| | |
Summary: E. Alpaydin AERFAISS 2010 2
Introduction
Questions:
Is the error rate of my classifier less than 2%?
Is k-NN more accurate than MLP?
Does having PCA before improve accuracy?
Which kernel leads to highest accuracy with SVM?
E. Alpaydin AERFAISS 2010 3
Material
Training/validation/test sets
Resampling methods
Comparing multiple algorithms on a single data set
Comparison on multiple data sets
E. Alpaydin AERFAISS 2010 4
Algorithm Preference
Criteria (Application-dependent):
Misclassification error, or risk (loss functions)
Training time/space complexity
Testing time/space complexity
Interpretability
|
