perform in the future?
In statistics, cross-validation checks: How does a trained model generalise to an unseen data set?
In our context of trading strategies, we want to know: how well will our strategy perform in the future?
2
Q
In-Sample Test
A
Optimise on one subset (training set).
Investigate the fitness landscape e.g. via a grid search.
Used to choose parameters.
3
Q
Out-of-Sample Test
A
Validate on the other (testing set).
Apply chosen parameters to unseen data.
4
Q
Fitness Landscapes
A
Describes how complex adaptive systems evolve over time through the processes of evolution.
Sometimes, instead of looking at the best in-sample parameters, we should look at the fitness landscape more holistically.
